Game program, game system, and game control method

Abstract

This provides a new way to utilize non-player characters, which are controlled based on past battle data. [Solution] The computer is configured to function as: a player character control means for controlling a player character to perform actions in response to user operations; a situation acquisition means for acquiring the battle situation during an AI battle between an AI character (a non-player character) and a player character; an AI character control means for controlling an AI character to perform actions output from a trained model by inputting the acquired battle situation to a trained model that has undergone machine learning to acquire character actions appropriate to the battle situation using battle data from the battle game as training data; and an evaluation means for evaluating the user's operations in an AI battle and outputting evaluation results that include tasks that the user should accomplish related to the operations.

Description

Technical Field

[0001] The present invention relates to a game program, a game system, and a game control method.

Background Art

[0002] Patent Document 1 discloses a game program that operates a non-player character based on past battle data of a player character operated by a user or another user in order to make the operation of the non-player character closer to the operation of a character operated by a human.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] Patent Document 1 does not describe utilization methods of non-player characters operated based on past battle data other than having them battle against player characters.

[0005] Therefore, an object of the present invention is to provide a game program, a game system, and a game control method that can provide a new utilization method for non-player characters operated based on past battle data.

Means for Solving the Problems

[0006] A game program according to one aspect of the present invention is a game program that causes a computer to execute a competitive game in which characters compete against each other in a game space, wherein the computer functions as: a player character control means for controlling a player character to perform actions in response to user operations; a situation acquisition means for acquiring the battle situation during an AI battle between an AI character, which is a non-player character, and the player character; an AI character control means for controlling the AI ​​character to perform actions output from a trained model by inputting the acquired battle situation to a trained model that has been trained using the battle data of the competitive game as training data to acquire character actions appropriate to the battle situation; and an evaluation means for evaluating the user's operations in the AI ​​battle and outputting an evaluation result that includes tasks that the user should accomplish related to the operations.

[0007] According to this game program, AI battles closely resemble human-versus-human matches, where a player character controlled by the user competes against a player character controlled by another user. Therefore, the evaluation system assesses the user's actions during AI battles and outputs tasks related to those actions that the user should accomplish. This allows the user to gain valuable skills for human-versus-human matches through AI battles. Consequently, this provides a new way to utilize non-player characters that operate based on past battle data.

[0008] The evaluation means may determine whether the task has been achieved in the AI ​​battle after outputting the task, and output the determination result regarding the achievement of the task.

[0009] The evaluation means may calculate the frequency of occurrence of a predetermined battle situation associated with each of the predetermined evaluation items during the AI ​​battle, and based on the frequency of occurrence, select an evaluation item from the plurality of evaluation items and output predetermined evaluation comment information associated with the selected evaluation item.

[0010] The evaluation means may include a first evaluation means that outputs a first evaluation result which is the evaluation result of evaluating the user's actions in the most recent AI match, and a second evaluation means that outputs a second evaluation result which is the evaluation result of evaluating the user's actions in a plurality of past AI matches.

[0011] The competitive game may include a plurality of game modes selectable by the user, and the computer may further function as a game mode suggestion means that determines a game status indicating at least one of the user's play status in the competitive game and the status of each of the game modes, and suggests a game mode to the user from among the plurality of game modes based on the determined game status.

[0012] The computer may further function as a mode selection means that, in response to user operation, selects a spectator mode as the game mode of the battle game for watching a battle between a first AI character and a second AI character, which are non-player characters. The situation acquisition means acquires the battle situation during the battle between the first AI character and the second AI character when the spectator mode is selected by the mode selection means. The AI ​​character control means controls the first AI character to perform an action output from the first trained model by inputting the acquired battle situation to a first trained model, which has been machine-trained using the battle data relating to the first AI character's battle as training data, when the spectator mode is selected by the mode selection means. The second AI character may also be controlled to perform an action output from the second trained model by inputting the acquired battle situation to a second trained model, which has been machine-trained using the battle data relating to the second AI character's battle as training data.

[0013] The computer may further function as a storage means for storing battle data related to a battle when a battle takes place between characters in the battle game, and as a model creation means for creating a trained model. The model creation means may set model creation conditions according to user operations, select battle data that satisfies the model creation conditions from the stored battle data, and create the trained model by performing machine learning using the selected battle data as training data.

[0014] The model creation means may change the number of match data used for machine learning, depending on the user's selection.

[0015] The model creation means may estimate the time required to create the trained model according to the number of match data used to create the trained model, and output the estimation result.

[0016] The model creation means may output warning information if it determines that the number of match data that satisfy the model creation conditions does not reach a predetermined number.

[0017] A game system according to one aspect of the present invention comprises a program storage unit that stores the above-mentioned game program, and a computer that executes the program stored in the program storage unit.

[0018] A game control method according to an aspect of the present invention is a game control method for controlling a battle game in which characters battle against each other in a game space on a computer. The game control method includes a player character control step of controlling a player character to perform an operation according to a user's operation, a situation acquisition step of acquiring the battle situation during an AI battle, which is a battle between an AI character, which is a non-player character, and the player character, a machine learning for obtaining an operation of a character according to the battle situation by using the battle data of the battle game as learning data, and an AI character control step of controlling the AI character to perform the operation output from the learned model by inputting the acquired battle situation to the learned model, and an evaluation step of evaluating the operation content of the user in the AI battle and outputting an evaluation result including tasks that the user related to the operation content should achieve.

Effect of the Invention

[0019] According to an aspect of the present invention, there are provided a game program, a game system, and a game control method capable of providing a new utilization method of a non-player character that operates based on past battle data.

Brief Description of the Drawings

[0020] [Figure 1] It is a block diagram showing a game system in one embodiment. [Figure 2] It is a diagram showing an example of a battle screen. [Figure 3] It is a diagram showing an example of a mode selection screen. [Figure 4] It is a block diagram showing a functional configuration of a game system. [Figure 5] It is a flowchart showing an example of a flow of online battle processing. [Figure 6] It is a flowchart showing an example of a flow of AI battle related processing. [Figure 7] It is a diagram showing an example of a creation condition setting screen. [Figure 8]It is a diagram showing an example of a creation method selection screen. [Figure 9] It is a diagram showing an example of a result screen. [Figure 10] It is a table showing an example of the correspondence relationship among evaluation items, count conditions, praise messages, and problem messages. [Figure 11] It is a diagram showing an example of an AI character selection screen. [Figure 12] It is a diagram showing an example of a request screen. [Figure 13] It is a flowchart showing an example of the flow of the second evaluation process. [Figure 14] It is a flowchart showing an example of the flow of the mode proposal process.

Mode for Carrying Out the Invention

[0021] <Embodiment> Hereinafter, a game system and a game program according to an embodiment of the present invention will be described with reference to the drawings.

[0022] [Hardware Configuration] First, the configuration of a game system 1 according to an embodiment of the present invention will be described. FIG. 1 is a block diagram showing the hardware configuration of the game system 1. As shown in FIG. 1, the game system 1 includes a plurality of game devices 2 and a server device 4 that can be communicatively connected to each of these plurality of game devices 2 via a communication network NW.

[0023] The game device 2 includes a game device main body 20, one or more controllers 31 (which may also be referred to as an operation unit) connected to the game device main body 20 by wire or wirelessly, a display 32 (which may also be referred to as a display unit or an output unit), and a speaker 33 (which may also be referred to as an acoustic output unit or an output unit).

[0024] The game device body 20 includes a control unit 21, which is a computer that controls its operation. The control unit 21 includes a processor, such as a CPU (Central Processing Unit). The control unit 21 is connected to a media I / F unit 22, a storage unit 23, a network I / F unit 24 (which may also be called a communication unit), a wireless communication control unit 25, a graphics processing unit 26, and an audio processing unit 27, each via a bus 21a.

[0025] The media I / F unit 22 is an interface into which a game medium 34 for running a game can be loaded. The game medium 34 is a disc-type recording medium such as a DVD-ROM, and contains the game program 34a and game data 34b necessary for running the game. This game data 34b includes various data necessary for running the game.

[0026] The memory unit 23 is composed of, for example, a large-capacity recording medium such as an HDD, a semiconductor memory such as a mask ROM or PROM, and RAM consisting of DRAM or SRAM. The memory unit 23 stores game programs 34a and game data 34b read from the game media 34, as well as save data, etc. Note that some or all of the game programs 34a and game data 34b stored in the memory unit 23 do not have to be read from the game media 34, but may be downloaded from a server device 4 or the like via the internet.

[0027] The network interface unit 24 is an interface that connects the game device 2 to a communication network NW such as the Internet or a LAN. The game device 2 can communicate with other game devices 2 and server devices 4 via the communication network NW. For example, when the game device 2 is playing a game while communicating with the outside world, it sends and receives data to and from other game devices 2 and / or server devices 4 via the communication network NW.

[0028] The wireless communication control unit 25 is wirelessly connected to the controller 31 attached to the game device main unit 20, enabling data transmission and reception between the controller 31 and the control unit 25. The controller 31 is an operating device that accepts user input. The controller 31 has multiple control elements. When the controller 31 is operated by the user, operation information indicating the content of that operation is sent from the controller 31 to the control unit 21 of the game device 2 via the wireless communication control unit 25. The control unit 21 proceeds with the game based on the operation information from the controller 31.

[0029] The graphics processing unit 26 renders game images in video format according to instructions from the control unit 21. The game images rendered by the graphics processing unit 26 are displayed on the display 32 as a game screen. The display 32 is a display device that displays game images. The display 32 is, for example, a liquid crystal display.

[0030] The audio processing unit 27 plays and synthesizes digital game audio according to instructions from the control unit 21. The audio processing unit 27 is also connected to an external speaker 33. The game audio played and synthesized by the audio processing unit 27 is output to the outside through the speaker 33, which acts as an audio output unit. For example, the speaker 33 could be a stationary speaker or a wearable speaker such as headphones.

[0031] The server device 4 includes a control unit 41, which is a computer that controls its operation. The control unit 41 includes a processor, such as a CPU (Central Processing Unit). A storage unit 42 and a network interface unit 43 are connected to the control unit 41 via a bus 41a. The server device 4 may also include input devices such as a keyboard and output devices such as a liquid crystal display, as needed.

[0032] The storage unit 42 consists of, for example, a large-capacity recording medium such as an HDD, a semiconductor memory such as a mask ROM or PROM, and RAM such as DRAM or SRAM. The storage unit 42 stores the game program 42a and game data 42b. The network I / F unit 43 is an interface that connects the server device 4 to a communication network NW such as the Internet or a LAN. The server device 4 can communicate with multiple game devices 2 via the communication network NW.

[0033] [Game Overview] Next, an example of a competitive game implemented by the game system 1 in this embodiment will be described. The competitive game described below is a fighting game in which a user controls one character and engages in a one-on-one battle with another character in a virtual game space.

[0034] Figure 2 shows an example of the battle screen D1. The battle screen D1 includes an image of the 3D game space S captured by a virtual camera. Two characters P1 and P2 are placed in the virtual game space S, and the two characters P1 and P2 battle each other.

[0035] The battle screen D1 contains various icons. For example, at the top of the battle screen D1, two health gauges G1 and G2 are placed with a gap between them horizontally. The two health gauges G1 and G2 represent the health values ​​of two characters P1 and P2, respectively. When an attack from one of the two characters P1 or P2 hits the other character, the other character's remaining health decreases by the amount of damage corresponding to that attack.

[0036] Additionally, a time display area G3 is positioned between the two health gauges G1 and G2 on the battle screen D1. The time limit is displayed in the time display area G3. Furthermore, at the bottom of the battle screen D1, two technique gauges G4 and G5 are positioned with a gap between them horizontally. The two technique gauges G4 and G5 each represent the technique points consumed when a specific technique (for example, a special technique described later) unique to the two characters P1 and P2 is activated. Of the two characters P1 and P2, the one whose health value reaches zero within the time limit, or the one with less health value when the time limit reaches zero, becomes the loser of the match, and the other becomes the winner.

[0037] This game offers several game modes. For example, in one versus mode, one of two characters, P1 and P2, is the player character controlled by the user, while the other character is the opponent of the user's character. For the sake of explanation, the player character controlled by the user may be referred to as the "user character," and the character fighting against that user character may be referred to as the "opponent character." Furthermore, any character among the user character and the opponent character may simply be referred to as "character."

[0038] The actions of the player character are determined by the user's input to the controller 31. Specifically, each input on the controller 31 is assigned a predetermined action for the player character. For example, several buttons on the controller 31 are assigned attack actions such as "weak punch," "strong punch," "weak kick," and "strong kick." For example, several directional keys on the controller 31 are assigned actions such as "move to the right," "move to the left," "jump," and "crouch."

[0039] In this game, a predetermined sequence of actions (hereinafter referred to as "special actions" or "special techniques") by the player's character is activated by a predetermined sequence of actions (such as pressing multiple buttons simultaneously, pressing a predetermined sequence of buttons in a predetermined order within a predetermined time, or a combination thereof) performed on the controller 31. In this specification, a sequence of actions performed by the user on the controller 31 may be referred to as a "command." Special actions may be attack techniques specific to the character.

[0040] Furthermore, in this game, users can select the type of character they will use in battle from several different character types. Special moves vary depending on the character type. In this game, the command inputs required to activate special moves also vary depending on the character type.

[0041] Furthermore, in this game, users can choose between a standard control type and a simplified control type for activating special moves, i.e., the commands used. When the standard control type is selected, activating a special move for your character requires a relatively complex series of operations. However, when the simplified control type is selected, special moves can be activated with simpler operations compared to when the standard control type is selected. Therefore, even users unfamiliar with fighting games can activate special moves without complex inputs (i.e., complex command inputs) by selecting the simplified control type and starting a match. However, the amount of damage dealt by a special move when the simplified control type is selected is less than the amount of damage dealt when the standard control type is selected. In addition, the simplified control type has limitations on the types of moves that can be used; some moves that can be used in the standard control type cannot be used in the simplified control type.

[0042] On the other hand, the method for determining the opponent character's actions differs depending on the game mode the user is playing. Figure 3 shows an example of the mode selection screen D2. The user can select which game mode to play by manipulating options B1, B2, B3, B4, B5, B6, etc. on the mode selection screen D2 via the controller 31. The game modes in this game include player vs. player mode, CPU vs. player mode, and AI vs. player mode.

[0043] The player-versus-player (PvP) mode is a game mode in which a user competes against another user. In other words, in PvP mode, the opponent's character is the player's character, and the opponent's character's actions are determined by the other user's input to the controller 31. Furthermore, PvP mode is classified into various types, such as online PvP mode and offline PvP mode.

[0044] For example, online battle mode is a game mode in which users compete against other users online via the internet. More specifically, in online battle mode, users access server device 4 from game device 2 via a communication network NW and compete against each other. When a user selects online battle mode as the game mode, game device 2 synchronizes with other users' game devices 2 through a well-known matching process, and the character controlled by the user and the character controlled by the other user battle each other.

[0045] Furthermore, the online battle mode is divided into ranked match mode and casual match mode. Ranked match mode is a battle mode in which rank points change depending on the results of the battle. Rank points are a score that represents the rating of the user's skill level in this game. Rank points increase by winning battles in designated battle modes such as ranked match mode, and decrease by losing battles. This game has a ranking system, and each user is placed in a rank that represents the degree of their skill level according to the rank points assigned to them, and they can be promoted to a higher rank or demoted to a lower rank depending on the increase or decrease in rank points. In other words, rank points and rank are indicators of the user's strength in this game. Note that "assigned to a user" means that it is stored in association with the user's identification information.

[0046] Casual Match mode is a battle mode where rank points do not change based on the outcome of the match. In Casual Match mode, rank points do not decrease even if you lose a match, so users can enjoy playing matches more casually compared to Ranked Match mode.

[0047] For example, the offline versus mode is a game mode in which two users compete against each other offline without using the internet.

[0048] CPU Versus Mode is a game mode where the user plays against the computer. In other words, in CPU Versus Mode, the opponent character is a non-player character, and the opponent character's actions are determined according to a predetermined algorithm. Furthermore, CPU Versus Mode is classified into various types, such as VS Mode and Story Mode. For example, in VS Mode, the user can freely specify the type of character they want to play against. In Story Mode, the story unfolds according to the type of character the user has selected, and the types of characters the user will play against are determined sequentially according to the story.

[0049] In AI battle mode, the opponent character is a non-player character, and the opponent character's actions are determined by a trained model created through machine learning. Hereafter, non-player characters whose actions are determined by a trained model will be referred to as AI characters. AI characters are the same as non-player characters in CPU battles in that their actions are automatically controlled by the CPU, but in this specification, AI characters are distinguished from non-player characters in CPU battle mode in that their actions are determined by a trained model.

[0050] The trained model that outputs the actions of the AI ​​character is created by machine learning using battle data from one or more users as training data. The method for collecting battle data to be used as training data will be explained later, along with the online battle processing.

[0051] Furthermore, the AI ​​battle mode is classified into Self-AI Battle Mode, Ranked AI Battle Mode, and AI Spectator Mode. The differences between Self-AI Battle Mode, Ranked AI Battle Mode, and AI Spectator Mode will be explained later, along with the AI ​​battle-related processing described below.

[0052] In addition, in this game, the evaluation results of the user's actions during AI battles are displayed on display 32. Details of the evaluation of user actions will be explained later along with the first and second evaluation processes.

[0053] Furthermore, this game suggests game modes that are suitable for the user. Details about game mode suggestions will be explained later along with the mode suggestion process.

[0054] [Functional configuration] Figure 4 is a block diagram showing the functional configuration of the game system 1. As shown in Figure 4, the control unit 5 (computer) of the game system 1, which is formed by the cooperation of the control unit 21 of the game device 2 and the control unit 41 of the server device 4, functions as various means 51, 52, 53, 54, 55, 56, 57, 58, 61, and 62 by executing the game programs 34a and 42a.

[0055] (Functional configuration of the control unit of the game device) Specifically, the control unit 21 of the game device 2 functions as a game progression means 51, a mode selection means 52, a player character control means 53, a situation acquisition means 54, a non-player character control means 55, an evaluation means 56, a game mode presentation means 57, and a model creation means 58 by executing the game program 34a. In this specification, "player character" may be referred to as "PC," and "non-player character" may be referred to as "NPC."

[0056] The game progression means 51 performs various processes to advance the game. For example, the game progression means 51 communicates with other game devices 2 and server devices 4 via a communication network NW. For example, the game progression means 51 generates a two-dimensional or three-dimensional virtual game space S and places the aforementioned characters P1 and P2 in the game space S.

[0057] The mode selection means 52 selects one game mode from among several game modes in response to user operation.

[0058] The PC control means 53 controls the player character to perform actions in response to user operations. Specifically, the PC control means 53 controls the actions of its own character placed in the game space S based on operation information generated by the user's operations on the controller 31. Furthermore, in player-versus-player matches, such as online matches, the PC control means 53 controls the actions of the opponent's character placed in the game space S based on operation information received from the other user's game device 2 (operation information generated by the other user's operations on the controller 31 on the other game device 2).

[0059] The situation acquisition means 54 acquires the battle status between characters during a match. For example, the situation acquisition means 54 acquires the battle status during an AI match, which is a match between an AI character (a non-player character) and a player character.

[0060] Examples of the battle situation include various parameters of characters P1 and P2 (for example, the character's health value, the technique points consumed when using special techniques, etc.), the elapsed time of the battle or the remaining time until the set time limit for the battle, the type of action of one or both characters P1 and P2 (type of attack, type of defense, jump, whether or not they are moving, direction of movement, speed of movement, posture, type of technique being performed or performed, etc.), the relative position of one character to the other (distance between characters P1 and P2, etc.), and the absolute position of one or both characters P1 and P2 (the position of characters P1 and P2 in game space S).

[0061] The NPC control means 55 controls non-player characters. The NPC control means 55 includes standard character control means 55a and AI character control means 55b.

[0062] The standard character control means 55a controls the non-player character to perform actions determined according to a predetermined algorithm. For example, during a CPU match, the standard character control means 55a controls the actions of the non-player character that is the opponent of the player character in the CPU match.

[0063] The AI ​​character control means 55b controls the actions of the AI ​​character, which is a non-player character. Specifically, the AI ​​character control means 55b controls the AI ​​character to perform actions output from the trained model by inputting the battle situation acquired by the situation acquisition means 54 into a trained model that has been trained using battle data from a competitive game as training data to acquire character actions appropriate to the battle situation. The battle data used as training data is data collected from human-versus-human battles.

[0064] The evaluation means 56 evaluates the user's actions during AI battles and outputs the evaluation results. The evaluation results include, for example, tasks that the user must accomplish related to the user's actions.

[0065] The evaluation means 56 includes a first evaluation means 56a and a second evaluation means 56b. The first evaluation means 56a performs a first evaluation process. The first evaluation process includes outputting a first evaluation result, which is an evaluation result of evaluating the user's actions in the most recent AI match. The second evaluation means 56b performs a second evaluation process. The second evaluation process includes outputting a second evaluation result, which is an evaluation result of evaluating the user's actions in multiple past AI matches.

[0066] The game mode suggestion means 57 determines the game situation in a competitive game and, according to the determined game situation, suggests a game mode to the user from among several game modes.

[0067] The model creation means 58 creates a trained model that outputs the character's actions by inputting the battle situation acquired by the situation acquisition means 54. Specifically, the model creation means 58 sets model creation conditions according to the user's operation and selects battle data that satisfies the model creation conditions from the stored battle data. The model creation means 58 creates a trained model by performing machine learning using the selected battle data as training data.

[0068] (Functional configuration of the server device's control unit) The control unit 41 of the server device 4 functions as a storage means 61 and a model creation means 62 by executing the game program 42a.

[0069] The storage means 61 stores battle data related to battles when characters battle in a competitive game. Specifically, the storage means 61 stores operation information received from the game device 2 as battle data in the storage unit 42, associating it with battle situation information that indicates the battle situation when the user operation corresponding to that operation information was performed. The storage means 61 also stores the battle data in the storage unit 42 associating it with user identification information.

[0070] User identification information refers to information that makes it possible to identify a user. Examples of information that makes it possible to identify a user include registration information such as a username registered in advance before playing the game, and a user ID assigned to the user by the server device 4.

[0071] The storage means 61 stores the battle data in the storage unit 42, associating it not only with user identification information but also with information about the character used, rank information, and operation type information. The character used information is information about the type of player character used in the battle. The rank information is information indicating the user's rank, and as mentioned above, the rank indicates the degree of skill in the game. The operation type information is information indicating the operation type used by the user during the battle.

[0072] The model creation means 62 creates a trained model that outputs the character's actions by inputting the battle situation acquired by the situation acquisition means 54. Specifically, the model creation means 62 sets model creation conditions and selects battle data that satisfies the model creation conditions from the stored battle data. The model creation means 62 creates a trained model by performing machine learning using the selected battle data as training data.

[0073] The storage unit 42 of the server device 4 stores multiple trained models, which are described later as rank AI models, created by the model creation means 62.

[0074] [How to collect battle data] This section describes how to collect battle data used to create a trained model for determining the behavior of AI characters. Below, as an example of how to collect battle data, we will explain the method of collecting battle data through the aforementioned online battles, referring to Figure 5.

[0075] Figure 5 is a flowchart illustrating an example of the online battle processing flow. As shown in Figure 5, in online battle mode, two users compete against each other via a network. For convenience, in the following explanation, the two users competing will be referred to as the first user and the second user, the game device 2 operated by the first user will be referred to as the first game device 2a, and the game device 2 operated by the second user will be referred to as the second game device 2b. Since the processing performed by the first game device 2a and the second game device 2b is the same, the following explanation will mainly focus on the processing performed by the first game device 2a, and the processing performed by the second game device 2b will be omitted as appropriate.

[0076] In the first game device 2a, when online battle is selected as the game mode based on the operation of the first user to the controller 31, the game progress means 51 sends a matching request from the first game device 2a to the server device 4 (step S101). Similarly in the second game device 2b, the game progress means 51 sends a matching request from the second game device 2b to the server device 4.

[0077] In server device 4, the server control unit 41 performs a well-known matching process (step S201). This causes game devices 2a and 2b to synchronize with each other via server device 4 (step S202). Thus, the match begins. For example, the server control unit 41 performs the matching process so that users of the same rank are synchronized.

[0078] When a match begins, the game progression means 51 of the first game device 2a transmits operation information generated by the first game device 2a to the second game device 2b via the server device 4, and receives operation information generated by the second game device 2b from the second game device 2b via the server device 4 (step S102).

[0079] Furthermore, the game progression means 51 of the first game device 2a transmits match status information indicating the match status during a match to the server device 4 (step S103). That is, in the first game device 2a, the status acquisition means 54 acquires the match status during a match, and the game progression means 51 transmits match status information indicating the match status when a user operation corresponding to the operation information is performed in the first game device 2a to the server device 4. Note that the match status information does not necessarily have to be sent from game device 2 (e.g., first game device 2a) to the opponent's game device 2 (e.g., second game device 2b).

[0080] In the first game device 2a, the PC control means 53 controls the movement of its own character based on operation information generated by the first user's operation of the controller 31. The PC control means 53 also controls the movement of the opponent's character based on operation information received from the second game device 2b (i.e., operation information generated by the second user's operation of the controller 31 in the second game device 2b) (step S104). In this way, the online battle mode proceeds.

[0081] Meanwhile, in the server device 4, the storage means 61 performs a battle data storage process to store data related to AI battles (step S203). Specifically, in the battle data storage process, the storage means 61 stores the operation information and battle status information received from the game device 2 in steps S102 and S103 as battle data in the storage unit 42 for each user. For example, the storage means 61 stores the operation information received from the first game device 2a as battle data in the storage unit 42, associating it with battle status information that shows the battle situation when the operation of the first user corresponding to that operation information was performed.

[0082] The storage means 61 stores the battle data in the storage unit 42, associating it not only with user identification information but also with various other information such as character information used, rank information, and operation type information. Similarly, the storage means 61 also stores the battle data, including operation information and battle status information received from the second game device 2b, in the storage unit 42, associating it with various other information.

[0083] As described above, each time an online match takes place between users, the storage unit 42 of the server device 4 stores the online match data and various information related to that match. During a match, operation information may be sent from one of the first game device 2a and the second game device 2b to the other without going through the server device 4. In this case, after the match ends, the match data and various information may be sent from the first game device 2a and the second game device 2b to the server device 4 and stored therein.

[0084] [AI battle-related processing] Figure 6 is a flowchart showing an example of the AI ​​battle-related processing flow. For example, when the mode selection screen D2 in Figure 3 is displayed and AI battle mode B3 is selected, the game progression means 51 starts the AI ​​battle-related processing.

[0085] The mode selection means 52 determines, in response to user operation, which of the following has been selected: self-AI battle mode, ranked AI battle mode, or spectator mode. For example, when the mode selection screen D2 shown in Figure 3 is displayed on the display 32, the mode selection means 52 determines whether the user has selected self-AI battle mode B4, ranked AI battle mode B5, or AI spectator mode B6.

[0086] The Self-AI Battle Mode and Ranked AI Battle Mode are battle modes in which the player's character competes against an AI character. The AI ​​character's actions are determined by a trained model created using machine learning with battle data from one or more specific users as training data. The trained models that determine the AI ​​character's actions in this game are broadly classified into two types depending on the type of training data used for machine learning. Specifically, the trained models are classified into Self-AI models and Ranked AI models.

[0087] The Self-AI Model is a pre-trained model created by machine learning using the user's own battle data, which is stored in the battle data of Server Device 4, as training data. The AI ​​character whose actions are determined by the Self-AI Model will be referred to as the Self-AI Character. Because the Self-AI Model uses the user's own past battle data as training data, the Self-AI Character will perform actions similar to those of the player character that the user controlled in past battles. In other words, the Self-AI Character can be said to be a non-player character that acts as if the user themselves were controlling it.

[0088] The Rank AI model is a pre-trained model created by machine learning using battle data from multiple users of the same rank, which are stored in the server device 4. The AI ​​character whose actions are determined by the Rank AI model will be called the Rank AI character. Because the Rank AI model uses battle data from multiple users of the same rank as its training data, the Rank AI character will behave similarly to a player character controlled by a user with a skill level similar to the user's. In other words, the Rank AI character can be described as a non-player character that behaves as if it were being controlled by a user with a skill level similar to the user's.

[0089] In this specification, generating or creating an AI character is synonymous with creating a trained model for determining the character's actions in response to the game situation.

[0090] The AI ​​Spectator Mode is a game mode that allows users to watch matches between AI characters.

[0091] (Self-AI match) First, let's explain what happens when the Self-AI Battle mode is selected. In Self-AI Battle mode, the user can create a self-AI model for the character type they want to battle against, and then battle against an AI character running on that self-AI model.

[0092] As shown in Figure 6, the mode selection means 52 determines, in response to user operation, whether the self-AI battle mode has been selected (step S111), the ranked AI battle mode has been selected (step S116), or the spectator mode has been selected (step S120). If the mode selection means 52 determines, in response to user operation, that the self-AI battle mode has been selected (step S111: Yes), the model creation means 58 sets model creation conditions in response to user operation (step S112). The model creation means 58 selects battle data from the stored battle data that satisfies the set model creation conditions, and creates a self-AI character by performing machine learning using the selected battle data as training data (step S113).

[0093] Figure 7 shows an example of the creation condition setting screen D3. For example, if the self-AI battle mode is selected in step S111, the model creation means 58 displays the creation condition setting screen D3 on the display 32. The creation condition setting screen D3 is a screen for the user to set what kind of self-AI character to create. In other words, the creation condition setting screen D3 is a screen for setting which of the user's past battle data to use as training data. The creation condition setting screen D3 includes a character selection area B11, an operation type selection area B12, and a battle count display area B13.

[0094] In the character selection area B11, one character can be selected from several types of characters. By selecting one character from the multiple types of characters in the character selection area B11, the model creation conditions will include the condition that the data must be battle data using the selected character.

[0095] In the operation type selection area B12, two operation types, Normal Operation Type and Simplified Operation Type, are displayed as selectable methods for activating special actions. By selecting operation type 1 in the operation type selection area B12, the model creation conditions will include the condition that the battle data was created using the selected operation type.

[0096] The model creation means 58 selects from the battle data associated with the user identification information battle data that satisfies the model creation conditions, that is, battle data in which the type of character selected in the character selection area B11 was used and the operation type selected in the operation type selection area B12 was used.

[0097] The match count display area B13 shows the number of match data that meet the model creation conditions.

[0098] Figure 8 shows an example of the creation method selection screen D4. For example, when the user selects and executes "Setting Complete B14" in the creation condition setting screen D3, the model creation means 58 displays the creation method selection screen D4 on the display 32. The creation method selection screen D4 is a screen for deciding how much of the match data that satisfies the model creation conditions to select and use as training data.

[0099] In this game, two methods are available for creating your own AI model: "Quick Learning" and "Thorough Learning." The creation method selection screen D4 includes option images B16 representing "Quick Learning" and B17 representing "Thorough Learning," allowing the user to choose either method. "Quick Learning" is a model creation method that uses battle data from 10 to 50 matches as training data. "Quick Learning" allows you to create your own AI model in a relatively short time. "Thorough Learning" uses battle data from 50 to 100 matches as training data. "Thorough Learning" takes a relatively longer time to create the AI ​​model than "Quick Learning," but it allows you to create an AI character that behaves more closely to the actions of the player character controlled by the user.

[0100] The model creation means 58 can change the number of match data used for machine learning depending on whether "easy learning" or "hard learning" is selected. Furthermore, after selecting each creation method, the model creation means 58 can adjust the number of match data used for machine learning within the range of the number of data that can be changed for each creation method. For example, when "easy learning" is selected, the user can specify any number of match data to select, which is less than or equal to the number of match data displayed in the match number display area B13, and is between 10 and 50 matches. For example, when "hard learning" is selected, the user can specify any number of match data to select, which is less than or equal to the number of match data, and is greater than or equal to 50 matches. The number of match data used for training data may be fixed for the selected creation method. For example, when "easy learning" is selected, the number of match data to select may be fixed at 10 matches.

[0101] "Simple learning" and "thorough learning" differ in the number of battle data sets required for training (essential number of battles). If the number of battle data sets that meet the model creation conditions does not reach the essential number of battles, it may not be possible to create a self-AI model. Furthermore, if the number of battle data sets that meet the model creation conditions does not reach the recommended number of battles, even if a self-AI model can be created, the created AI character may not behave in a manner very similar to that of a player character controlled by a user. For this reason, the model creation means 58 outputs warning information if it determines that the number of battle data sets that meet the model creation conditions does not reach a predetermined number (e.g., essential number of battles, recommended number of battles, etc.).

[0102] For example, the creation condition setting screen D3 in Figure 7 includes a warning information display area B15. In the example in Figure 7, the number of match data that satisfies the model creation conditions, which is 27 matches, does not reach the required number of matches for the "thorough learning" creation method. Therefore, the model creation means 58 displays the message "Not enough matches to meet the required number of matches for thorough learning" in the warning information display area B15. For example, if the number of match data that satisfies the model creation conditions is less than 10 matches, the model creation means 58 may display the warning information "Too few matches to create an AI character" in the warning information display area B15. The warning information does not have to be displayed on the creation condition setting screen D3; it may be displayed after selecting a self-AI model creation method or after the user has set the number of match data to select.

[0103] If the model creation means 58 determines that there is a small amount of match data that satisfies the model creation conditions and that the number of match data that satisfies the model creation conditions does not reach a predetermined number, it may perform machine learning using not only the match data of past matches of users who satisfy the model creation conditions, but also the match data of past matches of other users of the same rank who satisfy the model creation conditions. In this case, the model creation means 58 may display a warning message in the warning information display area B15 stating, "Because there is a small number of matches that satisfy the conditions, match data of other users of the same rank will also be used to create the AI ​​character."

[0104] Thus, warning information should simply inform the user of situations that may arise due to a lack of battle data that meets the model creation conditions. For example, the warning information may notify the user that an AI character cannot be created due to a lack of battle data that meets the model creation conditions, or that the AI ​​character intended by the user may not be created.

[0105] Returning to Figure 6, once the creation of the self-AI character, i.e., the creation of the self-AI model, in step S113 is complete, the AI ​​battle process is executed (step S114). In the AI ​​battle process, an AI battle takes place between the self-character controlled by the user and the self-AI character controlled using the self-AI model created in step S113.

[0106] Specifically, in an AI battle, the PC control means 53 controls the actions of its own character based on operation information generated by the user's operation on the controller 31. During the AI ​​battle, the situation acquisition means 54 acquires the battle situation, and the AI ​​character control means 55b controls the AI ​​character to perform actions output from the self-AI model created in step S113 by inputting the battle situation acquired by the situation acquisition means 54 to the self-AI model. In this way, the AI ​​battle in the self-AI battle mode proceeds.

[0107] Furthermore, the first evaluation means 56a executes the first evaluation process (step S115). After the start of the AI ​​match in step S114, the first evaluation means 56a evaluates the user's actions during the AI ​​match, and when the match ends, it outputs the first evaluation result, which is the evaluation result. For example, when the AI ​​match ends, the first evaluation means 56a displays a results screen including the first evaluation result on the display 32.

[0108] Figure 9 shows an example of the results screen D5. After an AI match, the first evaluation means 56a displays the results screen D5 on the display 32. The results screen D5 includes a win / loss display area B21 and a message display area B22. The win / loss display area B21 displays the results regarding the win / loss in the AI ​​match. The message display area B22 displays the first evaluation result, which is the evaluation result of the user's actions in the AI ​​match. Specifically, the message display area B22 outputs two praise comments B23 and B24 and one challenge comment B25. The praise comments are comments that praise the user's actions. The challenge comment is a comment that presents challenges regarding the user's actions.

[0109] The user can see the praise comments B23 and B24 to learn about the positive aspects of their own actions during the AI ​​match. The user can also see the challenge comment B25 to learn about areas for improvement in their own actions during the AI ​​match. For example, a challenge might be advice on actions the user should take in a given situation to increase their win rate. The memory unit 23 stores multiple comment pieces indicating praise and challenge comments, and the first evaluation means 56a determines which comment piece to output from the stored comment pieces based on the content of the AI ​​match.

[0110] Referring to Figure 10, an example of how to determine the comments to be output will be explained. Figure 10 is a table showing an example of the correspondence between evaluation items, counting conditions, praise comments, and problem comments. Information regarding evaluation items, information regarding counting conditions, and evaluation comment information are stored in the storage unit 23 in relation to each other. The evaluation comment information includes information on at least one of the praise comments and problem comments.

[0111] The first evaluation means 56a calculates the frequency of occurrence of a predetermined battle situation associated with each of the predetermined evaluation items during an AI battle. The count conditions in Figure 10 show the battle situations associated with the evaluation items. In other words, in this embodiment, the first evaluation means 56a calculates the frequency of occurrence of a predetermined battle situation associated with each of the predetermined evaluation items during an AI battle, based on the count conditions corresponding to each evaluation item. Based on the calculated frequency of occurrence, the first evaluation means 56a selects an evaluation item from among the multiple evaluation items and outputs predetermined evaluation comment information associated with the selected evaluation item.

[0112] As shown in Figure 5, the evaluation comments associated with an evaluation item include at least one of a praise comment and a problem comment. For example, the evaluation item "throwing technique" shown in Figure 5 is associated with a praise comment, "You're using throwing techniques to get your opponents to move! Keep it up!" and a problem comment, "Try using throwing techniques more often!". For example, if throwing techniques are used frequently, the praise comment, "You're using throwing techniques to get your opponents to move! Keep it up!" is more likely to be output, and if throwing techniques are used infrequently, the problem comment, "Try using throwing techniques more often!" is more likely to be output.

[0113] The frequency of occurrence is, for example, the number of times a match situation corresponding to an evaluation item occurred. For example, the evaluation item "throwing technique" shown in Figure 5 is associated with the counting condition "the player character throws the opponent character with a throwing technique." Therefore, during an AI match, the first evaluation means 56a counts the number of times the player character, the player's own character, throws the non-player character, the AI ​​character, with a throwing technique. Similarly, during an AI match, the first evaluation means 56a counts the number of times each of the evaluation items in Figure 5, "avoiding a throwing technique," "hitting with a special move," and "receiving a projectile," occurred.

[0114] However, the degree of occurrence is not limited to the number of times the counting condition is met. For example, the degree of occurrence may be the amount of time that has elapsed while the counting condition is met.

[0115] The comment for the evaluation item with the highest occurrence rate is not necessarily output. In this embodiment, first, the first evaluation means 56a corrects the occurrence rate and calculates a priority (in other words, an evaluation score) for determining which comment to output first. Then, based on the calculated priority, the first evaluation means 56a determines which comment to output first.

[0116] For each evaluation item, a formula is predetermined to calculate its priority based on its frequency of occurrence. The formula is adjusted to prevent bias in the output comments. In other words, the formula for calculating priority may differ for each evaluation item. For example, depending on the formula, even if the number of "throws" in a match is greater than the number of "special moves that hit," the priority of "throws" may be lower than the priority of "special moves that hit."

[0117] After calculating priority, for example, the first evaluation means 56a extracts a predetermined number of evaluation items in order of priority from the evaluation items, and then randomly selects evaluation items from the extracted evaluation items and outputs comments corresponding to the selected evaluation items. For example, the first evaluation means 56a extracts five evaluation items with high priority from among the evaluation items associated with praise comments, and then randomly selects two evaluation items from among the extracted evaluation items and displays the praise comments corresponding to the selected evaluation items in the message display area B22. For example, the first evaluation means 56a extracts five evaluation items with high priority from among the evaluation items associated with problem comments, and then randomly selects one evaluation item from among the extracted evaluation items and displays the problem comment corresponding to the selected evaluation item in the message display area B22.

[0118] In step S115 and step S119 described later, the first evaluation process not only displays the first evaluation result but also saves it for a certain period of time. In addition, the priority of each evaluation item calculated by the first evaluation means 56a is also saved for a certain period of time as an evaluation list for that AI battle.

[0119] Specifically, in step S115 of Figure 6, the first evaluation means 56a stores the AI ​​match data and the first evaluation result in the storage unit 23 of the game device 2, relating them to each other. The stored first evaluation result is accessible to the user along with past match results, so that the user can check the first evaluation result presented in past AI matches not only immediately after the match, but also after some time has passed.

[0120] Furthermore, in step S115, the first evaluation means 56a sends an evaluation list, which is a list of the priority levels of each evaluation item, from the game device 2 to the server device 4. In this way, each time an AI match is executed, the evaluation list is stored in the memory unit 42 of the server device 4. The evaluation list is used in the second evaluation process described later.

[0121] (Ranked AI match) Next, we will explain what happens when Ranked AI Battle mode is selected. In Ranked AI Battle mode, the user can specify the character type, rank, and control type of the Ranked AI character they want to battle, and then battle an AI character that operates on the Ranked AI model corresponding to those specifications.

[0122] As shown in Figure 6, if the mode selection means 52 determines that the ranked AI battle mode has been selected in response to the user's operation (step S116: Yes), the game progression means 51 selects a ranked AI character, which is the opponent character, in response to the user's operation (step S117).

[0123] Figure 11 shows an example of the AI ​​character selection screen D6. When the ranked AI battle mode is selected in step S116, the model creation means 58 displays the AI ​​character selection screen D6 on the display 32. The AI ​​character selection screen D6 is a screen for determining which of the multiple ranked AI models stored in the server device 4 to use.

[0124] Multiple rank AI models are pre-stored in the memory unit 42 of the server device 4. Each rank AI model is created by machine learning using battle data from the battle data stored in the server device 4, where battle data associated with the same type of character, the same control type, and the same rank is used as training data.

[0125] Specifically, the memory unit 42 stores battle data from various users. In the server device 4, the stored battle data is used as training data, and the model creation means 62 periodically or irregularly inputs battle conditions to create a trained model that outputs the character's actions. The model creation means 62 sets model creation conditions regarding the character type, operation type, and rank, selects battle data that satisfies the model creation conditions from the stored battle data, and creates a trained model by performing machine learning using the selected battle data as training data. The rank AI model stored in the server device 4 is updated periodically or irregularly using the latest multiple battle data stored in the server device 4.

[0126] The AI ​​character selection screen D6 includes a character selection area B31, an operation type selection area B32, and a rank selection area B33. In the character selection area B31, one character can be selected from several types of characters. In the operation type selection area B32, one operation type can be selected from the normal operation type and the simplified operation type, which are methods for activating special actions. In the rank selection area B33, one rank can be selected from several ranks. As mentioned above, rank is a class that represents the degree of skill of each user, positioned according to the rank points assigned to each user.

[0127] When a user selects a character type, control type, and rank, a rank AI model corresponding to the selected character type, control type, and rank combination is selected. For example, if a user selects "Character B" as the character type, "Normal Control Type" as the control type, and "S Rank" as the rank, the rank AI model obtained through machine learning using multiple battle data obtained from S Rank users who select "Character B" as their player character and "Normal Control Type" as the control type will be selected.

[0128] After selecting a rank AI character in step S117, the selected rank AI model is sent from the server device 4 to the game device 2, and the AI ​​battle process is executed (step S118). In the AI ​​battle process, an AI battle takes place between the user's own character and the user's own AI character controlled using the user's own AI model created in step S113. The first evaluation means 56a also executes the first evaluation process (step S119).

[0129] Unlike the AI ​​battle process in step S114, the AI ​​battle process in step S118 uses a ranked AI model instead of a self-AI model to determine the actions of the AI ​​characters. Aside from this point, the AI ​​battle process in step S118 is the same as that in step S114, therefore, a detailed explanation of the AI ​​battle process in step S118 will be omitted.

[0130] Furthermore, the first evaluation process in step S119 is the same as the first evaluation process in step S115, so its explanation is omitted.

[0131] (AI spectator) Next, we will explain what happens when AI spectator mode is selected. In AI spectator mode, the user can watch a match between AI characters.

[0132] As shown in Figure 6, if the mode selection means 52 determines that the ranked AI battle mode has been selected in response to the user's operation (step S120: Yes), the game progression means 51 selects two ranked AI characters to battle with based on the user's operation (step S121).

[0133] In step S121, the selection of AI characters differs from that in step S117, where two characters are selected. Aside from this difference, step S121 is the same as step S117, so its explanation is omitted.

[0134] After selecting a ranked AI character in step S121, the AI ​​spectator process is executed (step S122). In the AI ​​spectator process, the two AI characters selected in step S121 compete against each other.

[0135] The two ranked AI characters selected in step S121 will be referred to as the first AI character and the second AI character, the ranked AI model that outputs the actions of the first AI character will be referred to as the first trained model, and the ranked AI model that outputs the actions of the second AI character will be referred to as the second trained model. In the AI ​​spectator processing, the situation acquisition means 54 acquires the match situation during the match between the first AI character and the second AI character. During the match between the first AI character and the second AI character, the AI ​​character control means 55b controls the first AI character to perform actions output from the first trained model by inputting the acquired match situation to the first trained model. Also, during the match between the first AI character and the second AI character, the AI ​​character control means 55b controls the second AI character to perform actions output from the second trained model by inputting the acquired match situation to the second trained model.

[0136] In spectator mode, the battle screen can be displayed in a different format than in battle mode. For example, the game progression means 51 displays a history of operation information corresponding to the AI ​​character's actions on the battle screen, according to the user's settings.

[0137] [Response to user requests] Figure 12 shows an example of the request screen D7. This game features a support character P3, which is displayed on the request screen D7. Support character P3 is a character that responds to user requests for advice.

[0138] The request screen D7 includes a choice image B41 for the user to provide their play evaluation and a choice image B42 for asking the user which game mode suits them best. If the user selects choice image B41, the second evaluation means 56b executes the second evaluation process. If the user selects choice image B42, the second evaluation means 56b executes the mode suggestion process. In the mode suggestion process, a game mode suitable for the user's game play status is suggested.

[0139] [Second evaluation process] First, let's explain the second evaluation process. Figure 13 is a flowchart showing an example of the flow of the second evaluation process. The second evaluation process includes a task completion check process and an evaluation result output process. The task completion check process checks whether the predetermined continuous tasks have been completed and outputs the results. The task completion check process corresponds to steps S131 to S136 in Figure 13.

[0140] In this game, ongoing tasks are challenges that the user must complete in relation to their actions during AI matches, and are set as tasks to be addressed now. Ongoing tasks are the content of task comments associated with one of the aforementioned evaluation items. Users can check which tasks are set as ongoing tasks. For example, request screen D7 includes a task display area B43 that shows the content of ongoing tasks. Ongoing tasks may also be viewable from other screens.

[0141] The evaluation result output process outputs a second evaluation result, which is an evaluation result of the user's actions in the most recent multiple AI matches. The evaluation result output process corresponds to steps S137 and S138 in Figure 13.

[0142] In the second evaluation process, the second evaluation means 56b first determines whether a predetermined continuing task has been set (step S131). The setting of a predetermined continuing task may mean that continuing task information indicating the continuing task is stored in association with user identification information. In step S131, the second evaluation means 56b accesses the server device 4 and determines that if the continuing task information is stored in association with user identification information, it determines that a continuing task has been set. If it determines that the continuing task information is not stored in association with user identification information, it determines that no continuing task has been set.

[0143] If it is determined that an ongoing task has been set (Step S131: Yes), the second evaluation means 56b determines whether or not the ongoing task has already been achieved (Step S132).

[0144] In step S132, the determination of whether or not the continuing task has already been completed is made using the evaluation list of each AI match that took place after the continuing task was presented.

[0145] Specifically, in steps S115 and S119 of Figure 6, an evaluation list is stored in the storage unit 42 of the server device 4 each time an AI match is executed. The second evaluation means 56b receives a predetermined number of evaluation lists of the most recently stored AI matches from the server device 4. As described above, each evaluation list of a match includes multiple priorities calculated for multiple evaluation items. The second evaluation means 56b extracts the priority of the evaluation item corresponding to the ongoing task from the received evaluation list, and determines whether the ongoing task has already been achieved based on predetermined criteria from the extracted priorities.

[0146] The criteria for judgment are not particularly limited. For example, when game device 2 receives an evaluation list of the latest predetermined number of matches (e.g., 1 match, 2 matches, etc.) from server device 4, the criteria for judgment may be that the continuation task has been achieved if, among the evaluation lists of the received predetermined number of matches, the predetermined number of priorities corresponding to the continuation task all exceed a predetermined threshold value. Alternatively, the criteria for judgment may be that the continuation task has been achieved if, for example, more than half of the predetermined number of priorities exceed a threshold value.

[0147] If it is determined that the ongoing task has been completed (step S132: Yes), the second evaluation means 56b outputs an achievement notification indicating that the ongoing task has been completed (step S133). For example, if the second evaluation means 56b determines that the ongoing task "Try blocking projectiles with a guard or dodging them with a jump" has been completed, it displays a line from support character P3, such as "You've gotten good at dealing with projectiles!", on the display 32 as an achievement notification.

[0148] Furthermore, in step S133, the second evaluation means 56b deletes the continuing task that it has determined to have been achieved. Specifically, the second evaluation means 56b sends a task deletion instruction to the server device 4, and the control unit 41 of the server device 4 deletes the continuing task information associated with the user identification information.

[0149] If it is determined that the ongoing task has not been completed (Step S132: No), the second evaluation means 56b determines whether the number of online matches since the presentation of the ongoing task is equal to or greater than a predetermined number (Step S134).

[0150] If the system determines that the number of online matches played since the presentation of the ongoing task is greater than or equal to a predetermined number (Step S134: Yes), the second evaluation means 56b outputs a notification of non-completion (Step S135). For example, for the ongoing task "Try blocking projectiles with a guard or dodging them with a jump," the second evaluation means 56b displays a line from support character P3, such as "It seems you're still not good at dealing with projectiles," on the display 32 as a notification of non-completion.

[0151] If it is determined that the number of online matches since the presentation of the ongoing task is not equal to or greater than a predetermined number (step S134: No), the second evaluation means 56b outputs a notification of insufficient number of plays (step S136). For example, as a notification of insufficient number of plays, the second evaluation means 56b displays a line from support character P3 on the display 32, such as, "There aren't enough matches, so we can't determine if the ongoing task has been completed."

[0152] If no continuing task has been set (Step S131: No), or after Steps S133, S135, or S136, the second evaluation means 56b outputs a praise comment (Step S137). Subsequently, the second evaluation means 56b outputs a task comment (Step S138).

[0153] This section explains how to determine the praise comments and problem comments in the second evaluation process. In the second evaluation process, the comments to be output are determined based on the evaluation list obtained in the first evaluation process.

[0154] Specifically, the second evaluation means 56b extracts a predetermined number of evaluation items in order of priority from the most recently stored predetermined number of evaluation lists received from the server device 4, then randomly selects evaluation items from among the extracted evaluation items, and outputs comments corresponding to the selected evaluation items.

[0155] For example, the second evaluation means 56b extracts five evaluation items with high priority from among the evaluation items associated with praise comments included in the received evaluation list of, for example, 50 evaluation items, and then randomly selects two more evaluation items from among the extracted evaluation items and outputs the praise comments corresponding to the selected evaluation items as lines of dialogue from support character P3. For example, the second evaluation means 56b extracts five evaluation items with high priority from among the evaluation items associated with problem comments included in the received evaluation list of, for example, 50 evaluation items, and then randomly selects one more evaluation item from among the extracted evaluation items and outputs the problem comment corresponding to the selected evaluation item as lines of dialogue from support character P3.

[0156] In step S138, if no continuing task has been set (step S131: No) or if a continuing task that has been completed in step S133 has been removed from the set tasks, the second evaluation means 56b sets the content of the new task comment output in step S138 as a continuing task. That is, the second evaluation means 56b sends information indicating the new task comment output in step S138 to the server device 4, and the server device 4 stores the received information as new continuing task information associated with the user identification information.

[0157] [Mode proposal processing] Next, we will explain the mode proposal process. Figure 14 is a flowchart showing an example of the flow of the mode proposal process.

[0158] The game mode presentation means 57 determines the game situation (step S141). The game situation indicates at least one of the user's play status in a competitive game and the status of each game mode. Based on the determined game situation, the game mode presentation means 57 presents a game mode to the user from among multiple game modes (step S142).

[0159] For example, the user's play status in a competitive game and the status of each game mode in a competitive game include situations where the user has not yet played each game mode. The game mode presentation means 57 presents the unplayed game modes to the user as recommended game modes when there are unplayed game modes.

[0160] The game mode presentation means 57 does not need to determine whether or not a user has played all game modes included in the competitive game; for example, it may determine whether or not a specific game mode has been played. For example, the online competitive mode is often something that beginners may hesitate to play. Therefore, the game mode presentation means 57 may determine whether or not the online competitive mode has been played, and if it has not been played, it may present the online competitive mode as a recommended game mode to the user. This can encourage beginners to play the online competitive mode.

[0161] For example, the status of each game mode in a competitive game includes situations where a limited-time event is taking place in some of the multiple game modes. For instance, if the game mode presentation means 57 determines that a limited-time tournament is being held in a game mode with multiple game modes, it recommends that the user participate in the tournament. This makes it easier for users to notice limited-time events and encourages them to participate in them.

[0162] For example, in a competitive game, the user's play status includes the user's skill level. For instance, the game mode presentation means 57 may present a game mode appropriate to the user's skill level based on the determined skill level. The play status may include an indicator of the user's skill level, such as a rank or rank points associated with user identification information. The game mode presentation means 57 may also determine the user's skill level as the play status from matches against non-player characters, such as CPU matches. For example, the game mode presentation means 57 may present a game mode appropriate to the user's skill level based on the determined skill level.

[0163] [summary] To summarize, the game program according to this embodiment is a game program that causes a control unit 5 (computer) to execute a competitive game in which characters P1 and P2 compete against each other in a game space S, and the control unit 5 functions as: a player character control means 53 that controls the player character to perform actions in response to user operations; a situation acquisition means 54 that acquires the battle situation during an AI battle, which is a battle between an AI character, which is a non-player character, and the player character; an AI character control means 55b that controls the AI ​​character to perform actions output from a trained model by inputting the acquired battle situation to a trained model that has been trained using the battle data of the competitive game as training data to acquire character actions in accordance with the battle situation; and an evaluation means 56 that evaluates the user's operations in the AI ​​battle and outputs an evaluation result that includes tasks that the user should achieve related to the operations.

[0164] [Effects and Effects] According to the game program of this embodiment, in AI battles, a battle similar to a human-versus-human battle is realized in which a player character controlled by the user is battled against a player character controlled by another user. Therefore, the evaluation means 56 evaluates the user's actions in the AI ​​battle and outputs tasks that the user should accomplish related to those actions, so that the user can obtain tasks that are useful in human-versus-human battles through the AI ​​battle. Accordingly, it is possible to provide a new way to utilize non-player characters that operate based on past battle data.

[0165] Furthermore, in this embodiment, since the result of determining whether or not the ongoing task has been achieved is output, it is possible to motivate the user to operate the player character in order to achieve the task, thereby further improving the enjoyment of the game.

[0166] Furthermore, in this embodiment, the first evaluation means 56a calculates the frequency of occurrence of a predetermined battle situation associated with each of the predetermined evaluation items during an AI battle, selects an evaluation item from among the multiple evaluation items based on the frequency of occurrence, and outputs predetermined evaluation comment information associated with the selected evaluation item. In this way, by providing multiple evaluation items related to the evaluation of operation content, a wide variety of evaluation results can be output.

[0167] Furthermore, in this embodiment, the first evaluation means 56a outputs a first evaluation result, which is an evaluation result of the user's actions in the most recent AI match, and the second evaluation means 56b outputs a second evaluation result, which is an evaluation result of the user's actions in multiple past AI matches. As a result, the user can check not only the evaluation obtained from the most recent match, but also the evaluation obtained from multiple matches, so that the user can obtain tasks that are more useful in human-versus-human matches.

[0168] Furthermore, in this embodiment, the game mode presentation means 57 presents a recommended game mode to the user from among multiple game modes based on the determined game situation, thereby guiding the user to select a game mode that is appropriate for the situation.

[0169] Furthermore, in this embodiment, there is an AI spectator mode, which is a game mode for users to watch battles between AI characters. Therefore, users can watch battles between AI characters without having to control a player character.

[0170] Furthermore, in this embodiment, the model creation means 58 sets model creation conditions according to user operations, selects battle data that satisfies the set model creation conditions from the stored battle data, and creates its own AI character by performing machine learning using the selected battle data as training data, thus allowing the user to freely generate the AI ​​character they desire.

[0171] Furthermore, in this embodiment, users can play against a self-AI character whose actions are determined by a self-AI model created using machine learning based on their own gameplay data. This allows users to learn about their own playing habits and weaknesses, which can help them further improve their playing skills.

[0172] Furthermore, in this embodiment, the model creation means 58 can change the number of match data selected according to the user's choice, thus allowing adjustment of the balance between the time required for machine learning and the accuracy of the AI ​​character's movements obtained through machine learning (i.e., the similarity of the AI ​​character's movements to the player character's movements in past matches).

[0173] Furthermore, in this embodiment, the model creation means 58 outputs warning information if it determines that the number of battle data that satisfy the model creation conditions has not reached a predetermined number (e.g., required number of battles, recommended number of battles, etc.). This allows, for example, the user to be guided to increase the number of battles, i.e., the battle data.

[0174] <Other Embodiments> The present invention is not limited to the embodiments described above, and various modifications are possible without departing from the spirit of the invention.

[0175] For example, the game program 34a does not have to be recorded on the game media 34, and may be downloaded to the game device 2 from an external source via a communication network NW. For example, the game device 2 included in the game system of the present invention may be a portable information terminal such as a smartphone, and the operation unit and display unit may be a touchscreen or the like. Also, the display unit that displays the game screen may be a head-mounted display or the like.

[0176] Furthermore, although the game system 1 in the above embodiment included a game device 2 and a server device 4, the game system of the present invention may be configured without a server device.

[0177] The game program of the present invention may be executed by at least one computer, either a game device 2 or a server device 4. For example, a portion of the game program of the present invention may be stored in the server device, and the remainder of the game program may be downloaded to the game device or recorded on a recording medium for loading into the game device. For example, the control unit 41 of the server device 4 may function as some or all of the various means 51 to 58 by executing the game program stored in the storage unit 42. In this case, the game program of the present invention is a program stored in the server device 4 or distributed and stored in both the game device 2 and the server device 4.

[0178] In the above embodiment, a fighting game was described as an example of a competitive game, but the present invention is not limited to this. For example, the present invention can also be applied to fighting games involving multiple characters versus multiple characters. For example, the present invention can be applied not only to competitive fighting games, but also to various competitive games such as action games, role-playing games, shooting games, sports games, and racing games.

[0179] In the above embodiment, the second evaluation process included a task achievement check process, but the second evaluation process may consist only of the evaluation result output process. That is, in the second evaluation process of Figure 13, steps S131 to S136 may be omitted. Conversely, the second evaluation process may consist only of the task achievement check process.

[0180] In the above embodiment, the first evaluation means 56a calculates the frequency of occurrence of a predetermined battle situation associated with each of the predetermined evaluation items during an AI battle, corrects the frequency of occurrence, calculates a priority for determining which comments to output preferentially, and determines which comments to output preferentially based on the calculated priority. However, the method for determining which comments to output is not limited to this. For example, the evaluation means may, after calculating the frequency of occurrence, prioritize comments for evaluation items with a high frequency of occurrence without calculating a priority. The evaluation means may also output the frequency of occurrence of all of the multiple evaluation items as an evaluation result. For example, the evaluation means may not even need to calculate the frequency of occurrence.

[0181] In the above embodiment, the evaluation means 56 includes a first evaluation means 56a and a second evaluation means 56b, but the evaluation means 56 may not include either the first evaluation means 56a or the second evaluation means 56b.

[0182] In the above embodiment, a mode suggestion process was performed, but the present invention is not limited to performing a mode suggestion process.

[0183] In the above embodiment, the self-AI model was created after selecting the self-AI battle mode, but it may also be possible to select a previously created self-AI character in the self-AI battle mode. The self-AI character may be created at the user's preferred timing before selecting the self-AI battle mode. The self-AI model created by the user's operation may be saved in memory so that it can be used for battles at any time later.

[0184] In the above embodiment, the AI ​​battle modes described were a self-AI battle mode, a ranked AI battle mode, and an AI spectator mode. However, the present invention is also applicable to games that do not offer some of these modes.

[0185] In the above embodiment, a self-AI model and a rank AI model were described as trained models for determining the behavior of an AI character, but the trained models are not limited to these. For example, a user may create a trained model using the battle data of a specific other user designated by the user as training data. In other words, a third-party AI model may be created. The model creation means may create a trained model by performing machine learning using the battle data of a specific other user (e.g., a friend) designated by the user as training data from the battle data stored in the server device 4.

[0186] In the above embodiment, the number of match data used for machine learning could be changed, but the number of match data used for machine learning may not be selectable by the user, or it may be fixed to a predetermined number.

[0187] In the above embodiment, the control unit 21 of the game device 2 functioned as the model creation means 58, but machine learning may be performed in the control unit 41 of the server device 4. That is, in the self-AI battle mode, information indicating the model creation conditions set by the user's operation may be sent from the game device 2 to the server device 4, and the model creation means 62 in the server device 4 may create a self-AI model based on the received model creation conditions.

[0188] In the above embodiment, the model creation conditions included conditions related to the character used, conditions related to the operation type, and conditions related to the rank. However, the model creation conditions may not include some of these conditions, or they may include other conditions.

[0189] In the above embodiment, the model creation conditions could be set by the user's actions, but it is not necessary for the model creation conditions to be set by the user's actions.

[0190] The battle data used for training does not have to be data obtained from online battles. Battle data from other human-versus-human battles may also be used as training data. For example, battle data from CPU battles or AI battles may also be used as training data.

[0191] Conversely, the battle data used for training may be limited to battle data from online battle modes, such as the ranked match mode described in the above embodiment, where a parameter indicating the degree of skill in the battle game increases or decreases depending on wins and losses. This is because users tend to take these types of battles seriously, and accurate battle data can be used for training.

[0192] In step S121 of Figure 6, the player could select the character type, rank, and control type of the AI ​​character to play against, but some of these may be determined automatically (for example, randomly). For example, the rank may be set to the same rank as the user. The character type may also be determined randomly.

[0193] The model creation means 58 may estimate the time required to create the trained model according to the number of battle data used to create the trained model, and output the estimation result. For example, the model creation means 58 may estimate the time required until the self-AI model is created according to the amount of data used for training, and may display the estimated time. For example, the model creation means 58 may display the estimated time before starting the creation of the self-AI model, or it may be displayed after the creation has started but before it is completed. The display according to the estimated time may be the estimated time itself, a countdown display of the time required until creation is completed, or a time gauge that visually shows the time required until creation is completed. This makes it easier for the user to look at the display according to the estimated time and decide whether to choose "easy learning" or "hard learning," or whether to stop creating the model after starting. By outputting the estimated time required to create the trained model, the user can use the outputted time as information to decide whether or not to proceed with creating the trained model.

[0194] In the above embodiment, two praise comments and one problem comment were output as evaluation results, but the number of comments to be output is not limited to this. For example, the evaluation means does not have to output praise comments, and may output only the problem comment as an evaluation result.

[0195] In the above embodiment, the evaluation means evaluated the user's actions in an AI match and output the evaluation results. However, the evaluation means may also evaluate the user's actions in matches other than AI matches (such as human vs. human matches or CPU matches) and output the evaluation results. Furthermore, in the above embodiment, the evaluation results output by the evaluation means included tasks that the user should accomplish in relation to the actions. However, the evaluation results do not necessarily need to include tasks that the user should accomplish in relation to the actions.

[0196] The effects and advantages of the present invention will also be exhibited when these other embodiments are adopted. Furthermore, it is possible to combine these embodiments and other embodiments as appropriate.

[0197] [Summary of Disclosures] Each of the following embodiments is a disclosure of a preferred embodiment.

[0198] [Aspect 1] A game program that causes a computer to run a competitive game in which characters fight each other in a game space, The aforementioned computer, A player character control means that controls the player character to perform actions in response to user input. A means for acquiring the status of a match during an AI match between a non-player character (an AI character) and a player character. An AI character control means controls the AI ​​character to perform actions output from the trained model by inputting the acquired battle situation to a trained model that has been trained using battle data from the aforementioned battle game as training data to acquire character actions appropriate to the battle situation, and An evaluation means that evaluates the user's actions in the AI ​​battle and outputs an evaluation result that includes tasks the user should accomplish related to those actions. A game program that functions as such.

[0199] According to Embodiment 1, in AI battles, a battle similar to a human-versus-human battle is realized in which a player character controlled by a user competes against a player character controlled by another user. Therefore, the evaluation means evaluates the user's actions in the AI ​​battle and outputs tasks that the user should accomplish in relation to those actions, so that the user can obtain tasks that are useful in human-versus-human battles through the AI ​​battle. Accordingly, it is possible to provide a new way to utilize non-player characters that operate based on past battle data.

[0200] [Aspect 2] The game program according to Embodiment 1, wherein the evaluation means determines whether the task has been achieved in the AI ​​battle after outputting the task, and outputs the determination result regarding the achievement of the task.

[0201] According to embodiment 2, since a result of determining whether or not the task has been achieved is output, it is possible to motivate the user to operate the player character in order to achieve the task, thereby further improving the enjoyment of the game.

[0202] [Aspect 3] The evaluation means is, For each of the predetermined multiple evaluation items, calculate the frequency of occurrence of the predetermined battle situation associated with each of the multiple evaluation items during the AI ​​battle. Based on the degree of occurrence, an evaluation item is selected from among the multiple evaluation items. A game program according to embodiment 1 or 2, which outputs predetermined evaluation comment information associated with the selected evaluation item.

[0203] According to embodiment 3, by providing multiple evaluation items, a wide variety of evaluation results can be output.

[0204] [Aspect 4] The evaluation means is, A first evaluation means that outputs a first evaluation result, which is the evaluation result of the user's actions in the preceding AI battle, and A second evaluation means that outputs a second evaluation result, which is the evaluation result of evaluating the user's actions in multiple AI battles that have been conducted in the past. A game program according to any one of embodiments 1 to 3, including the following.

[0205] According to aspect 4, the user can obtain tasks that are more useful in player-versus-player battles.

[0206] [Aspect 5] The aforementioned competitive game includes multiple game modes selectable by the user, A game program according to any one of embodiments 1 to 4, wherein the computer is further made to function as a game mode presentation means that determines a game status indicating at least one of the user's play status in the competitive game and the status of each of the game modes, and presents a game mode to the user from among the plurality of game modes based on the determined game status.

[0207] According to embodiment 5, the user can be guided to select a game mode appropriate to the situation.

[0208] [Aspect 6] The computer is further configured to function as a mode selection means, which, in response to user input, selects a spectator mode as the game mode of the battle game, for watching a battle between a first AI character and a second AI character, both of which are non-player characters. The status acquisition means, when the spectator mode is selected by the mode selection means, acquires the battle status during the battle between the first AI character and the second AI character. The game program according to any one of embodiments 1 to 5, wherein the AI ​​character control means controls the first AI character to perform actions output from the first trained model by inputting the acquired match status to a first trained model, which has been machine-trained using the match data relating to the first AI character's match as training data, when the mode selection means selects the spectator mode, during a match between the first AI character and the second AI character, and controls the second AI character to perform actions output from the second trained model by inputting the acquired match status to a second trained model, which has been machine-trained using the match data relating to the second AI character's match as training data.

[0209] According to embodiment 6, the user can watch the AI ​​characters compete without having to control a player character.

[0210] [Aspect 7] The aforementioned computer, A storage means for storing the battle data related to a battle when a battle takes place between characters in the aforementioned battle game. The aforementioned trained model is further made to function as a model creation means, The aforementioned model creation means is The model creation conditions are set according to the user's actions. From the stored match data, select the match data that satisfies the model creation conditions. A game program according to any one of embodiments 1 to 6, which creates the trained model by performing machine learning using the selected match data as the training data.

[0211] According to embodiment 7, a trained model is created using battle data that satisfies the model creation conditions set by the user, so the user can freely generate the AI ​​character they desire.

[0212] [Aspect 8] The game program according to embodiment 7, wherein the model creation means changes the number of battle data used for machine learning according to the user's selection.

[0213] According to aspect 8, the user can change the number of match data points used to create the trained model.

[0214] [Aspect 9] The game program according to embodiment 7 or 8, wherein the model creation means estimates the time required to create the trained model according to the number of battle data used to create the trained model, and outputs the estimation result.

[0215] According to embodiment 9, the estimated time required to create the trained model is output, allowing the user to use the outputted time as information to decide whether or not to proceed with creating the trained model.

[0216] [Aspect 10] The game program according to any one of embodiments 7 to 9, wherein the model creation means outputs warning information when it determines that the number of battle data that satisfy the model creation conditions does not reach a predetermined number.

[0217] According to embodiment 10, the output of warning information can guide the user to increase the number of matches played, i.e., the number of match data.

[0218] [Aspect 11] A program storage unit storing a game program according to any one of embodiments 1 to 10, A game system comprising a computer that executes a program stored in the program storage unit.

[0219] [Aspect 12] A game control method for controlling a competitive game in which characters compete against each other in a game space on a computer, A player character control step that controls the player character to perform actions in response to user input. A situation acquisition step to acquire the match status during an AI match, which is a match between an AI character that is a non-player character and a player character. An AI character control step involves controlling the AI ​​character to perform actions output from a trained model, which has been trained using the battle data of the aforementioned battle game as training data to acquire character actions appropriate to the battle situation, by inputting the acquired battle situation, and An evaluation step that evaluates the user's actions in the AI ​​battle and outputs an evaluation result that includes tasks the user should accomplish related to those actions. Game control methods, including [Explanation of Symbols]

[0220] 1: Game System 2: Game device 4: Server device 5: Control Unit 52: Mode selection means 53: Player Character Control Means 54: Means of obtaining the situation 55b: AI character control means 56: Evaluation methods 56a: First evaluation means 56b: Second evaluation method 57: Game Mode Presentation Method 58: Model creation methods 61: Storage means 62: Model creation methods

Claims

1. A game program that causes a computer to run a competitive game in which characters fight against each other in a game space, The aforementioned computer, A player character control means that controls the player character to perform actions in response to user input. A means for acquiring the status of a match during an AI match, which is a match between an AI character (a non-player character) and a player character. An AI character control means controls the AI ​​character to perform actions output from the trained model by inputting the acquired battle situation to a trained model that has been trained using battle data from the aforementioned battle game as training data to acquire character actions appropriate to the battle situation, and An evaluation means that evaluates the user's actions in the AI ​​battle and outputs an evaluation result that includes tasks that the user should accomplish related to the actions. A game program that functions as such.

2. The game program according to claim 1, wherein the evaluation means determines whether the task has been achieved in the AI ​​match after outputting the task, and outputs the determination result regarding the achievement of the task.

3. The evaluation means is For each of the predetermined multiple evaluation items, the frequency of occurrence of the predetermined match situation associated with each of the multiple evaluation items during the AI ​​match is calculated. Based on the degree of occurrence, an evaluation item is selected from among the multiple evaluation items. The game program according to claim 1 or 2, which outputs predetermined evaluation comment information associated with the selected evaluation item.

4. The evaluation means is A first evaluation means that outputs a first evaluation result, which is the evaluation result of the user's actions in the preceding AI battle, and A second evaluation means that outputs a second evaluation result, which is the evaluation result of evaluating the user's actions in multiple AI battles that have been conducted in the past. The game program according to claim 1 or 2, including the following:

5. The aforementioned competitive game includes multiple game modes selectable by the user, The game program according to claim 1 or 2, wherein the computer is further configured to function as a game mode presentation means that determines a game status indicating at least one of the user's play status in the competitive game and the status of each of the game modes, and presents a game mode to the user from among the plurality of game modes based on the determined game status.

6. The computer is further configured to function as a mode selection means, which, in response to user input, selects a spectator mode as the game mode of the versus game, for watching a match between a first AI character and a second AI character, which are non-player characters. The status acquisition means, when the spectator mode is selected by the mode selection means, acquires the battle status during the battle between the first AI character and the second AI character. The game program according to claim 1 or 2, wherein the AI ​​character control means, when the mode selection means selects the spectator mode, controls the first AI character to perform an action output from the first trained model by inputting the acquired match status to a first trained model, which has been machine-trained using the match data relating to the first AI character's match as training data, during a match between the first AI character and the second AI character, and controls the second AI character to perform an action output from the second trained model by inputting the acquired match status to a second trained model, which has been machine-trained using the match data relating to the second AI character's match as training data.

7. The aforementioned computer, A storage means for storing the battle data related to a battle when a battle takes place between characters in the aforementioned battle game. The aforementioned trained model is further made to function as a model creation means, The aforementioned model creation means is The model creation conditions are set according to the user's actions. From the stored match data, select the match data that satisfies the model creation conditions. The game program according to claim 1 or 2, which creates the trained model by performing machine learning using the selected match data as the training data.

8. The game program according to claim 7, wherein the model creation means changes the number of battle data used for machine learning according to the user's selection.

9. The game program according to claim 7, wherein the model creation means estimates the time required to create the trained model according to the number of battle data used to create the trained model, and outputs the estimation result.

10. The game program according to claim 7, wherein the model creation means outputs warning information when it determines that the number of battle data that satisfy the model creation conditions does not reach a predetermined number.

11. A program storage unit storing the game program described in claim 1 or 2, A game system comprising a computer that executes a program stored in the program storage unit.

12. A game control method for controlling a competitive game in which characters compete against each other in a game space on a computer, A player character control step that controls the player character to perform actions in response to user input. A situation acquisition step to acquire the match status during an AI match, which is a match between an AI character that is a non-player character and a player character. An AI character control step in which the AI ​​character is controlled to perform actions output from the trained model by inputting the acquired battle situation to a trained model that has been trained using battle data from the aforementioned battle game as training data to acquire character actions appropriate to the battle situation, and An evaluation step that evaluates the user's actions in the AI ​​battle and outputs an evaluation result that includes tasks that the user should accomplish related to the actions. Game control methods, including

Images