Game team forming method and device, storage medium and program product

Abstract

This invention discloses a game team formation method, apparatus, storage medium, and program product. First, based on the desired team positions of each participant in the game, the positions of the participants are adjusted to obtain multiple game combinations. Then, the game feature information of all participants in each game combination is concatenated to obtain the game feature information of each game combination. Next, win rate prediction is performed based on the game feature information of each game combination to obtain the predicted win rate of each game combination. Finally, the game combination corresponding to the one that meets the target win rate among all predicted win rates is determined as the target game combination, and the participants are teamed up according to the target game combination. This invention can reduce the probability of lopsided game outcomes without affecting the satisfaction of position allocation, thereby improving the gaming experience. This invention can be applied to fields such as artificial intelligence, data processing, and game technology.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a game team formation method, apparatus, storage medium, and program product. Background Technology

[0002] Team games are games in which multiple players form teams and compete against each other. Before players enter the game, the game system automatically matches them with teammates and opponents to form competitive matches, allowing players to cooperate with their teammates to fight against opponents.

[0003] To avoid lopsided matches, related technologies adjust player team positions after the initial lineup is determined, taking into account individual skill levels to ensure teams are evenly matched and improve the overall gaming experience. While this approach balances team strength, the actual assigned positions may deviate significantly from players' desired ones, impacting player satisfaction and ultimately hindering a better gaming experience. Summary of the Invention

[0004] The following is an overview of the subject matter described in detail herein. This overview is not intended to limit the scope of the claims.

[0005] This invention provides a game team formation method, apparatus, storage medium, and program product that can reduce the probability of lopsided matches without affecting the satisfaction of position allocation, thereby improving the gaming experience.

[0006] On one hand, embodiments of the present invention provide a game team-up method, including the following steps:

[0007] Obtain the desired team positions of each participant in the game;

[0008] Based on the expected team positions, the positions of each participating entity are adjusted between teams to obtain multiple match combinations;

[0009] Obtain game characteristic information of each of the participating objects;

[0010] The game feature information of all the participants in each game combination is concatenated to obtain the game feature information of each game combination;

[0011] Win rate prediction is performed based on the game feature information of each game combination to obtain the predicted win rate of each game combination.

[0012] Among all the predicted win rates, the match combination that meets the target win rate is determined as the target match combination, wherein the target win rate represents that the teams have balanced game strength;

[0013] The game teams are formed for each of the participants based on the target match combination.

[0014] On the other hand, embodiments of the present invention also provide a game team-up device, comprising:

[0015] The team position acquisition unit is used to acquire the expected team positions of each participant in the game.

[0016] The match combination acquisition unit is used to adjust the positions of each participating object between teams according to the expected team positions, so as to obtain multiple match combinations.

[0017] A game feature acquisition unit is used to acquire game feature information of each of the participating objects;

[0018] The game feature splicing unit is used to splice the game feature information of all the participating objects in each game combination to obtain the game feature information of each game combination.

[0019] The game win rate prediction unit is used to predict the win rate based on the game feature information of each game combination, and obtain the predicted win rate of each game combination.

[0020] The target match determination unit is used to determine the match combination that meets the target win rate from all the predicted win rates as the target match combination, wherein the target win rate represents that the teams have balanced game strength;

[0021] The game match teaming unit is used to team up the participants in the game according to the target match combination.

[0022] Optionally, the game combination acquisition unit is further configured to:

[0023] Get the current team position of each of the participating objects;

[0024] Based on the current team position and the expected team position, the positions of each participating entity are adjusted between teams to obtain multiple match combinations.

[0025] Optionally, the game feature splicing unit is further used for:

[0026] Determine the team position of each participant in each of the aforementioned match combinations;

[0027] The game feature information of all the participants in each game combination is spliced ​​together according to their positions in the team to obtain the game feature information of each game combination.

[0028] Optionally, the game win rate prediction unit is further used for:

[0029] The game feature information of each game combination is feature encoded to obtain the game feature encoding information of each game combination;

[0030] Win rate prediction is performed based on the game feature encoding information of each game combination to obtain the predicted win rate of each game combination.

[0031] Optionally, the game win rate prediction unit is further used for:

[0032] Vector embedding is performed on the game feature information of each game combination to obtain the game feature embedding vector of each game combination;

[0033] Feature encoding is performed on the game feature embedding vectors of each game combination to obtain the game feature encoding information of each game combination.

[0034] Optionally, the game feature information includes lineup feature information, auxiliary feature information, and position feature information; the game win rate prediction unit is further used for:

[0035] For each of the aforementioned game combinations, vector embedding is performed on the lineup feature information to obtain a lineup feature embedding vector, vector embedding is performed on the auxiliary feature information to obtain an auxiliary feature embedding vector, and vector embedding is performed on the position feature information to obtain a position feature embedding vector.

[0036] The lineup feature embedding vector, the auxiliary feature embedding vector, and the position feature information are concatenated to obtain the game feature embedding vector.

[0037] Optionally, the game feature acquisition unit is further configured to:

[0038] Obtain the historical game information of each of the aforementioned participants;

[0039] The historical game information of each participating object is feature-encoded to obtain the game feature information of each participating object.

[0040] Optionally, the game feature acquisition unit is further configured to:

[0041] Based on the historical game information of each participating object, obtain multiple non-quantitative feature information and multiple quantitative feature information of each participating object;

[0042] The multiple non-quantized feature information of each of the participating objects is feature-encoded to obtain multiple non-quantized feature-encoded information;

[0043] The multiple quantization feature information of each of the participating objects is feature-encoded to obtain multiple quantization feature-encoded information;

[0044] By concatenating all the non-quantized feature encoding information and all the quantized feature encoding information of each participating object, the game feature information of each participating object is obtained.

[0045] Optionally, the game feature acquisition unit is further configured to:

[0046] Obtain multiple quantitative indicator information and multiple non-quantitative indicator information of each participating object from the historical game information of each participating object;

[0047] Feature extraction is performed on multiple quantitative indicator information of each of the participating objects to obtain multiple quantitative feature information of each of the participating objects;

[0048] Feature extraction is performed on multiple non-quantitative indicator information of each of the participating objects to obtain multiple non-quantitative feature information of each of the participating objects.

[0049] Optionally, the game feature acquisition unit is further configured to:

[0050] Discretize the multiple non-quantized feature information of each of the participating objects to obtain multiple discrete feature information sets;

[0051] Each of the non-quantized feature information in each of the discrete feature information sets is feature-encoded to obtain multiple non-quantized feature-encoded information.

[0052] Optionally, the game feature acquisition unit is further configured to:

[0053] Determine the number of categories;

[0054] Based on the number of classifications, the non-quantitative feature information of each of the participating objects is mapped to obtain the feature classification number of each of the non-quantitative feature information;

[0055] The non-quantitative feature information is categorized according to the feature classification number to obtain multiple discrete feature information sets.

[0056] Optionally, the game team-up device further includes:

[0057] A sample acquisition unit is used to acquire training samples and label information, wherein the training samples include game feature sample information of each participant in historical game matches;

[0058] The sample splicing unit is used to splice the game feature sample information of all the participants in the historical game matches to obtain the game feature sample information of the historical game matches.

[0059] The win rate prediction unit is used to input the game feature sample information into the win rate prediction model to predict the win rate and obtain the training predicted win rate of the historical game game.

[0060] The loss calculation unit is used to calculate the loss value based on the training prediction win rate and the label information;

[0061] The parameter correction unit is used to correct the parameters of the win rate prediction model based on the loss value.

[0062] Optionally, the sample splicing unit is further used for:

[0063] Determine the in-team position information of each of the participants in the historical game;

[0064] Based on the team's position information, the game feature sample information of all the participants in the historical game matches is spliced ​​together to obtain the game feature sample information of the historical game matches.

[0065] Optionally, the win rate prediction unit is further configured to:

[0066] The game feature sample information is input into the win rate prediction model, and the game feature sample information is feature encoded to obtain game feature encoded sample information;

[0067] Win rate prediction is performed on the game feature encoding sample information to obtain the training prediction win rate of the historical game games.

[0068] Optionally, the win rate prediction unit is further configured to:

[0069] The game feature sample information is vector-embedded to obtain the game feature sample embedding vector;

[0070] Feature encoding is performed on the embedding vector of the game feature sample to obtain the game feature encoded sample information.

[0071] Optionally, the game feature sample information includes lineup feature sample information, auxiliary feature sample information, and position feature sample information; the win rate prediction unit is further used for:

[0072] The lineup feature sample information is vector-embedded to obtain a lineup feature sample embedding vector, the auxiliary feature sample information is vector-embedded to obtain an auxiliary feature sample embedding vector, and the position feature sample information is vector-embedded to obtain a position feature sample embedding vector.

[0073] The game feature sample embedding vector is obtained by concatenating the lineup feature sample embedding vector, the auxiliary feature sample embedding vector, and the position feature sample information.

[0074] On the other hand, embodiments of the present invention also provide a game team-up device, comprising:

[0075] At least one processor;

[0076] At least one memory for storing at least one program;

[0077] The game teaming method as described above is implemented when at least one of the programs is executed by at least one of the processors.

[0078] On the other hand, embodiments of the present invention also provide a computer-readable storage medium storing a processor-executable program, which, when executed by a processor, is used to implement the game team-up method as described above.

[0079] On the other hand, embodiments of the present invention also provide a computer program product, including a computer program or computer instructions, the computer program or computer instructions being stored in a computer-readable storage medium, a processor of a computer device reading the computer program or computer instructions from the computer-readable storage medium, and the processor executing the computer program or computer instructions to cause the computer device to perform the game team formation method as described above.

[0080] The embodiments of the present invention include at least the following beneficial effects: First, the positions of each participant in the game are adjusted according to their expected team positions to obtain multiple game combinations. Then, the game feature information of all participants in each game combination is spliced ​​together to obtain the game feature information of each game combination. Next, the win rate is predicted based on the game feature information of each game combination to obtain the predicted win rate of each game combination. Then, the game combination that meets the target win rate is determined as the target game combination from all the predicted win rates, and the game teams of each participant are formed according to the target game combination. Since the various match combinations used for win rate prediction are obtained by adjusting the team positions of each participant based on their desired team roles, these match combinations can satisfy the position allocation satisfaction of each participant. Furthermore, since the target win rate represents a balanced game strength among all teams, when the match combination corresponding to the target win rate is determined as the target match combination from all predicted win rates, and the participants are teamed up according to the target match combination, the probability of lopsided matches can be reduced, maximizing the likelihood of balanced matches and thus achieving the goal of balanced matches. Therefore, the solution of this embodiment can reduce the probability of lopsided matches without affecting position allocation satisfaction, thereby improving the gaming experience.

[0081] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the description, claims, and drawings. Attached Figure Description

[0082] The accompanying drawings are provided to further understand the technical solutions of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the technical solutions of the present invention, and do not constitute a limitation on the technical solutions of the present invention.

[0083] Figure 1 This is a schematic diagram of an implementation environment provided by an embodiment of the present invention;

[0084] Figure 2 This is a schematic diagram of another implementation environment provided by an embodiment of the present invention;

[0085] Figure 3 This is a flowchart of a game team formation method provided in an embodiment of the present invention;

[0086] Figure 4 This is a schematic diagram of the structure of a win rate prediction model provided in an embodiment of the present invention;

[0087] Figure 5 This is a flowchart illustrating a specific example of a game team-up method provided by the present invention;

[0088] Figure 6 This is a schematic diagram of the game interface after assigning initial team positions to each participant, as provided in an example of the present invention.

[0089] Figure 7 This is a schematic diagram of a game interface for predicting the win rates of both teams, provided as an example of the present invention.

[0090] Figure 8 This is a schematic diagram of the game interface after teaming up the participants according to the target match combination in an example of the present invention;

[0091] Figure 9 This is a complete flowchart of a game team-up method provided by a specific example of the present invention;

[0092] Figure 10 This is a predicted win rate distribution chart after adopting the game team formation method of this invention.

[0093] Figure 11 This is a predicted win rate analysis chart after adopting the game team formation method of this invention embodiment;

[0094] Figure 12 This is a schematic diagram of a game team-up device provided in an embodiment of the present invention;

[0095] Figure 13 This is a schematic diagram of another game team-up device provided in an embodiment of the present invention. Detailed Implementation

[0096] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. The described embodiments should not be considered as limitations on the present invention, and all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of the present invention.

[0097] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0098] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein is for the purpose of describing embodiments of the invention only and is not intended to limit the invention.

[0099] Before providing a further detailed description of the embodiments of the present invention, the nouns and terms involved in the embodiments of the present invention will be explained, and the nouns and terms involved in the embodiments of the present invention shall be interpreted as follows.

[0100] 1) Participants: This can refer to users participating in the game, or it can refer to game accounts participating in the game.

[0101] 2) Virtual character: A character in a game controlled by the participant, such as a "hero" in a multiplayer online battle arena (MOBA) game.

[0102] 3) Multiplayer online tactical battle royale games (MOTs) refer to games in which multiple players form multiple teams and engage in tactical battles against each other. In MOTs, the multiple players are divided into two or more teams. Each player controls their chosen virtual character to fight on a virtual map. By eliminating players from other teams or artificial intelligence (AI) units on the map, players acquire virtual resources to purchase virtual equipment. The ultimate goal of the game is to destroy the bases of other teams.

[0103] 4) Artificial Intelligence (AI) is the theory, methods, technology, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. In other words, AI is a comprehensive technology within computer science that attempts to understand the essence of intelligence and produce a new kind of intelligent machine that can react in a way similar to human intelligence. AI studies the design principles and implementation methods of various intelligent machines, enabling them to have perception, reasoning, and decision-making capabilities. AI technology is a comprehensive discipline involving a wide range of fields, encompassing both hardware and software technologies. Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operating / interactive systems, and mechatronics. AI software technologies mainly include computer vision, speech processing, natural language processing, and machine learning / deep learning.

[0104] With the research and advancement of artificial intelligence (AI) technology, AI is being studied and applied in various fields, such as smart homes, smart wearable devices, virtual assistants, smart speakers, smart marketing, autonomous driving, drones, robots, smart healthcare, and smart customer service. It is believed that with the development of technology, AI will be applied in more fields and play an increasingly important role.

[0105] 5) Machine Learning (ML) is a multidisciplinary field involving probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. It specifically studies how computers can simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to endow computers with intelligence; its applications span all areas of artificial intelligence. Machine learning and deep learning typically include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and instructional learning.

[0106] 6) Blockchain is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and cryptographic algorithms. Essentially, a blockchain is a decentralized database, a chain of data blocks linked together using cryptographic methods. Each data block contains information about a batch of network transactions, used to verify the validity of the information (anti-counterfeiting) and generate the next block. A blockchain can include an underlying platform, a platform product service layer, and an application service layer.

[0107] In current multiplayer online battle arena (MOBA) games, the matchmaking system is its core component, used to group multiple participants into different teams to ensure fairness and enhance the gaming experience. The quality of the matchmaking system directly impacts the core gameplay experience, ultimately determining the game's reputation and player retention rate. To avoid lopsided matches, matchmaking systems typically adjust the team positions based on the skill level of each participant after initial selection, aiming to create teams with similar skill levels and improve the gaming experience. While these systems can balance team strength, the actual assigned team positions may deviate significantly from the player's desired positions after the position adjustment, affecting player satisfaction and hindering the overall gaming experience.

[0108] To reduce the probability of lopsided matches without affecting position allocation satisfaction, this invention provides a game teaming method, a game teaming device, a computer-readable storage medium, and a computer program product. First, the positions of each participant in the game are adjusted according to their desired teaming positions, resulting in multiple match combinations. Then, the game feature information of all participants in each match combination is concatenated to obtain the match feature information of each match combination. Next, win rate prediction is performed based on the match feature information of each match combination to obtain the predicted win rate of each match combination. Finally, the match combination that matches the target win rate from all predicted win rates is determined as the target match combination, and the participants are teamed up according to the target match combination. Since the various match combinations used for win rate prediction are obtained by adjusting the team positions of each participant based on their desired team roles, these match combinations can satisfy the position allocation satisfaction of each participant. Furthermore, since the target win rate represents a balanced game strength among all teams, when the match combination corresponding to the target win rate is determined as the target match combination from all predicted win rates, and the participants are teamed up according to the target match combination, the probability of lopsided matches can be reduced, maximizing the likelihood of balanced matches and thus achieving the goal of balanced matches. Therefore, the solution of this embodiment can reduce the probability of lopsided matches without affecting position allocation satisfaction, thereby improving the gaming experience.

[0109] The solutions provided in the embodiments of the present invention relate to technical fields such as artificial intelligence, data processing, and game technology, and are specifically described through the following embodiments.

[0110] Figure 1 This is a schematic diagram of an implementation environment provided by an embodiment of the present invention. (Refer to...) Figure 1 The implementation environment includes a server 101 and a terminal 102, which are directly or indirectly connected via wired or wireless communication. The server 101 and terminal 102 can be nodes in a blockchain, but this embodiment does not specifically limit their presence.

[0111] Server 101 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0112] Server 101 may integrate a game team-up function, or server 101 may be equipped with a game team-up device for implementing the game team-up function.

[0113] Server 101 has at least the following functions: adjusting the positions of participating objects in teams, predicting the win rate of the game combinations obtained after the position adjustment, determining the target game combination based on the predicted win rate, and forming game teams for participating objects based on the target game combination. For example, it can adjust the positions of each participating object in teams according to the expected team positions of each participating object in the game, obtain multiple game combinations, then concatenate the game feature information of all participating objects in each game combination to obtain the game feature information of each game combination, then predict the win rate based on the game feature information of each game combination to obtain the predicted win rate of each game combination, then determine the game combination corresponding to the one that meets the target win rate from all the predicted win rates as the target game combination, and form game teams for each participating object based on the target game combination.

[0114] Terminal 102 may include, but is not limited to, mobile phones, computers, smart voice interaction devices, smart home appliances, in-vehicle terminals, and smart wearable devices. Optionally, terminal 102 may have a game client installed. When a participant logs into the game client to play a team game such as a multiplayer online tactical competitive game, after the participant selects their desired team position and enters the game room, the game client can receive the game team result sent by server 101 and display the game team result through terminal 102. Optionally, a participant can also log into a lightweight game program through terminal 102. When a participant plays a team game such as a multiplayer online tactical competitive game based on the lightweight game program, after the participant selects their desired team position and enters the game room, the lightweight game program can receive the game team result sent by server 101 and display the game team result through terminal 102. A lightweight game program refers to a game program that can be used without downloading and installation; for example, a lightweight game program can be a mini-game. The game room is a virtual room created by server 101 for the target game. Before starting the game, participants can choose a virtual character or chat in the game room.

[0115] Reference Figure 2As shown, in one application scenario, when a participant logs into a multiplayer online tactical competitive game via terminal 102, selects their desired team position, and enters the game room, server 101 adjusts the positions of each participant in the current game match according to their desired team positions, resulting in multiple match combinations. At this point, server 101 concatenates the game feature information of all participants in each match combination to obtain the match feature information for each match combination. Then, server 101 predicts the win rate based on the match feature information of each match combination, obtaining the predicted win rate for each match combination. Server 101 identifies the match combination that matches the target win rate from all predicted win rates as the target match combination, and teams up the participants according to the target match combination. After teaming up the participants, server 101 sends the team formation results to terminal 102 for display. In response to receiving the team formation results from server 101, terminal 102 displays the team positions of each participant in the game room. At this time, the participants can select virtual characters. After all participants have selected their virtual characters, the multiplayer online tactical competitive game can begin.

[0116] It should be noted that in various specific embodiments of the present invention, when processing data related to the characteristics of the participating object, such as attribute information or sets of attribute information, is required, the permission or consent of the participating object will be obtained first. Furthermore, the collection, use, and processing of this data will comply with the relevant laws, regulations, and standards of the relevant countries and regions. In addition, when embodiments of the present invention need to obtain the attribute information of the participating object, separate permission or consent from the participating object will be obtained through pop-up windows or redirection to a confirmation page. Only after obtaining the separate permission or consent of the participating object will the necessary data related to the participating object for the normal operation of the embodiments of the present invention be obtained.

[0117] Figure 3 This is a flowchart of a game team-up method provided in an embodiment of the present invention. In this embodiment, a server is used as the execution subject for illustration. (Refer to...) Figure 3 The team formation method for this game includes, but is not limited to, steps 110 to 170.

[0118] Step 110: Obtain the desired team positions for each participant in the game.

[0119] In this step, the desired team position refers to the team position that a participant hopes to be assigned during a game. In team games, each team includes multiple participants. To provide a more flexible gaming experience for each participant, multiple game routes or attack routes are generally offered. Therefore, each participant can choose different game routes or attack routes according to their game preferences. To ensure the fairness of the game, different game routes or attack routes correspond to different team positions, and different team positions correspond to different virtual characters. For example, in some multiplayer online tactical competitive games, different game routes or attack routes such as "jungler," "top lane," "mid lane," and "bottom lane" can be offered to each participant. Before entering the game, participants can select their preferred desired team position so that the server can assign actual team positions to each participant based on their desired team positions, thereby improving the participant's satisfaction with their team position.

[0120] In some possible implementations, before a participant enters a game, the terminal can display different team positions for the participant to choose from via pop-ups or page redirects. After the participant determines their desired team position, the terminal can send the selected desired team position to the server, so that the server can adjust the positions of each participant's team based on the desired team position in subsequent steps to obtain multiple game combinations.

[0121] It's important to clarify that "jungler," "top lane," "mid lane," "bottom lane," and "support" are all game terms. "Jungler" refers to a non-lane position that primarily acquires experience and gold through jungle resources, aiming to reduce the enemy's advantage or increase their team's resources. "Top lane" refers to the attack lane located at the top of the map in a multiplayer online battle arena (MOBA) game. "Mid lane" refers to the attack lane located in the middle of the map in a MOBA game. "Bottom lane" refers to the attack lane located at the bottom of the map in a MOBA game.

[0122] Step 120: Adjust the positions of each participant in the team according to the expected team positions to obtain multiple match combinations.

[0123] In this step, since the expected team positions of each participant were obtained in step 110, in order to balance the game strength between the teams without affecting the participants' satisfaction with the position allocation and to reduce the probability of unequal matches, the positions of each participant's teams can be adjusted according to their expected team positions to obtain multiple match combinations. This allows subsequent steps to determine the match combination that can maintain the satisfaction with the position allocation and balance the game strength between the teams.

[0124] In some possible implementations, when adjusting the positions of each participant's teams based on their desired team positions to obtain multiple match combinations, the current team positions of each participant can be obtained first. Then, the positions of each participant's teams can be adjusted based on their current team positions and desired team positions to obtain multiple match combinations. Specifically, after a participant has confirmed their participation in a game and entered the game room, an initial team position can be assigned to each participant. After assigning initial team positions, the current team position of each participant may not match their desired team position. Therefore, the positions of each participant's teams can be adjusted based on their current team position and their desired team position.

[0125] In some possible implementations, the initial team position of each participant can be determined in different ways. For example, it can be determined by random assignment or based on the participant's historical game data (such as historical team positions). No specific limitation is made here. Specifically, when determining the initial team position based on the participant's historical game data, the participant's frequently used historical team positions can be identified first, and then these frequently used historical team positions can be used as the initial team positions.

[0126] In some possible implementations, different methods can be used to adjust the positions of participating players based on their current and desired team positions; these are not specifically limited here. For example, to obtain a wider variety of match combinations, the positions of all participating players can be adjusted based on their current and desired team positions. Alternatively, to improve the efficiency of obtaining multiple match combinations, the position adjustments can primarily be made for participating players whose current and desired team positions do not match. The following example illustrates the process of adjusting the positions of participating players whose current and desired team positions do not match. For example, suppose team A includes participants a1, a2, and a3, and team B includes participants b1, b2, and b3. For participants a1 and b3, their current team positions and desired team positions match. However, for participants a2, a3, b1, and b2, their current team positions and desired team positions do not match. Therefore, when adjusting the team positions of participants based on their current and desired team positions, the team positions of participants a2 and b3 can be adjusted to match either participant b1 or b2, the team position of participant a3 can be adjusted to match either participant b1 or b2, the team position of participant b1 can be adjusted to match either participant a2 or a3, and the team position of participant b2 can be adjusted to match either participant a2 or a3.

[0127] Step 130: Obtain game characteristic information for each participating entity.

[0128] To better reflect the gaming skills of each participant and enable subsequent steps to make more accurate win rate predictions based on their skill levels, the gaming characteristic information obtained in this step can be multi-dimensional. For example, in some possible implementations, the gaming characteristic information of the participants can include three types of gaming characteristic information. The first type of gaming characteristic information can be the participant's first number of game data within a first time period (e.g., the last 7 days) (e.g., 5 game data). The second type of gaming characteristic information can be the participant's second number of game data within a second time period (e.g., the last 30 days) (e.g., 15 game data). The third type of gaming characteristic information can be the participant's game data before starting the current game. The first category of game characteristic information can include various features such as kills, assists, gold, last hits, virtual characters used, win / loss ratio, rank, KDA (KILL DEATH ASSIST), AFK status, proficiency, and satisfaction level. The second category can include various features such as the proportion of characters in a team and their corresponding win rates, the top five most used virtual characters and their corresponding win rates, average number of likes, and total number of likes. The third category can include various features such as current hidden skill rating, initial team position, team position satisfaction, rank, and team-related hidden skill rating. By acquiring the game characteristic information of each participant, the true game skill and match status of each participant can be accurately determined, thereby improving the accuracy of subsequent win rate predictions based on game characteristic information.

[0129] It should be noted that kills, assists, gold, last hits, rank, KDA, AFK status, and hidden skill rating are all common gaming terms in this field. For explanations of these gaming terms, please refer to the relevant descriptions in related technical documents, which will not be elaborated here.

[0130] In some possible implementations, when obtaining the game characteristic information of each participant, the historical game information of each participant can be obtained first. Then, the historical game information of each participant is feature-encoded to obtain the game characteristic information of each participant. The historical game information of each participant in past game matches is stored on the server. By periodically analyzing this historical game information, the stability and rationality of the game can be determined, thereby enabling reasonable maintenance and optimization of the game and ensuring the gaming experience of the participants. Since the historical game information of each participant in past game matches is stored on the server, when obtaining the game characteristic information of each participant, the historical game information of each participant can be obtained from the server, and then feature-encoded to obtain the game characteristic information of each participant.

[0131] In some possible implementations, when encoding the historical game information of each participant to obtain their game feature information, the process can begin by acquiring multiple non-quantitative and multiple quantitative feature information for each participant based on their historical game information. Then, the multiple non-quantitative feature information is encoded to obtain multiple non-quantitative feature-encoded information, and the multiple quantitative feature information is also encoded to obtain multiple quantitative feature-encoded information. Finally, all non-quantitative and quantitative feature-encoded information for each participant is concatenated to obtain their game feature information. Non-quantitative feature information represents historical game information where relationships exist between different historical game matches, such as total kills, total assists, average win rate, and average likes, which change accordingly with statistics on historical game matches. Quantitative feature information represents historical game information where no relationships exist between different historical game matches, such as virtual character identifiers, team positions, match results, virtual character levels, and team factions, which only correspond to the current historical game match. In some embodiments, when concatenating all non-quantized feature encoding information and all quantized feature encoding information, the non-quantized feature encoding information can be concatenated end-to-end first, and then the quantized feature encoding information can be concatenated end-to-end first. Finally, the concatenated non-quantized feature encoding information and the concatenated quantized feature encoding information can be concatenated end-to-end again. Of course, depending on the specific needs of the application, different concatenation methods can be used to concatenate all non-quantized feature encoding information and all quantized feature encoding information; no specific limitation is made here. It should be noted that during feature encoding, pre-trained models such as One-Hot Encoding, word2vec, Embeddings from Language Models (ELMo), OpenAI-GPT, or Bidirectional Encoder Representation from Transformer (BERT) can be used to perform feature encoding processing. Among them, one-hot encoding, Word2vec, ELMo, OpenAI-GPT and BERT models are common models in the field of natural language processing. For the principles and structures of one-hot encoding, Word2vec, ELMo, OpenAI-GPT and BERT models, please refer to the relevant descriptions in related technologies, which will not be repeated here.

[0132] In some possible implementations, when obtaining multiple non-quantitative and multiple quantitative feature information of each participant based on their historical game information, multiple quantitative and multiple non-quantitative indicator information of each participant can be obtained from their historical game information first. Then, feature extraction is performed on the multiple quantitative indicator information of each participant to obtain multiple quantitative feature information, and feature extraction is performed on the multiple non-quantitative indicator information of each participant to obtain multiple non-quantitative feature information. Quantitative indicator information refers to historical game information where there is no correlation between different historical game matches, such as virtual character identifiers, team positions, match results, virtual character levels, team factions, etc., which only correspond to the current historical game match. Non-quantitative indicator information refers to historical game information where there is a correlation between different historical game matches, such as total kills, total assists, average win rate, average number of likes, etc., which change accordingly with the statistics of historical game matches. In some embodiments, only virtual character identifiers, team positions, match results, virtual character levels, and team factions can be used as quantitative indicator information, and all other historical game information besides quantitative indicator information can be used as non-quantitative indicator information. Of course, depending on the specific needs of the application, different methods can be used to divide historical game information into non-quantitative and quantitative indicator information; no specific limitations are made here. It should be noted that when extracting features from both quantitative and non-quantitative indicator information, pre-trained models such as One-Hot encoding, word2vec, ELMo, OpenAI-GPT, or BERT can be used; no specific limitations are made here.

[0133] Since non-quantized feature information is continuous, in order to effectively encode it, in some possible implementations, when encoding multiple non-quantized feature information of various participating objects to obtain multiple non-quantized feature encoded information, the multiple non-quantized feature information of each participating object can first be discretized to obtain multiple discrete feature information sets. Then, each non-quantized feature information in each discrete feature information set is encoded to obtain multiple non-quantized feature encoded information. Discretization refers to the process of dividing continuous features into discrete segments. For example, the original continuous features are divided into multiple feature intervals, and then each feature interval is mapped to a single feature value. Therefore, when discretizing multiple non-quantized feature information of various participating objects to obtain multiple discrete feature information sets, the number of discretization categories can be determined first. Then, based on the number of categories, each non-quantized feature information of each participating object is mapped to obtain a feature classification number for each non-quantized feature information. Next, the non-quantized feature information is categorized according to the feature classification number to obtain multiple discrete feature information sets. The number of categories is used to divide the multiple non-quantized feature information into feature intervals. When mapping non-quantitative feature information of each participating object based on the number of classifications, a mapping function between the number of classifications and non-quantitative feature information can be established. Then, the non-quantitative feature information can be mapped based on this mapping function. For example, a logarithmic function with the number of classifications as the base can be established, and the non-quantitative feature information can be mapped based on this logarithmic function. To illustrate the process of discretizing non-quantitative feature information, suppose the number of classifications is 20. A logarithmic function with base 20 (i.e., the mapping function) can be established. Each non-quantitative feature information is then input as an independent variable into this logarithmic function with base 20. The result is the feature classification number of each non-quantitative feature information. At this point, non-quantitative feature information with the same feature classification number can be grouped together to obtain multiple discrete feature information sets.

[0134] Step 140: Concatenate the game feature information of all participants in each game combination to obtain the game feature information of each game combination.

[0135] In this step, since the game feature information of each participant was obtained in step 130, the game feature information of all participants in each match combination can be concatenated to obtain the match feature information of each match combination. This allows subsequent steps to predict the win rate of each match combination based on the match feature information. Specifically, when concatenating the game feature information of all participants in each match combination to obtain the match feature information of each match combination, the current team position of each participant in each match combination can be determined first. The game feature information of all participants in each match combination can then be concatenated based on their team positions to obtain the match feature information of each match combination. In addition, in some embodiments, when splicing the game feature information of the participants based on their positions within the team, the game feature information of the participants can also be spliced ​​based on the team order. For example, assuming a certain match combination includes team A and team B, and both team A and team B include 5 participants, then when splicing the game feature information of all participants in this match combination, the game feature information of the 5 participants in team A can be spliced ​​first according to the order of their positions within the team, and the game feature information of the 5 participants in team B can be spliced ​​according to the order of their positions within the team. Then, the spliced ​​game feature information of team A and the spliced ​​game feature information of team B can be spliced ​​together according to the team order of team A and team B to obtain the match feature information of this match combination.

[0136] Step 150: Based on the game characteristic information of each game combination, predict the win rate to obtain the predicted win rate of each game combination.

[0137] In this step, since the game feature information of each game combination was obtained in step 140, the win rate can be predicted based on the game feature information of each game combination, so as to obtain the predicted win rate of each game combination. This allows subsequent steps to select the game combination that meets the target win rate from these predicted win rates and determine it as the target game combination.

[0138] In some possible implementations, when predicting win rates based on the game feature information of each game combination, the game feature information of each game combination can be input into a pre-trained win rate prediction model. After inputting the game feature information of each game combination into the pre-trained win rate prediction model, the win rate prediction model can first perform feature encoding on the game feature information of each game combination to obtain the game feature encoding information of each game combination. Then, the win rate prediction model performs win rate prediction based on the game feature encoding information of each game combination to obtain the predicted win rate of each game combination. Specifically, when the win rate prediction model performs feature encoding on the game feature information of each game combination to obtain the game feature encoding information of each game combination, the win rate prediction model can first perform vector embedding on the game feature information of each game combination to obtain the game feature embedding vector of each game combination. Then, the win rate prediction model performs feature encoding on the game feature embedding vector of each game combination to obtain the game feature encoding information of each game combination. In some embodiments, the match feature information may include lineup feature information, auxiliary feature information, and position feature information. In this case, when the win rate prediction model performs vector embedding on the match feature information of each match combination to obtain the match feature embedding vector for each match combination, for each match combination, the win rate prediction model can first perform vector embedding on the lineup feature information to obtain the lineup feature embedding vector, perform vector embedding on the auxiliary feature information to obtain the auxiliary feature embedding vector, and perform vector embedding on the position feature information to obtain the position feature embedding vector. Then, the win rate prediction model concatenates the lineup feature embedding vector, the auxiliary feature embedding vector, and the position feature information to obtain the match feature embedding vector. The lineup feature information represents the strength and synergy of the virtual characters in the team lineup. For example, the lineup feature information of a team represents the overall strength and synergy of all virtual characters in that team. The auxiliary feature information represents other information about the team besides strength and synergy. For example, the auxiliary feature information of a team represents the team's total kills, total assists, total minions, average rank, and average hidden MMR. Positional feature information represents the sequential positional relationship of game feature information. For example, the positional feature information of a certain team represents the positional relationship between the team's lineup feature information and auxiliary feature information.

[0139] In some possible implementations, the win rate prediction model can be an OmniNet model, a network model that utilizes omnidirectional attention to connect all tokens across the entire network via self-attention. The OmniNet model allows each token to process not only all other tokens in the same layer, but also all tokens across all layers of the network, rather than strictly maintaining a single horizontal receptive field. This global access of the OmniNet model enables tokens to have a complete view of the network, thus accessing the knowledge and intermediate representations of each token at each layer.

[0140] In some possible implementations, the structure of the win rate prediction model can be as follows: Figure 4 As shown, in Figure 4In this model, the win rate prediction model can include an embedding layer, a transform layer, an omnidirectional attention layer, and an output layer, wherein the embedding layer, transform layer, omnidirectional attention layer, and output layer are connected in sequence. The embedding layer can include a lineup embedding layer, an auxiliary embedding layer, and a position embedding layer. After the game feature information of the game combination is input into the win rate prediction model, the win rate prediction model first inputs the lineup feature information, auxiliary feature information, and position feature information from the game feature information into the lineup embedding layer, the auxiliary embedding layer, and the position embedding layer respectively. At this time, the lineup embedding layer can perform vector embedding on the lineup feature information to obtain the lineup feature embedding vector, the auxiliary embedding layer can perform vector embedding on the auxiliary feature information to obtain the auxiliary feature embedding vector, and the position embedding layer can perform vector embedding on the position feature information to obtain the position feature embedding vector. After obtaining the lineup feature embedding vector, the auxiliary feature embedding vector, and the position feature embedding vector, the win rate prediction model concatenates the lineup feature embedding vector, the auxiliary feature embedding vector, and the position feature embedding vector, and then inputs the concatenation result into the transformation layer for feature encoding to obtain the game feature encoding information of the game combination. It should be noted that since the concatenation result obtained by combining the lineup feature embedding vector, auxiliary feature embedding vector, and position feature embedding vector is sequence information, in some implementations, the transformation layer can use the Transformer model, which is adept at handling sequence information, as its basic architecture. Furthermore, the number of transformation layers can be set to 6, making the OmniNet model a lightweight model, easy to parallelize, and with a fast interface, thus meeting the performance requirements of online real-time prediction. In addition, since the OmniNet model uses the Transformer model as its basic architecture, it can better support model pre-training, facilitating rapid iteration and updates of model versions. After the transformation layer outputs the game feature encoding information, this information is input into the omnidirectional attention layer for subsequent information processing. The omnidirectional attention layer allows each token to "interact" with all tokens in the entire network, giving it a receptive domain across the entire width and depth of the network. The omnidirectional attention layer uses a self-attention mechanism as a meta-learner. This dense residual connection helps gradient propagation and enhances the model's expressive power, resulting in higher accuracy than traditional Transformer models. After the omnidirectional attention layer completes the processing of the game feature encoding information, it inputs the processing result into the output layer for win rate prediction. In some embodiments, the output layer can be a multilayer perceptron (MLP), which is a feedforward neural network model that can map multiple input datasets to a single output dataset.A multilayer perceptron can be viewed as a directed graph consisting of multiple node layers, each fully connected to the next. Except for the input node, each node is a neuron (or processing unit) with a non-linear activation function. Multilayer perceptrons are effective for win rate prediction tasks. Compared to traditional sequence deep learning models (such as recurrent neural networks and long short-term memory networks), the OmniNet model in this embodiment uses a self-attention mechanism instead of self-circular dependencies, which not only makes the model structure easier to parallelize but also enhances interpretability.

[0141] Step 160: Among all predicted win rates, the matchup that matches the target win rate is identified as the target matchup, where the target win rate represents a balanced game strength among the teams.

[0142] In this step, since the predicted win rates of each match combination were obtained in step 150, the predicted win rates that meet the target win rate can be determined from these predicted win rates. Then, the match combinations corresponding to the predicted win rates that meet the target win rate are determined as the target match combinations. Since the target win rate represents that the game strength between each team is balanced, the target match combinations can reduce the probability of lopsided matches. In addition, since each match combination is obtained after adjusting the positions between the teams of each participant according to the expected team positions of each participant, these match combinations can satisfy the position allocation satisfaction of each participant. Therefore, when the subsequent steps team up the participants according to the target match combinations, the probability of lopsided matches can be reduced without affecting the position allocation satisfaction, thereby improving the game experience.

[0143] It should be noted that among all predicted win rates, there may be only one or more that match the target win rate. Therefore, in some possible implementations, when there is only one predicted win rate that matches the target win rate, the game combination corresponding to that predicted win rate can be directly determined as the target game combination. When there are multiple predicted win rates that match the target win rate, a predicted win rate can be randomly selected from these predicted win rates, and then the game combination corresponding to that randomly selected predicted win rate can be determined as the target game combination.

[0144] Step 170: Team up the participants in the game according to the target match combination.

[0145] In this step, since the target match combination was determined in step 160, the participants can be teamed up according to the target match combination. This reduces the probability of lopsided matches without affecting the satisfaction of position allocation, thereby improving the gaming experience.

[0146] The process of the game team formation method in this embodiment will be described below with a specific example.

[0147] Reference Figure 5 As shown, Figure 5 This is a flowchart illustrating a specific example of a game team-up method. Figure 5 In this game, when multiple participants apply to join a match, they wait in the matchmaking queue for the game to start. Once the number of participants reaches the required number for a match to start, they are temporarily assigned to different teams, and each participant is temporarily assigned a corresponding initial team position. At this point, the positions of the participants are adjusted according to their desired team positions, resulting in multiple match combinations (such as...). Figure 5 The system first identifies game combinations (such as game combination 1, game combination 2, and game combination n). Then, it acquires the game feature information of each participant and concatenates this information to obtain the game feature information for each game combination. Next, it inputs this information into a pre-trained win rate prediction model to predict the win rate of each game combination. Finally, after obtaining the predicted win rates, it identifies the game combination that matches the target win rate from all predicted win rates and designates it as the target game combination. Then, it teams up the participants based on the target game combination. For example... Figure 6 As shown, Figure 6 This is an example illustrating the initial team positions assigned to each participant. Figure 6 Based on this, the positions of reference targets in both teams are adjusted. For example, swapping the positions of the second and third participants from the left team with the fourth and fifth participants from the right team yields a new matchup. After obtaining the new matchup, such as... Figure 7 As shown, win rate predictions can be performed for both teams separately to obtain the predicted win rate of the current matchup. By adjusting the positions of the participating teams, multiple matchups are obtained. Win rate predictions are then performed based on different matchups to obtain the predicted win rates for different matchups. Finally, the matchup that matches the target win rate from all predicted win rates can be identified as the target matchup. This allows subsequent steps to team up the participating teams based on the target matchup, for example... Figure 8 As shown, Figure 8 This is a diagram illustrating the team composition of participants based on the target matchup. Figure 6In comparison, it can be seen that by swapping the positions of the second and third participants of the left team with the first and second participants of the right team, the two teams after the position swap have balanced game strength, thus achieving the goal of balancing the game.

[0148] In this embodiment, the game teaming method, including steps 110 to 170 above, after obtaining the expected teaming positions and game feature information of each participant in the game, firstly, the positions of each participant in the team are adjusted according to the expected teaming positions to obtain multiple game combinations. Then, the game feature information of all participants in each game combination is spliced ​​to obtain the game feature information of each game combination. Then, the win rate is predicted according to the game feature information of each game combination to obtain the predicted win rate of each game combination. Next, the game combination that meets the target win rate among all the predicted win rates is determined as the target game combination, and the game teaming is performed for each participant according to the target game combination. Since the match combinations used for win rate prediction are obtained by adjusting the team positions of each participant based on their desired team roles, these match combinations can satisfy the position allocation satisfaction of each participant. Furthermore, since the target win rate represents a balanced game strength among teams, when the match combination corresponding to the target win rate is selected from all predicted win rates as the target match combination, and the participants are teamed up according to the target match combination, the probability of lopsided matches can be reduced, maximizing the likelihood of balanced matches and thus achieving the goal of balanced matches. Therefore, this embodiment can reduce the probability of lopsided matches without affecting position allocation satisfaction, thereby improving the gaming experience.

[0149] In some possible implementations, the win rate prediction model can be trained using the following steps: First, acquire training samples and label information. The training samples include game feature sample information of each participant in historical game matches. Then, concatenate the game feature sample information of all participants in historical game matches to obtain the match feature sample information of the historical game matches. Next, input the match feature sample information into the win rate prediction model to predict the win rate, obtaining the training predicted win rate of the historical game matches. At this point, calculate the loss value based on the training predicted win rate and label information, and then adjust the parameters of the win rate prediction model based on the loss value. Specifically, in the process of concatenating the game feature sample information of all participants in historical game matches to obtain the match feature sample information of the historical game matches, the team position information of each participant in the historical game matches can be determined first. Then, the game feature sample information of all participants in the historical game matches can be concatenated based on the team position information to obtain the match feature sample information of the historical game matches. Furthermore, when inputting game feature sample information into a win rate prediction model to obtain the training predicted win rate of historical game matches, the game feature sample information can be input into the win rate prediction model first, and feature encoding can be performed on the game feature sample information to obtain game feature encoded sample information. Then, win rate prediction can be performed on the game feature encoded sample information to obtain the training predicted win rate of historical game matches. In one embodiment, when performing feature encoding on the game feature sample information to obtain game feature encoded sample information, the game feature sample information can first be vector-embedded to obtain game feature sample embedding vectors, and then feature-encoded on the game feature sample embedding vectors to obtain game feature encoded sample information. In another embodiment, the game feature sample information may include lineup feature sample information, auxiliary feature sample information, and position feature sample information. In this case, when performing vector embedding on the game feature sample information to obtain the game feature sample embedding vector, the lineup feature sample information can be vector embedded first to obtain the lineup feature sample embedding vector, the auxiliary feature sample information can be vector embedded first to obtain the auxiliary feature sample embedding vector, and the position feature sample information can be vector embedded first to obtain the position feature sample embedding vector. Then, the lineup feature sample embedding vector, the auxiliary feature sample embedding vector, and the position feature sample information are concatenated to obtain the game feature sample embedding vector.It should be noted that the training process of the win rate prediction model can be either offline or online. When the training process is offline, offline game data can be used to train the win rate prediction model. After training, the win rate prediction model is deployed to a server so that it can be used for win rate prediction. When the training process is online, the win rate prediction model can be deployed to a server first, and then online game data can be acquired in real time. The online game data is then input into the win rate prediction model, achieving the goal of both training the win rate prediction model and using it for win rate prediction.

[0150] The game team formation method of this invention will be described in detail below with specific examples.

[0151] Reference Figure 9 As shown, Figure 9 This is a complete flowchart illustrating the process of forming a team in a game, using a specific example. Figure 9In this system, the server executing the game team formation method may include a model server 901, a database 902, and a feature processing platform 903. The win rate prediction model is deployed in the model server 901. Before using the win rate prediction model to predict the win rate of game combinations, game feature sample information and label information of each participant in historical game matches are obtained. Then, the win rate prediction model is trained using the game feature sample information and label information. After training is complete, the win rate prediction model is enabled in the model server 901. When participants enter a game room through game client 904, the server obtains the expected team positions and game characteristic information of each participant, and stores this information in database 902. At this time, feature processing platform 903 retrieves the expected team positions and game characteristic information of each participant from database 902 in real time, adjusts the team positions of each participant based on the expected team positions to obtain multiple match combinations, and then concatenates the game characteristic information of all participants in each match combination to obtain the match characteristic information of each match combination. Finally, it sends the match characteristic information of each match combination to model server 901, which then processes the received feature information. After the platform 903 sends the game feature information of each game combination, it calls the win rate prediction model to predict the win rate based on the game feature information of each game combination, obtaining the predicted win rate of each game combination. Then, the model server 901 determines the game combination that meets the target win rate from all the predicted win rates as the target game combination. After obtaining the target game combination, the server teams up the participants according to the target game combination. After completing the game teaming for all participants, the server sends the game teaming result to the game client 904. After receiving the game teaming result, the game client 904 performs secondary team matching for all participants according to the game teaming result (i.e., adjusts the positions of participants in different teams). Figure 10 and Figure 11 As shown, Figure 10 This is a predicted win rate distribution chart after using the game team formation method of this invention. Figure 11 This is a predicted win rate analysis chart after employing the game team formation method of this invention. According to... Figure 10 The data shows that after adopting the game team formation method of this invention, the overall prediction value conforms to a Gaussian distribution, and it can also provide intuitive prediction values ​​for extreme matches distributed on both sides. Figure 11The data shows that after adopting the game team formation method of this invention, a high accuracy rate can be achieved in predicting extreme matches, with an accuracy rate of over 90% for extreme matches and an overall accuracy rate of 65%. In addition, the predicted values ​​are also in line with expectations for balanced matches distributed in the middle. In other words, the game team formation method of this invention can significantly improve the quality of matches, reduce the probability of lopsided matches, and thus improve the game experience.

[0152] The following examples illustrate the application scenarios of the embodiments of the present invention.

[0153] It should be noted that the game team formation method provided in this embodiment of the invention can be applied to different application scenarios, such as game scenarios with automatic game quality balancing and game scenarios with manual game quality balancing. The following description will take game scenarios with automatic game quality balancing and game scenarios with manual game quality balancing as examples.

[0154] Scene 1

[0155] The game teaming method provided in this embodiment of the invention can be applied to game scenarios that automatically balance the quality of matches. Specifically, when a player logs into a multiplayer online tactical competitive game through a terminal, selects their desired team position, and enters the game room, the server automatically adjusts the positions of each player's team based on their desired team positions, resulting in multiple match combinations. At this time, the server concatenates the game feature information of all players in each match combination to obtain the match feature information of each match combination. Then, the server predicts the win rate based on the match feature information of each match combination to obtain the predicted win rate of each match combination. Next, the server selects the match combination that matches the target win rate from all predicted win rates as the target match combination and teams up the players according to the target match combination. After completing the game teaming for each player, the server sends the game teaming result to the terminal for display. In response to receiving the game teaming result sent by the server, the terminal displays the team positions of each player in the game room. At this time, players can select virtual characters. After all players have completed the selection of virtual characters, the multiplayer online tactical competitive game can begin.

[0156] Scene 2

[0157] The game team formation method provided in this invention can be applied to game scenarios where match quality is manually balanced. Specifically, when a player logs into a multiplayer online tactical competitive game via a terminal, selects their desired team position, and enters a game room, and the player has enabled the match quality balancing function, the server, based on the player's enabled match quality balancing function and the desired team positions of each player participating in the current game, adjusts the positions of each player's team, resulting in multiple match combinations. At this time, the server concatenates the game characteristic information of all players in each match combination to obtain the match characteristic information of each match combination. Then, the server, based on each... The server predicts the win rate of each match combination by analyzing its match characteristics. Then, it selects the match combination that matches the target win rate from all the predicted win rates and assigns it to the target match combination. Based on the target match combination, the server teams up the players. After teaming up the players, the server sends the teaming results to the terminal for display. In response to receiving the teaming results from the server, the terminal displays the team positions of each player in the game room. At this time, players can select virtual characters. After all players have selected their virtual characters, the multiplayer online tactical competitive game can begin.

[0158] It is understood that although the steps in the above flowcharts are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated in this embodiment, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the above flowcharts may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps.

[0159] Reference Figure 12 The present invention also discloses a game team-up device 1200, which can implement the game team-up method as described in the previous embodiments. The game team-up device 1200 includes:

[0160] Team position acquisition unit 1210 is used to acquire the expected team positions of each participant in the game;

[0161] The game combination acquisition unit 1220 is used to adjust the positions of each participant in the team according to the expected team position to obtain multiple game combinations.

[0162] The game feature acquisition unit 1230 is used to acquire game feature information of each participating object;

[0163] The game feature splicing unit 1240 is used to splice the game feature information of all participants in each game combination to obtain the game feature information of each game combination.

[0164] The game win rate prediction unit 1250 is used to predict the win rate based on the game characteristic information of each game combination, and obtain the predicted win rate of each game combination.

[0165] The target match determination unit 1260 is used to determine the match combination that meets the target win rate from all predicted win rates as the target match combination, wherein the target win rate represents that the teams have a balanced game strength.

[0166] Game match teaming unit 1270 is used to team up various participants in the game based on the target match combination.

[0167] In one embodiment, the game combination acquisition unit 1220 is further configured to:

[0168] Get the current team position of each participating object;

[0169] Based on the current team positions and expected team positions, the positions of each participant are adjusted between teams to obtain multiple match combinations.

[0170] In one embodiment, the game feature splicing unit 1240 is further configured to:

[0171] Determine the team positions of each participant in each match combination;

[0172] The game characteristic information of all participants in each game combination is obtained by splicing together their game characteristics information based on their positions within the team.

[0173] In one embodiment, the game win rate prediction unit 1250 is further configured to:

[0174] Feature encoding is performed on the game feature information of each game combination to obtain the game feature encoding information of each game combination;

[0175] Win rate prediction is performed based on the game feature encoding information of each game combination to obtain the predicted win rate of each game combination.

[0176] In one embodiment, the game win rate prediction unit 1250 is further configured to:

[0177] Vector embedding is performed on the game feature information of each game combination to obtain the game feature embedding vector of each game combination.

[0178] Feature encoding is performed on the game feature embedding vectors of each game combination to obtain the game feature encoding information of each game combination.

[0179] In one embodiment, the game feature information includes lineup feature information, auxiliary feature information, and position feature information; the game win rate prediction unit 1250 is further used for:

[0180] For each game combination, vector embedding is performed on the lineup feature information to obtain the lineup feature embedding vector, vector embedding is performed on the auxiliary feature information to obtain the auxiliary feature embedding vector, and vector embedding is performed on the position feature information to obtain the position feature embedding vector.

[0181] The lineup feature embedding vector, auxiliary feature embedding vector, and position feature information are concatenated to obtain the game feature embedding vector.

[0182] In one embodiment, the game feature acquisition unit 1230 is further configured to:

[0183] Obtain historical game information for each participant;

[0184] By encoding the historical game information of each participant, the game feature information of each participant can be obtained.

[0185] In one embodiment, the game feature acquisition unit 1230 is further configured to:

[0186] Based on the historical game information of each participant, obtain multiple non-quantitative feature information and multiple quantitative feature information of each participant;

[0187] Multiple non-quantized feature information of each participating object is feature-encoded to obtain multiple non-quantized feature-encoded information;

[0188] Multiple quantitative feature information of each participating object is feature-encoded to obtain multiple quantitative feature-encoded information;

[0189] By concatenating all non-quantized feature encoding information and all quantized feature encoding information of each participant, the game feature information of each participant is obtained.

[0190] In one embodiment, the game feature acquisition unit 1230 is further configured to:

[0191] Extract multiple quantitative and non-quantitative indicator information for each participant from their historical game information.

[0192] Feature extraction is performed on multiple quantitative indicators of each participating object to obtain multiple quantitative feature information of each participating object;

[0193] Feature extraction is performed on multiple non-quantitative indicators of each participant to obtain multiple non-quantitative feature information of each participant.

[0194] In one embodiment, the game feature acquisition unit 1230 is further configured to:

[0195] Discretize multiple non-quantitative feature information of each participating object to obtain multiple sets of discrete feature information;

[0196] Feature encoding is performed on each non-quantized feature information in each discrete feature information set to obtain multiple non-quantized feature encoding information.

[0197] In one embodiment, the game feature acquisition unit 1230 is further configured to:

[0198] Determine the number of categories;

[0199] Based on the number of classifications, the non-quantitative feature information of each participating object is mapped to obtain the feature classification number of each non-quantitative feature information;

[0200] Based on the feature classification number, each non-quantitative feature information is categorized to obtain multiple discrete feature information sets.

[0201] In one embodiment, the game team-up device 1200 further includes:

[0202] The sample acquisition unit is used to acquire training samples and label information, wherein the training samples include game feature sample information of each participant in historical game matches;

[0203] The sample splicing unit is used to splice the game feature sample information of all participants in the historical game matches to obtain the game feature sample information of the historical game matches.

[0204] The win rate prediction unit is used to input the game feature sample information into the win rate prediction model to predict the win rate and obtain the training prediction win rate of historical game games.

[0205] The loss calculation unit is used to calculate the loss value based on the training prediction win rate and label information;

[0206] The parameter correction unit is used to correct the parameters of the win rate prediction model based on the loss value.

[0207] In one embodiment, the sample splicing unit is further configured to:

[0208] Determine the in-team position information of each participant in historical game matches;

[0209] Based on the team's position information, the game feature sample information of all participants in the historical game is spliced ​​together to obtain the game feature sample information of the historical game.

[0210] In one embodiment, the win rate prediction unit is further configured to:

[0211] The game feature sample information is input into the win rate prediction model, and the game feature sample information is feature encoded to obtain the game feature encoded sample information.

[0212] Win rate prediction is performed on the game feature encoding sample information to obtain the training prediction win rate of historical game games.

[0213] In one embodiment, the win rate prediction unit is further configured to:

[0214] Vector embedding is performed on the game feature sample information to obtain the game feature sample embedding vector;

[0215] Feature encoding is performed on the embedding vector of the game feature samples to obtain the game feature encoded sample information.

[0216] In one embodiment, the game feature sample information includes lineup feature sample information, auxiliary feature sample information, and position feature sample information; the win rate prediction unit is further used for:

[0217] Vector embedding is performed on the lineup feature sample information to obtain the lineup feature sample embedding vector, vector embedding is performed on the auxiliary feature sample information to obtain the auxiliary feature sample embedding vector, and vector embedding is performed on the position feature sample information to obtain the position feature sample embedding vector.

[0218] The game feature sample embedding vector is obtained by concatenating the lineup feature sample embedding vector, the auxiliary feature sample embedding vector, and the position feature sample information.

[0219] It should be noted that since the game team-up device 1200 of this embodiment can implement the game team-up method of the previous embodiment, the game team-up device 1200 of this embodiment and the game team-up method of the previous embodiment have the same technical principle and the same beneficial effect. In order to avoid repetition, it will not be described again here.

[0220] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0221] Reference Figure 13 The present invention also discloses a game team-up device, the game team-up device 1300 comprising:

[0222] At least one processor 1301;

[0223] At least one memory 1302 is used to store at least one program;

[0224] When at least one program is executed by at least one processor 1301, the game team-up method as described in any of the preceding embodiments is implemented.

[0225] This invention also discloses a computer-readable storage medium storing a processor-executable program, which, when executed by a processor, is used to implement the game team-up method as described in any of the preceding embodiments.

[0226] This invention also discloses a computer program product, including a computer program or computer instructions, which are stored in a computer-readable storage medium. A processor of a computer device reads the computer program or computer instructions from the computer-readable storage medium and executes the computer program or computer instructions, causing the computer device to perform the game team formation method as described in any of the preceding embodiments.

[0227] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatuses.

[0228] It should be understood that in this invention, "at least one (item)" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0229] In the several embodiments provided by this invention, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the couplings or direct couplings or communication connections shown or discussed may be indirect couplings or communication connections through some interfaces, apparatuses, or units, and may be electrical, mechanical, or other forms.

[0230] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0231] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0232] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0233] The step numbers in the above method embodiments are set only for ease of explanation and do not limit the order of the steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

Claims

1. A game team forming method, characterized by, Includes the following steps: Obtain the desired team positions of each participant in the game; Based on the expected team positions, the positions of each participating entity are adjusted between teams to obtain multiple match combinations; Obtain game characteristic information for each of the participating objects; wherein, the game characteristic information includes the proportion and win rate of each participating object in each team position in historical game matches; Determine the team position of each participant in each of the aforementioned match combinations; The game feature information of all the participants in each game combination is spliced ​​together according to their positions in the team to obtain the game feature information of each game combination. Win rate prediction is performed based on the game feature information of each game combination to obtain the predicted win rate of each game combination. Among all the predicted win rates, the match combination that meets the target win rate is determined as the target match combination, wherein the target win rate represents that the teams have balanced game strength; The game teams are formed for each of the participants based on the target match combination.

2. The method of claim 1, wherein, The step of adjusting the positions of each participating entity between teams based on the expected team positions results in multiple match combinations, including: Get the current team position of each of the participating objects; Based on the current team position and the expected team position, the positions of each participating entity are adjusted between teams to obtain multiple match combinations. 3.The teaming method of claim 1, wherein, The step of predicting the win rate based on the game feature information of each game combination to obtain the predicted win rate of each game combination includes: The game feature information of each game combination is feature encoded to obtain the game feature encoding information of each game combination; Win rate prediction is performed based on the game feature encoding information of each game combination to obtain the predicted win rate of each game combination.

4. The method of claim 3, wherein, The step of performing feature encoding on the game feature information of each game combination to obtain the game feature encoding information of each game combination includes: Vector embedding is performed on the game feature information of each game combination to obtain the game feature embedding vector of each game combination; Feature encoding is performed on the game feature embedding vectors of each game combination to obtain the game feature encoding information of each game combination.

5. The method of claim 4, wherein, The game feature information includes lineup feature information, auxiliary feature information, and position feature information; The step of embedding the game feature information of each game combination into a vector to obtain the game feature embedding vector of each game combination includes: For each of the aforementioned game combinations, vector embedding is performed on the lineup feature information to obtain a lineup feature embedding vector, vector embedding is performed on the auxiliary feature information to obtain an auxiliary feature embedding vector, and vector embedding is performed on the position feature information to obtain a position feature embedding vector. The lineup feature embedding vector, the auxiliary feature embedding vector, and the position feature information are concatenated to obtain the game feature embedding vector.

6. The method of claim 1, wherein, The step of obtaining the game characteristic information of each of the participating objects includes: Obtain the historical game information of each of the aforementioned participants; The historical game information of each participating object is feature-encoded to obtain the game feature information of each participating object.

7. The method of claim 6, wherein, The step of encoding the historical game information of each participant to obtain the game feature information of each participant includes: Based on the historical game information of each participating object, obtain multiple non-quantitative feature information and multiple quantitative feature information of each participating object; The multiple non-quantized feature information of each of the participating objects is feature-encoded to obtain multiple non-quantized feature-encoded information; The multiple quantization feature information of each of the participating objects is feature-encoded to obtain multiple quantization feature-encoded information; By concatenating all the non-quantized feature encoding information and all the quantized feature encoding information of each participating object, the game feature information of each participating object is obtained.

8. The method of claim 7, wherein, The step of obtaining multiple non-quantitative feature information and multiple quantitative feature information of each participant based on the historical game information of each participant includes: Obtain multiple quantitative indicator information and multiple non-quantitative indicator information of each participating object from the historical game information of each participating object; Feature extraction is performed on multiple quantitative indicator information of each of the participating objects to obtain multiple quantitative feature information of each of the participating objects; Feature extraction is performed on multiple non-quantitative indicator information of each of the participating objects to obtain multiple non-quantitative feature information of each of the participating objects. 9.The teaming method of claim 7, wherein, The step of encoding multiple non-quantized feature information of each of the participating objects to obtain multiple non-quantized feature encoded information includes: Discretize the multiple non-quantized feature information of each of the participating objects to obtain multiple discrete feature information sets; Each of the non-quantized feature information in each of the discrete feature information sets is feature-encoded to obtain multiple non-quantized feature-encoded information.

10. The game team formation method according to claim 9, characterized in that, The discretization of multiple non-quantized feature information of each of the participating objects yields multiple sets of discrete feature information, including: Determine the number of categories; Based on the number of classifications, the non-quantitative feature information of each of the participating objects is mapped to obtain the feature classification number of each of the non-quantitative feature information; The non-quantitative feature information is categorized according to the feature classification number to obtain multiple discrete feature information sets.

11. A game team-up device, characterized in that, include: The team position acquisition unit is used to acquire the expected team positions of each participant in the game. The match combination acquisition unit is used to adjust the positions of each participating object between teams according to the expected team positions, so as to obtain multiple match combinations. The game feature acquisition unit is used to acquire game feature information of each of the participating objects; wherein, the game feature information includes the proportion of each participating object in each team position and the win rate in historical game matches; A game feature splicing unit is used to determine the team position of each participating object in each game combination; and splice the game feature information of all participating objects in each game combination according to the team position to obtain the game feature information of each game combination. The game win rate prediction unit is used to input the game feature information of each game combination into the win rate prediction model to predict the win rate and obtain the predicted win rate of each game combination. The target match determination unit is used to determine the match combination that meets the target win rate from all the predicted win rates as the target match combination, wherein the target win rate represents that the teams have balanced game strength; The game match teaming unit is used to team up the participants in the game according to the target match combination.

12. A game team-up device, characterized in that, include: At least one processor; At least one memory for storing at least one program; The game teaming method as described in any one of claims 1 to 10 is implemented when at least one of the programs is executed by at least one of the processors.

13. A computer-readable storage medium, characterized in that, It stores a processor-executable program, which, when executed by the processor, is used to implement the game team-up method as described in any one of claims 1 to 10.

14. A computer program product, comprising a computer program or computer instructions, characterized in that, The computer program or the computer instructions are stored in a computer-readable storage medium, the processor of the computer device reads the computer program or the computer instructions from the computer-readable storage medium, and the processor executes the computer program or the computer instructions, causing the computer device to perform the game team formation method as described in any one of claims 1 to 10.

Images