Method, apparatus, electronic device, and storage medium for real-time display of game winning rate

Abstract

The present application provides a method, apparatus, electronic device, and storage medium for real-time display of game win rates, relating to the field of game technologies. The method includes: obtaining player game characteristics of a first player character at different historical moments in a current game session, inputting the player game characteristics into a target game win rate prediction model corresponding to the current game session that has been pre-trained (obtained by training with player game characteristics of historical player characters at different historical moments in historical game sessions, and the player game characteristics at different historical moments have different contribution degree time effect weights for representing the player game characteristics at different historical moments), obtaining a first current game win rate, and displaying it on the graphical user interface of a first terminal device. The present application can perform win rate prediction in real time through the target game win rate prediction model, improving the accuracy of real-time display of game win rates; at the same time, it enables players to make accurate decisions during the game and saves players' time.

Description

Technical Field

[0001] The present application relates to the technical field of machine learning, and in particular, to a method, device, electronic device, and storage medium for real-time display of game win rates. Background Art

[0002] Currently, some games will prompt the clearance probability when players play the game, such as "there is a certain percentage probability of clearing this game". For the method of predicting and displaying the win rate in the game, related technologies usually predict the current win rate of the player according to the relative importance of the pre-set prediction time points in the game for game feature types (such as lineup features, economic features, minion wave features, etc.). However, in this way, the importance of data in prediction only depends on the static prediction time point, resulting in a low accuracy of the win rate prediction result. Summary of the Invention

[0003] The present application provides a method, device, electronic device, and storage medium for real-time display of game win rates, which can perform win rate prediction in real time through a target game win rate prediction model, improving the accuracy of real-time display of game win rates; at the same time, it enables players to make accurate decisions in the game and saves players' time.

[0004] In a first aspect, the present invention provides a method for real-time display of game win rates, including:

[0005] Obtain the player game features of the first player character at different historical moments in the current game session; the first player character is the player character controlled by the first terminal device;

[0006] Input the player game features at different historical moments into the target game win rate prediction model corresponding to the current game session that has been pre-trained, and obtain the first current game win rate; the target game win rate prediction model is trained by the player game features of historical player characters at different historical moments in historical game sessions; wherein, the player game features at different historical moments in the target game win rate prediction model have different time-dependent weights, and the time-dependent weights represent the contribution degrees of the player game features at different historical moments;

[0007] Display the first current game win rate on the graphical user interface of the first terminal device.

[0008] In a second aspect, the present invention provides a device for real-time display of game win rates, including:

[0009] A feature acquisition module, configured to obtain the player game features of the first player character at different historical moments in the current game session; the first player character is the player character controlled by the first terminal device;

[0010] A winning rate prediction module, configured to input the player game features at different historical moments into a target game winning rate prediction model corresponding to the current game round that has been pre-trained, to obtain a first current game winning rate; the target game winning rate prediction model is trained by using player game features of a historical player character at different historical moments in historical game rounds; wherein, the player game features at different historical moments in the target game winning rate prediction model have different time-effect weights, and the time-effect weights represent the contribution degrees of the player game features at different historical moments.

[0011] A winning rate display module, configured to display the first current game winning rate on a graphical user interface of the first terminal device.

[0012] In a third aspect, the present invention provides an electronic device, including a processor and a memory, where the memory stores computer executable instructions that can be executed by the processor, and the processor executes the computer executable instructions to implement the method for real-time display of game winning rate according to any one of the foregoing embodiments.

[0013] In a fourth aspect, the present invention provides a computer-readable storage medium, where the computer-readable storage medium stores computer executable instructions, and when the computer executable instructions are called and executed by a processor, the computer executable instructions cause the processor to implement the method for real-time display of game winning rate according to any one of the foregoing embodiments.

[0014] The method, device, electronic device, and storage medium for real-time display of game winning rate provided by this application first obtain player game features of a first player character controlled by a first terminal device at different historical moments in the current game round, input the player game features at different historical moments into a target game winning rate prediction model corresponding to the current game round that has been pre-trained, to obtain a first current game winning rate, so as to display the first current game winning rate on a graphical user interface of the first terminal device. Among them, the target game winning rate prediction model is trained by using player game features of a historical player character at different historical moments in historical game rounds; wherein, the player game features at different historical moments in the target game winning rate prediction model have different time-effect weights, and the time-effect weights represent the contribution degrees of the player game features at different historical moments. The above method determines the first current game winning rate corresponding to the current game round through the target game winning rate prediction model trained by using player game features of a historical player character at different historical moments in historical game rounds, which can make the prediction result of the obtained model closer to the current game and improve the accuracy of the game; by predicting and displaying the game winning rate in real time, players can make accurate decisions during the game. Especially when it is prompted that the game winning rate is relatively low, players can make decisions to quit or surrender as soon as possible, thus saving players' time. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] To more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0016] Figure 1 It is a flowchart of a method for real-time display of game win rate provided by an embodiment of the present application;

[0017] Figure 2 It is a training flowchart of a target game win rate prediction model provided by an embodiment of the present application;

[0018] Figure 3 It is a flowchart of a specific method for real-time display of game win rate provided by an embodiment of the present application;

[0019] Figure 4 It is a structural diagram of a device for real-time display of game win rate provided by an embodiment of the present application;

[0020] Figure 5 It is a structural diagram of an electronic device provided by an embodiment of the present application. Specific Embodiments

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some, rather than all, of the embodiments of the present application. Usually, the components of the embodiments of the present application described and illustrated in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the present application claimed, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0022] In one embodiment of the present disclosure, the method for real-time display of game win rate can run on a local terminal device or a server. When the method for real-time display of game win rate runs on the server, the method can be implemented and executed based on a cloud interaction system, where the cloud interaction system includes a server and client devices.

[0023] In an alternative embodiment, various cloud applications can run under the cloud interaction system, such as cloud games. Taking cloud games as an example, cloud games refer to a game mode based on cloud computing. In the operation mode of cloud games, the running entity of the game program and the presenting entity of the game screen are separated. The storage and operation of the method for real-time display of game winning rate are completed on the cloud game server. The function of the client device is to receive and send data and present the game screen. For example, the client device can be a display device with data transmission function near the user side, such as a mobile terminal, a television, a computer, a palm computer, etc.; however, the information processing is performed by the cloud game server in the cloud. When playing a game, the player operates the client device to send an operation instruction to the cloud game server. The cloud game server runs the game according to the operation instruction, encodes and compresses data such as the game screen, returns it to the client device through the network, and finally, the game screen is decoded and output through the client device.

[0024] In an alternative embodiment, taking a game as an example, the local terminal device stores a game program and is used to present the game screen. The local terminal device is used to interact with the player through the graphical user interface, that is, conventionally, the game program is downloaded and installed on the electronic device and run. The way the local terminal device provides the graphical user interface to the player can include various methods. For example, it can be rendered and displayed on the display screen of the terminal, or provided to the player through holographic projection. For example, the local terminal device can include a display screen and a processor. The display screen is used to present the graphical user interface, which includes the game screen. The processor is used to run the game, generate the graphical user interface, and control the display of the graphical user interface on the display screen.

[0025] In a possible embodiment, the embodiment of the present invention provides a method for real-time display of game winning rate. A graphical user interface is provided through a terminal device, where the terminal device can be the aforementioned local terminal device or the client device in the aforementioned cloud interaction system. In this embodiment, the terminal device includes a first terminal device and a second terminal device. Among them, the first terminal device is the terminal device used by the first player, and the second terminal device is the terminal device used by the teammate players of the first player. And usually there are multiple second terminal devices. For the convenience of description in this embodiment, all the terminal devices used by the teammate players are referred to as the second terminal devices.

[0026] The method for real-time display of game winning rate provided in this embodiment is applied to the first terminal device. Refer to Figure 1 As shown, the method mainly includes the following steps:

[0027] Step S102, obtaining the player game characteristics of the first player character at different historical moments in the current game session; the first player character is the player character controlled by the first terminal device.

[0028] The above-mentioned player game features are such that after the player enters the game or game instance, the player game features may include features related to the outcome of the game, such as the player's current health, current combat power, remaining amount of current health potions, remaining amount of current mana potions, inventory items, player experience, player economy, etc.

[0029] After the player enters the game or game instance, the first terminal device obtains the player game features at different historical moments in the player's current game session at regular intervals. When obtaining the features, periodic acquisition of player game features can be performed at a specified frequency, such as 1 time / min, 3 times / min, 5 times / min, etc. This periodicity can be periodic acquisition at the same time interval or periodic acquisition at different time intervals. For example, the player's various game features during the game can be obtained at a time interval with the same frequency of 3 times / min, or a larger frequency of 5 times / min can be used at the start of the game, and as the game progresses, the features are obtained at 3 times / min. In actual implementation, the player game features can be obtained at the same time interval or different time intervals according to the specific situation.

[0030] Step S104: Input the player game features at different historical moments into the target game win rate prediction model corresponding to the current game session that has been pre-trained, and obtain the first current game win rate.

[0031] The above-mentioned target game win rate prediction model is trained by the player game features of historical player characters at different historical moments in historical game sessions; among them, the player game features at different historical moments in the target game win rate prediction model have different time-effect weights, and the time-effect weight represents the contribution degree of the player game features at different historical moments.

[0032] In a possible implementation manner, the game server communicates with the game client, and the game server trains the target game win rate prediction model at regular intervals. For example, the game server can train the model every day, and uses the game data of a preset time period before that day as a sample set to update and train the model, so that the trained target game win rate prediction model can better conform to the real situation of the game, and the prediction result obtained through this target game win rate prediction model will be more accurate.

[0033] When the first player logs in to the first terminal device, obtain the target game win rate prediction model that has been newly trained by the game server. The target game win rate prediction model can be a machine learning model, such as a CatBoost model. In practical applications, other models can also be selected. This is only an example here and is not specifically limited.

[0034] After obtaining the player's game characteristics each time, input the obtained player's game characteristics into the target game win rate prediction model pulled latest, and output the first current game win rate through the model. When obtaining the player's game characteristics periodically, correspondingly, the current game win rates corresponding to the player's game characteristics obtained in each period will also be obtained. When the current game win rate is obtained, the game win rate obtained based on the player's game characteristics obtained in the previous period will be replaced. That is, the current game win rate displayed at the current moment is the real-time win rate predicted corresponding to the player's game characteristics in the previous period closest to the current moment.

[0035] Step S106, display the first current game win rate on the graphical user interface of the first terminal device.

[0036] The graphical user interface of the first terminal device can display the first current game win rate predicted by the target game win rate prediction model in real time. For the target game win rate prediction model to predict the win rate according to a periodic frequency, the first terminal device can display according to the game win rate obtained from the previous prediction during the time period between two predictions. For example, the first current game win rate predicted by the target game win rate prediction model at the current moment is 65%, and the frequency of the model prediction is 1 time / min. Then the currently displayed is 65%, and within 1 min before the next prediction result comes out, the displayed result is still 65%.

[0037] In order to enable the game win rate to provide a reference for the player's game decision-making, the prediction frequency of the model can be adaptively adjusted, so that the first current game win rate displayed on the graphical user interface of the first terminal device can be more real-time. In one implementation, the corresponding win rate prediction and display can be performed after the player triggers an event that affects the game win rate. For example, when the player's current health or current economy grows, the game win rate is predicted through the increased player's game characteristics to display the win rate corresponding to the event that affects the game win rate.

[0038] The method for real-time display of the game win rate provided by the embodiments of the present application determines the first current game win rate corresponding to the current game match through the player's game characteristics of the historical player character at different historical moments in the historical game match. By predicting and displaying the game win rate in real time, players can make accurate decisions in the game. Especially when it is prompted that the game win rate is low, players can make decisions to exit or surrender as soon as possible, thus saving the players' time.

[0039] In order to illustrate the way to improve the prediction accuracy in this embodiment, first, the target game win rate prediction model in the embodiments of the present application will be described. Refer to Figure 2 As shown, the training steps of the target game win rate prediction model mainly include the following steps:

[0040] Step S202, obtain the game historical sample sets at different historical moments of each historical game session.

[0041] The above game historical sample sets include multiple pieces of historical game data; the historical game data includes at least player game characteristics, and the player game characteristics include one or more of the current health value, the current combat power value, and the remaining amount of current health potions. The health potion is an item that can be used to extend the health value, and the remaining amount of the health potion can be obtained through the data in the backpack.

[0042] In an optional implementation manner, the above historical game data is obtained according to the data generation time. For example, the game historical sample set is a sample set composed of historical game data in a preset time period before the current moment using timestamps. The game historical sample sets of multiple preset time periods with different durations can include, for example, time periods of 1 day, 5 days, 7 days, 15 days, etc. before the current moment, or sample sets composed of data corresponding to a certain time period in each day within the preset number of days, such as 18:00 - 22:00, 21:00 - 24:00, etc. By selecting the sample sets composed of data from multiple time periods and training the game win rate prediction model with each sample respectively, target game win rate prediction models for different time periods can be obtained, so as to perform win rate prediction targeted and improve the model prediction ability.

[0043] In order to obtain a target game win rate prediction model with relatively better prediction effect, in an optional implementation manner, the game historical sample sets at different historical moments of each historical game session within multiple historical preset time periods can be obtained first. For example, three game historical sample sets of different time periods are pre-selected to train the model respectively. For example, historical game data with n1 = 14, n2 = 3, and n3 = 30 are respectively selected to form the game historical sample sets, and the model is trained according to the corresponding game historical sample sets. Through this method, different training effects can be obtained, that is, the prediction effects of the models trained with different time periods are different. Thus, the model with the best effect among multiple trained models can be selected and pushed to the game client to improve the win rate prediction effect in the game client.

[0044] In an optional implementation manner, the server can also select the model trained according to the corresponding time period and push it according to the time when the player logs in to the game. For example, if the player logs in to the game at 8 pm, the model trained by the game server according to the game historical sample set in the 18:00 - 22:00 time period can be pushed to the game client, so that the prediction result of the game win rate is more in line with the win rate result in the same time period in the historical game time.

[0045] In order to ensure the effect of model training, each piece of historical game data should at least include a player identifier, a game version identifier, a timestamp, a data acquisition sequence identifier, a clearance flag, and player game characteristics. Among them, the player game characteristics can include one or more of the current health value, the current combat power value, the remaining amount of current health potions, and the remaining amount of current mana potions. The above game versions can include, for example, the main game version, dungeon games, etc.

[0046] Step S204: Determine the aging weight corresponding to the historical game data according to the data acquisition sequence identifier of each piece of historical game data in the game historical sample set.

[0047] In one implementation, the aging weight corresponding to the historical game data can be determined according to the data acquisition sequence identifier of each piece of historical game data and a preset weight threshold. The data acquisition sequence identifier is in a proportional relationship with the elapsed time of the game corresponding to the game end moment. During the game, the above historical game data obtained by the game client can be acquired according to the data generation time since the game started. For example, a dungeon game lasts for 29 minutes, and the game client collects one piece of training data per minute, so a total of 29 pieces are collected, and data acquisition sequence identifiers are marked for the 29 pieces of data: Data 1, Data 2, Data 3... Data 28, Data 29.

[0048] Furthermore, considering that the closer to the end of the game, the more relevant the data is to the game result, the aging weight corresponding to the historical game data can be determined according to the data acquisition sequence identifier of each piece of historical game data and a preset weight threshold; among them, the aging weight corresponding to the historical game data closer to the end of the game is larger, and the game end moment includes the clearance moment, the failure moment, or the early exit moment.

[0049] Taking a dungeon game as an example, the game result of a game, such as victory (clearance), failure, or early exit, has a certain correlation with the elapsed time of the game. The data farther away from the end of the dungeon game contributes less to the prediction of the game win rate. Therefore, in one implementation, the above aging weight can be assigned using a decaying weight. The above preset weight threshold can be set numerically according to actual needs, denoted as M, and the data acquisition sequence identifier is denoted as j. Then, in one example, M - j can be used to represent the aging weight, and M - tj can be used to represent it (where t is a constant). In short, it only needs to satisfy that the aging weight of the data farther away from the end of the dungeon game is smaller.

[0050] Step S206: Train the initialized game win rate prediction model based on the aging weight and historical game data until the model converges to obtain the target game win rate prediction model.

[0051] In one implementation, in order to accurately predict the winning rates in different time periods, there are multiple initialized game winning rate prediction models, and the number of initialized game winning rate prediction models corresponds to the number of historical preset time periods.

[0052] When performing model training, for each initialized game winning rate prediction model, step 206 may further include the following steps 1.1 and 1.2:

[0053] Step 1.1, based on the time effect weight corresponding to the historical game data within the corresponding historical preset time period and the historical game data within the historical preset time period, train the corresponding initialized game winning rate prediction model until the corresponding model converges, obtaining multiple trained game winning rate prediction models.

[0054] Step 1.2, based on the model loss values corresponding to the multiple trained game winning rate prediction models, determine the target game winning rate prediction model.

[0055] Regarding step 1.1, in specific implementation, it may further include the following steps 1.1.1 to 1.1.3:

[0056] Step 1.1.1, for the initialized game winning rate prediction model trained with the historical game data corresponding to each historical preset time period, input the player game features and the corresponding whether cleared flag into the initialized game winning rate prediction model for training to obtain an initial loss value.

[0057] After determining the time effect weight, the contribution of each piece of historical game data to model training can also be determined accordingly. In an optional implementation, step S206 may further include the following steps:

[0058] Step 1, input the player game features and the corresponding whether cleared flag into the initialized game winning rate prediction model for training to obtain an initial loss value.

[0059] Generally, the mathematical form of a linear model:

[0060] y = f(x) = w x + c = w1 * x1 + w2 * x2 +... + wn * xn + c;

[0061] Among them, y is the prediction target (also called the label, called label), f is the model (i.e., a function), and the x vector represents the features. For example, y is the whether cleared flag, and x is features such as the current health value, current combat power value, current remaining amount of health potions, and current remaining amount of mana potions. w is the weight vector of the model, indicating the importance of each feature.

[0062] The above clearance indication at least includes the indication corresponding to early exit; among them, the indication corresponding to early exit is characterized by a probability, and the probability is in a proportional relationship with the elapsed time of the game corresponding to the early exit moment.

[0063] The indication of whether the above game is cleared, in addition to the indication corresponding to early exit, also includes the indication corresponding to game clearance and the indication corresponding to game failure. For example, the indication corresponding to game clearance can be represented by Y, the indication corresponding to game failure can be represented by N, and the indication corresponding to early exit can be represented by Q.

[0064] In a possible implementation manner, before the training data is input into the model for model training when the game is cleared, the label value (label, that is, the target of the model task) is processed as "1.0", and when the game fails, the label is processed as "0.0".

[0065] For the situation of early exit from the game, it is more applied to the game process of the dungeon. When the player finds that the probability of clearing the dungeon is low during the game process, in order to avoid wasting too much time, the player can exit early so as to play the dungeon again when there is enough time. Therefore, in this embodiment, the indication corresponding to early exit can be characterized by a probability, and the probability is in a proportional relationship with the elapsed time of the game when the player exits. In one example, the probability of early exit is processed as "1.0" with P_cehua (the value of P_cehua ranges from 0% to 100%), and is processed as "0.0" with the probability 1 - P_cehua. Among them, 1.0 assumes "exit" as "victory", and 0.0 assumes "exit" as "failure". If it is assumed as "victory", then the trained model will give a prediction result that is more inclined to tell the player that they will win, so the player may play for a longer time.

[0066] During model training, the P_cehua can be set accordingly according to the planning requirements of the designers. If the game planner needs to set the player to play in this game / dungeon for a longer time, P_cehua can be appropriately increased; if the game planner needs to set the player to exit as soon as possible, P_cehua can be appropriately decreased.

[0067] By inputting the player's game characteristics and the corresponding indication of whether the game is cleared in the historical game data into the initialized game win rate prediction model, the final navigation weight vector w can be determined through continuous learning to ensure that the initial loss value corresponding to the game win rate y is the smallest under the weight vector w.

[0068] The above initial loss value is determined by a loss function, and the loss function is used to measure the quality of the model prediction result. For example, loss = (y - y1)*2, where y1 is the predicted value of the model and y is the actual value. The gap between the true value and the predicted value can be measured through the loss function loss.

[0069] When performing model training, the weight vector w is usually initialized to 0. Then, according to the formula of the loss function loss, the true x vector and the true y value of the training data are input, so that the w vector changes in the direction of "decreasing loss". Until, when new training data is input and the loss cannot continue to decrease, that is, the loss function converges, the model training stops, and the minimum loss at this time (i.e., the initial loss value of this embodiment) is obtained.

[0070] Step 1.1.2: Determine the model loss value corresponding to the game history sample set based on the timeliness weight corresponding to each game history data and the initial loss value.

[0071] Usually during model training, to improve training efficiency, the data input into the model is in batches. Therefore, when calculating the loss value, the average loss value is directly calculated according to the batch data as the loss value of model training. However, in this embodiment, considering that the win rate prediction is related to the game progress time, that is, the determination of this loss value is also related to the acquisition order of game history data. Therefore, in this embodiment, the timeliness weight corresponding to each game history data and the initial loss value are jointly used to determine the final loss value of each batch input into the model.

[0072] In an optional implementation manner, when calculating the model loss value, considering that the contributions of each game history data are different, the timeliness weights of each game history data are also different. Correspondingly, when calculating the model loss value, loss_avg = sum(v loss) / sum(v) can be used, where loss_avg is the model loss value, the v vector is the timeliness weight of each game history data, loss is the initial loss value. It should be noted that the timeliness weight v and the feature weight w corresponding to each player game feature above are not the same weight.

[0073] Step 1.1.3: Determine the corresponding trained game win prediction model based on the model loss value obtained from each model training.

[0074] In an optional implementation manner, when the determined target loss value converges, the weight corresponding to the player game feature obtained is the feature weight at which the model converges. Thus, when determining this feature weight, it can be determined that the model has completed training, and the target game win rate prediction model is obtained.

[0075] Through the above steps, the best target game win rate prediction model can be trained. Thus, the first terminal device can, in response to the operation of the first player logging in to the first terminal device, obtain the target game win rate prediction model with the minimum target loss value from the game server.

[0076] In addition to the above situation where the model convergence is determined by the loss value, the embodiments of the present application can also use the prediction accuracy rate, the number of training times, etc. as the conditions for judging the model convergence. In one implementation manner, the condition for the model convergence can be that the model prediction accuracy rate reaches a specified accuracy threshold. For example, it is preset that if the lowest accuracy rate of the test set can reach 90% during the model training, it is determined that the model converges. In another implementation manner, the condition for the model convergence can also be that the number of model training reaches a specified number threshold. For example, it is preset that the number of model training is 100,000 times. When the number of training times reaches 100,000 times, it is determined that the model converges.

[0077] For the above method of obtaining historical game data for different time periods, there is a corresponding initialized game win rate model for training, so as to select the optimal model from multiple trained target game win rate prediction models and push it to the game client.

[0078] Regarding the above step 1.2, the game win rate prediction model with the smallest model loss value among the multiple trained game win rate prediction models is determined as the target game win rate prediction model. This method can make the prediction accuracy rate of the finally determined target game win rate prediction model higher and the model effect better.

[0079] For the situation of multiple players forming a team, in order to conform to the overall game win rate of the whole team and give a macro win rate prompt, in one implementation manner, the second current game win rate of the second player character in the current game session can also be obtained, and the game win rate range is determined based on the first current game win rate and the second current game win rate, and then the game win rate range is displayed on the graphical user interface of the first terminal device.

[0080] The above second player character is a player character operated by a second terminal device. The number of second player characters is at least one, and the second player character is a character in the same game camp as the first player character.

[0081] The second player is a teammate player in the same game process as the first player. When the second player logs in to the second terminal, the target game win rate prediction model pre-trained by the game server is obtained. And during the game, each second terminal also periodically obtains the corresponding current player game characteristics of the second player, and determines the game win rate of the second player through the target game win rate prediction model. The method for the second player to determine the second current game win rate is the same as that of the first player, except that the execution entity is changed from the first terminal device controlled by the first player to the second terminal device controlled by the second player. Other methods are all referred to the above embodiments and will not be elaborated here.

[0082] After each second terminal determines the current game win rate, it will share its own game win rate to the game clients of the teammates in the same team, so that each game client receives the game win rate predicted corresponding to the player game characteristics of each player in its own team.

[0083] In a possible implementation manner, when determining the first current game win rate corresponding to the first player and the second current game win rates corresponding to each second player, the maximum value and the minimum value of the win rates can be determined from the first current game win rate and the multiple second current game win rates to determine the game win rate range. For example, when there are 5 people in the game team, the corresponding current game win rates are 35%, 42%, 41%, 73% and 45% respectively, then the determined and displayed game win rate range is 35% - 73%.

[0084] After determining the game win rate range, it can be displayed by directly reading the range, such as the predicted win rate of the current game / copy is M% - N% (where M and N are the minimum win rate and the maximum win rate determined by the first current game win rate and the multiple second current game win rates); it can also be displayed by using a bar identifier, a circular identifier, etc., and the win rate range is marked with a more prominent color or style, so that players can intuitively know the current game win rate range.

[0085] Furthermore, when the minimum value of the game win rate range is less than the preset threshold, a surrender request is initiated through the game client corresponding to the minimum value of the game win rate range, so as to determine the game process based on the number of surrender terminals.

[0086] In an alternative implementation manner, the above-mentioned surrender request initiated through the game client corresponding to the minimum value of the game win rate range is an automatic surrender request initiated by the game client. Surrender controls are displayed on the graphical user interfaces provided by all game clients. Within a certain period of time after initiating the surrender (such as 20 seconds), if the vote exceeds half of the people, the game or the game copy will automatically end; if the number of votes is less than half of the people, the game or the game copy can continue. In this example, the determination of the number of votes can be set according to actual needs, and this is only an example here and is not specifically limited.

[0087] In this embodiment, the game server trains the target game win rate prediction model, and the game client predicts the game win rate. The win rate can be predicted in real time through the target game win rate prediction model, which improves the accuracy of real-time display of the game win rate; at the same time, it enables players to make accurate decisions during the game and saves players' time.

[0088] In one example, this embodiment also provides a specific method for real-time display of the game win rate. Taking the win rate prediction of the game copy as an example, it is executed by the game server and the game client. See Figure 3As shown, the method mainly adopts the following approach:

[0089] The server collects copy data. The game server collects the data of all players in each copy into the database. It can also be collected on the game client and then sent to the game server and stored in the database.

[0090] In one implementation, each person's copy (a player plays a copy once) can record up to 300 entries at most. If 6 people participate in this copy, up to 1800 entries can be recorded. Generally, for each person's copy, one piece of data can be recorded per minute. Each piece of data can include: player id (i.e., player identifier), copy id (i.e., the aforementioned game version identifier), timestamp, the jth entry of this player in this copy (i.e., the aforementioned data acquisition sequence identifier), whether cleared flag (cleared is recorded as Y, failed is recorded as N, exited is recorded as Q. Since whether it is cleared is only known later, for example, the result has not come out at the 2nd minute and the copy is only completed at the 25th minute, it can be left blank first and then this value can be filled in later), current health value, current combat power value, current remaining amount of health potions, current remaining amount of mana potions, and one or more of other features that may be useful for model training.

[0091] The server regularly trains multiple models.

[0092] The aforementioned regular time can be, for example, 5 o'clock in the early morning every day for model training. Multiple models are trained using sample sets from different time periods. For example, the model is trained using the sample sets corresponding to n1 = 14, n2 = 3, and n3 = 30. That is, each copy needs to train 3 models. If there are 20 types of copy ids in the game, then 60 models need to be trained in this period.

[0093] The meaning of the aforementioned n1 = 14 is to use the data with timestamps from the previous 14 days to the present as the training data. If it is for the copy with the identifier A (A is a placeholder), the data with the copy identifier A is further selected from the 14 - day data for subsequent training to form the "training data set" of this model, that is, the aforementioned game historical sample set.

[0094] When conducting model training, for each model, according to the selected data, the data input into the model is processed, that is, the aforementioned player game features and whether cleared flag, which will not be elaborated here. It should be noted here that each piece of data input into the model has a corresponding time - effect weight. When the data is input into the model, the data line numbers (one piece of data is one line) can be shuffled and input. For example, the data that was originally in the front may be in the back after shuffling. The shuffled data is input into the model in batches (for example, 1024 pieces as a batch) for training. The loss function loss of the training is determined using the aforementioned loss_avg, thereby obtaining the trained model.

[0095] In addition, to improve the model training effect, the trained model can be reapplied to the "training data set" (without changing the w vector during application), so as to obtain the loss_avg on the entire "training data set". All 60 models are trained.

[0096] Select the model with the smallest "average loss" as the final model for this period. For example, for the 60 models trained for the above 20 replicas, the smallest loss_avg for each replica can be selected from them to obtain 20 models corresponding to the 20 replicas. That is to say, after the training is completed at 5:00 am every day, the "best 20 models" of yesterday on the server are overwritten.

[0097] The client pulls the model to the local. When the player logs in to the game for the first time every day. Download the latest 20 models to the local client.

[0098] The client enters a certain replica. When the player enters a replica, the model corresponding to this replica is started locally (preheated, that is, the model is loaded into the memory and the variables required for their respective calculations are set).

[0099] The client regularly calculates the current clearance probability. The player can collect the x vector every 5 seconds in the replica, that is, the current blood volume value, the current combat power value, the current remaining amount of blood medicine, the current remaining amount of blue medicine, and other features that may be useful for the model. Input the x vector into this preheated model to obtain the floating point number output by the model (that is, the probability that the model thinks will clear the level).

[0100] The probability output by the model is synchronously sent to the clients of the teammates in real time. The clients of each player in the team can know the clearance probability of the teammates. It can be displayed on the game interface, for example, "There is a 52% to 61% probability of clearing the level" is displayed directly above the screen.

[0101] Initiate a surrender when the threshold is reached. If the passing probability of the entire team is lower than the threshold (30%). Then the teammate with the lowest clearance probability initiates the surrender. Surrender controls are displayed on all client interfaces. Within a certain period of time (such as 20 seconds) after the surrender is initiated, if the vote exceeds half of the people, the replica will automatically end. Otherwise, the game continues.

[0102] In summary, the method for real-time display of the game winning rate provided in this embodiment can perform the winning rate prediction in real time through the target game winning rate prediction model, improving the accuracy of the real-time display of the game winning rate; at the same time, it enables players to make accurate decisions in the game and saves the players' time.

[0103] Based on the above method embodiment, the embodiment of the present application also provides a device for real-time display of the game winning rate, which is applied to the first terminal device. See Figure 4As shown, the device for real-time display of the game winning rate includes the following parts:

[0104] A model acquisition module 42, configured to acquire the player game characteristics of the first player character at different historical moments in the current game session; the first player character is the player character controlled by the first terminal device;

[0105] A winning rate prediction module 44, configured to input the player game characteristics at different historical moments into the target game winning rate prediction model corresponding to the current game session that has been pre-trained, and obtain the first current game winning rate; the target game winning rate prediction model is trained by the player game characteristics of the historical player character at different historical moments in the historical game session; wherein, the player game characteristics at different historical moments in the target game winning rate prediction model have different time-effect weights, and the time-effect weight represents the contribution degree of the player game characteristics at different historical moments;

[0106] A winning rate display module 46, configured to display the first current game winning rate on the graphical user interface of the first terminal device.

[0107] The device for real-time display of the game winning rate provided by the embodiment of the present application determines the first current game winning rate corresponding to the current game session through the target game winning rate prediction model trained by the player game characteristics of the historical player character at different historical moments in the historical game session, which can make the prediction result of the obtained model closer to the current game and improve the accuracy of the game; by predicting and displaying the game winning rate in real time, players can make accurate decisions during the game. Especially when the game winning rate is prompted to be low, players can make decisions to exit or surrender as soon as possible, thus saving players' time.

[0108] In a feasible implementation, the above device further includes a model training module, configured to:

[0109] Acquire the game historical sample set at different historical moments of each historical game session; wherein, the game historical sample set includes multiple historical game data; the historical game data at least includes player game characteristics, and the player game characteristics include one or more of the current blood volume value, the current combat power value, and the remaining amount of current blood medicine; determine the time-effect weight corresponding to the historical game data according to the data acquisition sequence identifier of each historical game data in the game historical sample set; train the initialized game winning rate prediction model based on the time-effect weight and the historical game data until the model converges to obtain the target game winning rate prediction model.

[0110] In a feasible implementation, the historical game data is acquired according to the data generation time; the above model training module is further configured to:

[0111] Determine the timeliness weight corresponding to the historical game data according to the data acquisition sequence identifier of each historical game data and the preset weight threshold; the data acquisition sequence identifier is in a proportional relationship with the elapsed time of the game corresponding to the game end moment, and the game end moment includes the clearance moment, the failure moment or the early exit moment.

[0112] In a feasible implementation, the above model training module is further configured to:

[0113] When obtaining the game historical sample set at different historical moments of each historical game match, obtain the game historical sample set at different historical moments of each historical game match within multiple historical preset time periods;

[0114] There are multiple initialized game win rate prediction models, and the number of initialized game win rate prediction models corresponds to the number of historical preset time periods; the above model training module is further configured to: when training the initialized game win rate prediction model based on the timeliness weight and historical game data until the model converges to obtain the target game win rate prediction model:

[0115] For each initialized game win rate prediction model, train the corresponding initialized game win rate prediction model based on the timeliness weight corresponding to the historical game data within the corresponding historical preset time period and the historical game data within the historical preset time period until the corresponding model converges to obtain multiple trained game win rate prediction models; determine the target game win rate prediction model based on the model loss values corresponding to the multiple trained game win rate prediction models.

[0116] In a feasible implementation, the above model training module is further configured to:

[0117] For the initialized game win rate prediction model trained with the historical game data corresponding to each historical preset time period, input the player game characteristics and the corresponding clearance identification into the initialized game win rate prediction model for training to obtain an initial loss value; determine the model loss value corresponding to the game historical sample set based on the timeliness weight corresponding to each game historical data and the initial loss value; determine the corresponding trained game victory prediction model based on the model loss value obtained by training each model.

[0118] In a feasible implementation, the above model training module is further configured to:

[0119] Determine the game win rate prediction model with the smallest model loss value among the multiple trained game win rate prediction models as the target game win rate prediction model.

[0120] In a feasible implementation, the clearance identification at least includes the identification corresponding to early exit; wherein, the identification corresponding to early exit is characterized by a probability, and the probability is in a proportional relationship with the duration of the game that has been played at the corresponding early exit moment.

[0121] In a feasible implementation, the above device further includes: a win rate range determination and display module, configured to:

[0122] Obtain the second current game win rate of the second player character in the current game session; the second player character is the player character operated through the second terminal device, and the number of the second player characters is at least one; determine the game win rate range based on the first current game win rate and the second current game win rate; and display the game win rate range on the graphical user interface of the first terminal device.

[0123] In a feasible implementation, the above device further includes a surrender initiation module, configured to:

[0124] When the minimum value of the game win rate range is less than a preset threshold, initiate a surrender request through the game client corresponding to the minimum value of the game win rate range, so as to determine the game progress based on the number of surrender terminals.

[0125] The device for real-time display of game win rate provided by the embodiments of the present application has the same implementation principle and the same technical effects as the foregoing method embodiments. For the sake of brief description, for the parts not mentioned in the embodiments of the device for real-time display of game win rate, reference may be made to the corresponding content in the foregoing method embodiments of the real-time display of game win rate.

[0126] Figure 5 The structural schematic diagram of an electronic device provided by the embodiments of the present application is shown. The electronic device referred to in this embodiment may include a game server and a game client. For the sake of convenience of description, this electronic device (including the game server or the game client, when performing the corresponding steps, the steps executed by the game server are executed by the corresponding processor of the game server, and the steps executed by the game client are executed by the corresponding processor of the game client) includes: a processor 501, a storage medium 502, and a bus 503. The storage medium 502 stores machine-readable instructions executable by the processor 501. When the electronic device runs a method for real-time display of game win rate as in the embodiment, the processor 501 communicates with the storage medium 502 through the bus 503, and the processor 501 executes the machine-readable instructions, the preamble part of the method item of the processor 501, to perform the following steps:

[0127] Obtain the player game characteristics of the first player character at different historical moments in the current game session; the first player character is the player character controlled through the first terminal device;

[0128] Input the player game characteristics at different historical moments into the target game win rate prediction model corresponding to the current game match that has been pre-trained. The target game win rate prediction model is trained by using the player game characteristics of historical player characters at different historical moments in historical game matches. Among them, the player game characteristics at different historical moments in the target game win rate prediction model have different time effect weights, and the time effect weight represents the contribution degree of the player game characteristics at different historical moments.

[0129] Display the first current game win rate on the graphical user interface of the first terminal device.

[0130] In a feasible implementation, when the processor 501 corresponding to the game server executes the training step of the target game win rate prediction model, it is specifically used for:

[0131] Obtain the game historical sample set at different historical moments of each historical game match. Among them, the game historical sample set includes multiple historical game data. The historical game data at least includes player game characteristics, and the player game characteristics include one or more of the current blood volume value, the current combat power value, and the remaining amount of current blood medicine. Determine the time effect weight corresponding to the historical game data according to the data acquisition sequence identifier of each historical game data in the game historical sample set. Train the initialized game win rate prediction model based on the time effect weight and the historical game data until the model converges to obtain the target game win rate prediction model.

[0132] In a feasible implementation, the historical game data is obtained according to the data generation time. When the processor 501 corresponding to the game server executes determining the time effect weight corresponding to the historical game data according to the data acquisition sequence identifier of each historical game data in the game historical sample set, it is specifically used for:

[0133] Determine the time effect weight corresponding to the historical game data according to the data acquisition sequence identifier of each historical game data and a pre-set weight threshold. The data acquisition sequence identifier is in a proportional relationship with the elapsed time of the game corresponding to the game end moment. The game end moment includes the clearance moment, the failure moment, or the early exit moment.

[0134] In a feasible implementation, when the processor 501 corresponding to the game server executes obtaining the game historical sample set at different historical moments of each historical game match, it is specifically used for:

[0135] Obtain the game historical sample set at different historical moments of each historical game match within multiple historical preset time periods.

[0136] In a feasible implementation, there are multiple initialized game win rate prediction models, and the number of the initialized game win rate prediction models corresponds to the number of the historical preset time periods; when the processor 501 corresponding to the game server performs training on the initialized game win rate prediction models based on the timeliness weight and the historical game data until the models converge to obtain the target game win rate prediction models, it is specifically used for:

[0137] For each initialized game win rate prediction model, based on the timeliness weight corresponding to the historical game data within the corresponding historical preset time period and the historical game data within the historical preset time period, train the corresponding initialized game win rate prediction model until the corresponding model converges to obtain multiple trained game win rate prediction models; based on the model loss values corresponding to the multiple trained game win rate prediction models, determine the target game win rate prediction model.

[0138] In a feasible implementation, when the processor 501 corresponding to the game client performs training on the corresponding initialized game win rate prediction model based on the timeliness weight corresponding to the historical game data within the corresponding historical preset time period and the historical game data within the historical preset time period until the corresponding model converges to obtain multiple trained game win rate prediction models, it is specifically used for:

[0139] For the initialized game win rate prediction model trained with the historical game data corresponding to each historical preset time period, input the player game features and the corresponding whether-cleared flag into the initialized game win rate prediction model for training to obtain an initial loss value; based on the timeliness weight corresponding to each game historical data and the initial loss value, determine the model loss value corresponding to the game historical sample set; based on the model loss values obtained by training each model, determine the corresponding trained game win prediction model.

[0140] In a feasible implementation, when the processor 501 corresponding to the game client performs determining the target game win rate prediction model based on the model loss values corresponding to the multiple trained game win rate prediction models, it is specifically used for:

[0141] Determine the game win rate prediction model with the smallest model loss value among the multiple trained game win rate prediction models as the target game win rate prediction model.

[0142] In a feasible implementation, the whether-cleared flag at least includes a flag corresponding to early exit; wherein, the flag corresponding to early exit is characterized by a probability, and the probability is in a proportional relationship with the duration of the game that has been played at the corresponding early exit moment.

[0143] In a feasible implementation, the processor 501 corresponding to the game client is further configured to:

[0144] Obtain the second current game win rate of the second player character in the current game session; the second player character is the player character operated by the second terminal device, and the number of the second player characters is at least one; determine the game win rate range based on the first current game win rate and the second current game win rate; display the game win rate range on the graphical user interface of the first terminal device.

[0145] In a feasible implementation, the processor 501 corresponding to the game client is further configured to:

[0146] When the maximum value of the game win rate range is less than the preset threshold, initiate a surrender request through the terminal device corresponding to the minimum value of the game win rate range, so as to determine the game process based on the number of surrender terminals.

[0147] In the above manner, the target game win rate prediction model trained by the player game characteristics of the historical player character at different historical moments in the historical game session is used to determine the first current game win rate corresponding to the current game session, which can make the prediction result of the obtained model closer to the current game and improve the accuracy of the game; by predicting and displaying the game win rate in real time, players can make accurate decisions in the game. Especially when it is prompted that the game win rate is low, players can make decisions to exit or surrender as soon as possible, thus saving players' time.

[0148] The embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is run by a processor, the processor executes the following steps:

[0149] Obtain the player game characteristics of the first player character at different historical moments in the current game session; the first player character is the player character controlled by the first terminal device;

[0150] Input the player game characteristics at different historical moments into the target game win rate prediction model corresponding to the current game session that has been pre-trained, and obtain the first current game win rate; the target game win rate prediction model is trained by the player game characteristics of the historical player character at different historical moments in the historical game session; wherein, the player game characteristics at different historical moments in the target game win rate prediction model have different time-effect weights, and the time-effect weights represent the contribution degrees of the player game characteristics at different historical moments.

[0151] Display the first current game win rate on the graphical user interface of the first terminal device.

[0152] In a feasible implementation, when the processor corresponding to the game server executes the training step of the target game win rate prediction model, it is specifically used for:

[0153] Obtain the game historical sample sets at different historical moments of each historical game session; wherein, the game historical sample sets include multiple pieces of historical game data; the historical game data includes at least player game characteristics, and the player game characteristics include one or more of the current health value, the current combat power value, and the current remaining amount of health potions; determine the time effect weight corresponding to the historical game data according to the data acquisition sequence identifier of each historical game data in the game historical sample set; train the initialized game win rate prediction model based on the time effect weight and the historical game data until the model converges to obtain the target game win rate prediction model.

[0154] In a feasible implementation, the historical game data is obtained according to the data generation time; when the processor corresponding to the game server executes determining the time effect weight corresponding to the historical game data according to the data acquisition sequence identifier of each historical game data in the game historical sample set, it is specifically used for:

[0155] Determine the time effect weight corresponding to the historical game data according to the data acquisition sequence identifier of each historical game data and a preset weight threshold; the data acquisition sequence identifier is in a proportional relationship with the elapsed time of the game corresponding to the game end moment; the game end moment includes the clearance moment, the failure moment, or the early exit moment.

[0156] In a feasible implementation, when the processor corresponding to the game server executes obtaining the game historical sample sets at different historical moments of each historical game session, it is specifically used for:

[0157] Obtain the game historical sample sets at different historical moments of each historical game session within multiple historical preset time periods;

[0158] In a feasible implementation, there are multiple initialized game win rate prediction models, and the number of the initialized game win rate prediction models corresponds to the number of the historical preset time periods; when the processor corresponding to the game server executes training the initialized game win rate prediction models based on the time effect weight and the historical game data until the models converge to obtain the target game win rate prediction model, it is specifically used for:

[0159] For each initialized game win rate prediction model, based on the time - effect weight corresponding to the historical game data within the corresponding historical preset time period and the historical game data within the historical preset time period, train the corresponding initialized game win rate prediction model until the corresponding model converges, obtaining multiple trained game win rate prediction models; based on the model loss values corresponding to the multiple trained game win rate prediction models, determine the target game win rate prediction model.

[0160] In a feasible implementation, when the processor corresponding to the game client executes training the corresponding initialized game win rate prediction model based on the time - effect weight corresponding to the historical game data within the corresponding historical preset time period and the historical game data within the historical preset time period until the corresponding model converges, obtaining multiple trained game win rate prediction models, it is specifically used for:

[0161] For the initialized game win rate prediction model trained with the historical game data corresponding to each historical preset time period, input the player game features and the corresponding pass - or - not flag into the initialized game win rate prediction model for training to obtain an initial loss value; based on the time - effect weight corresponding to each piece of game historical data and the initial loss value, determine the model loss value corresponding to the game historical sample set; based on the model loss values obtained from training each model, determine the corresponding trained game win prediction model.

[0162] In a feasible implementation, when the processor corresponding to the game client executes determining the target game win rate prediction model based on the model loss values corresponding to the multiple trained game win rate prediction models, it is specifically used for:

[0163] Determine the game win rate prediction model with the smallest model loss value among the multiple trained game win rate prediction models as the target game win rate prediction model.

[0164] In a feasible implementation, the pass - or - not flag at least includes a flag corresponding to early exit; wherein, the flag corresponding to early exit is characterized by a probability, and the probability is in a proportional relationship with the duration of the game that has been played at the moment of early exit.

[0165] In a feasible implementation, the processor corresponding to the game client is further used for:

[0166] Obtain the second current game win rate of the second player character in the current game session; the second player character is the player character operated through the second terminal device, and the number of the second player characters is at least one; based on the first current game win rate and the second current game win rate, determine the game win rate range; display the game win rate range on the graphical user interface of the first terminal device.

[0167] In a feasible embodiment, the processor corresponding to the game client is further configured to:

[0168] When the maximum value of the game win rate range is less than a preset threshold, initiate a surrender request through the terminal device corresponding to the minimum value of the game win rate range, so as to determine the game process based on the number of surrendering terminals.

[0169] In the above manner, the target game win rate prediction model trained by the player game characteristics of the historical player character at different historical moments in the historical game session is used to determine the first current game win rate corresponding to the current game session, which can make the prediction result of the obtained model closer to the current game and improve the accuracy of the game; by predicting and displaying the game win rate in real time, players can make accurate decisions during the game. Especially when the game win rate is prompted to be low, players can make decisions to quit or surrender as soon as possible, thus saving players' time.

[0170] In the embodiments of the present application, when the computer program is run by the processor, it can also execute other machine-readable instructions to execute the methods described in other embodiments. For the specific method steps and principles of execution, refer to the description of the embodiments and will not be elaborated here.

[0171] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only illustrative. For example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. For another example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some communication interfaces. The indirect couplings or communication connections of the devices or units can be in electrical, mechanical or other forms.

[0172] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0173] In addition, in the embodiments provided in the present application, the various functional units can be integrated in one processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit.

[0174] When the above-mentioned functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of this 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 for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of this application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs.

[0175] It should be noted that: similar reference numerals and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. In addition, the terms "first", "second", "third", etc. are only used for descriptive distinction and cannot be understood as indicating or implying relative importance.

[0176] Finally, it should be noted that: the above-mentioned embodiments are only specific implementation manners of this application, used to illustrate the technical solution of this application, rather than limiting it. The protection scope of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: any person skilled in the art within the technical scope disclosed in this application can still modify the technical solutions recorded in the foregoing embodiments, or can easily think of changes, or perform equivalent replacements on some of the technical features; and these modifications, changes, or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of this application. All should be covered within the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

Claims

1. A method for real-time display of game winning rates, characterized in that, Including: Obtain the player game characteristics of the first player character at different historical moments in the current game session; The first player character is the player character controlled by the first terminal device; Input the player game characteristics at different historical moments into the target game win rate prediction model corresponding to the current game session that has been pre-trained, and obtain the first current game win rate; the target game win rate prediction model is trained by the player game characteristics of historical player characters at different historical moments in historical game sessions; wherein, the player game characteristics at different historical moments in the target game win rate prediction model have different time-effect weights, and the time-effect weight represents the contribution degree of the player game characteristics at different historical moments; the time-effect weight is determined according to the data acquisition sequence identifier in the game historical sample set during model training; the historical game data is acquired according to the data generation time; Display the first current game win rate on the graphical user interface of the first terminal device.

2. The method for real-time display of game winning rate according to claim 1, wherein The training steps of the target game win rate prediction model include: Obtain the game historical sample set at different historical moments of each historical game session; wherein, the game historical sample set includes multiple historical game data; the historical game data at least includes player game characteristics, and the player game characteristics include one or more of the current blood volume value, the current combat power value, and the remaining amount of current blood medicine; Determine the time-effect weight corresponding to the historical game data according to the data acquisition sequence identifier of each historical game data in the game historical sample set; Train the initialized game win rate prediction model based on the time-effect weight and the historical game data until the model converges to obtain the target game win rate prediction model.

3. The method for real-time display of game winning rate according to claim 2, wherein Determine the time-effect weight corresponding to the historical game data according to the data acquisition sequence identifier of each historical game data in the game historical sample set, including: Determine the time-effect weight corresponding to the historical game data according to the data acquisition sequence identifier of each historical game data and the preset weight threshold; the data acquisition sequence identifier is in a proportional relationship with the elapsed time of the game corresponding to the game end moment; the game end moment includes the clearance moment, the failure moment, or the early exit moment.

4. The method for real-time display of game winning rate according to claim 2, characterized in that, Obtain the game historical sample set at different historical moments of each historical game session, including: Obtain the game historical sample set at different historical moments of each historical game session within multiple historical preset time periods; There are multiple initialized game win rate prediction models, and the number of the initialized game win rate prediction models corresponds to the number of the historical preset time periods; training the initialized game win rate prediction model based on the time-effect weight and the historical game data until the model converges to obtain the target game win rate prediction model, including: For each initialized game win rate prediction model, train the corresponding initialized game win rate prediction model based on the time-effect weight corresponding to the historical game data within the corresponding historical preset time period and the historical game data within the historical preset time period until the corresponding model converges to obtain multiple trained game win rate prediction models; The target game win rate prediction model is determined based on the model loss values corresponding to the multiple trained game win rate prediction models.

5. The method for real-time display of game winning rate according to claim 4, characterized in that, Based on the timeliness weight corresponding to the historical game data within the corresponding preset historical time period and the historical game data within the preset historical time period, the corresponding initialized game win rate prediction model is trained until the corresponding model converges, thereby obtaining multiple trained game win rate prediction models, including: The game win rate prediction model initialized by training the historical game data corresponding to each preset historical time period is inputted with the player's game features and the corresponding clearance flag into the initialized game win rate prediction model for training to obtain an initial loss value; Determine the model loss value corresponding to the game history sample set based on the timeliness weight corresponding to each game history data and the initial loss value; The corresponding trained game victory prediction model is determined based on the model loss value obtained from each model training.

6. The method for real-time display of game winning rate according to claim 5, wherein Determining the target game win rate prediction model based on the model loss values corresponding to the multiple trained game win rate prediction models includes: The game winning rate prediction model with the smallest model loss value among the multiple trained game winning rate prediction models is determined as the target game winning rate prediction model.

7. The method for real-time display of game winning rate according to claim 5, characterized in that, The clearance mark includes at least a mark corresponding to early exit; wherein, the mark corresponding to early exit is represented by probability, and the probability is proportional to the game duration corresponding to the early exit moment.

8. The method for real-time display of game winning rate according to claim 6, wherein The method further comprises: Obtaining a second current game win rate of a second player character in a current game match; the second player character is a player character operated by a second terminal device, and there is at least one second player character; determining a game win rate range based on the first current game win rate and the second current game win rate; The game win rate range is displayed on a graphical user interface of the first terminal device.

9. The method for real-time display of game winning rate according to claim 8, wherein The method further comprises: When the maximum value of the game winning rate range is less than a preset threshold, a surrender request is initiated through the terminal device corresponding to the minimum value of the game winning rate range, so as to determine the game progress based on the number of surrendered terminals.

10. A device for real-time display of game winning rates, characterized in that, include: A feature acquisition module is used to obtain the player game features of the first player character at different historical moments in the current game; The first player character is a player character controlled by a first terminal device; a win rate prediction module, configured to input the player's game features at different historical moments into a pre-trained target game win rate prediction model corresponding to the current game match, to obtain a first current game win rate; the target game win rate prediction model is trained using the player's game features at different historical moments in historical game matches; wherein the player's game features at different historical moments in the target game win rate prediction model have different timeliness weights, the timeliness weights representing the contribution of the player's game features at different historical moments; the timeliness weights are determined based on a data acquisition sequence identifier for each historical game data in a game history sample set during model training; the historical game data is acquired according to the time at which the data was generated; A winning rate display module, configured to display the first current game winning rate on the graphical user interface of the first terminal device.

11. An electronic device, characterized in that, It includes a processor and a memory. The memory stores computer-executable instructions that can be executed by the processor. The processor executes the computer-executable instructions to implement the method for real-time display of game winning rate according to any one of claims 1 to 9.

12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions. When the computer-executable instructions are called and executed by a processor, the computer-executable instructions cause the processor to implement the method for real-time display of game winning rate according to any one of claims 1 to 9.

Images