Confrontation competition result prediction method and related equipment

Abstract

The invention provides a confrontation competition result prediction method and related equipment. The method comprises the following steps: determining a plurality of target players participating in a target game; obtaining feature data of a plurality of target players; the feature data of each target player comprises implicit vectors which are obtained by training on the basis of competition data of historical competitions participated by the target player and are used for representing technical and tactical features of the target player; the competition data comprises competition results of historical competitions and competition indexes of all competitors, and the competition indexes are used for describing technical and tactical application conditions of the competitors in the competition process; and inputting the implicit vectors of the plurality of target players into the trained prediction model, and predicting a competition result of the target competition. According to the method, the implicit vectors obtained based on historical game data training can more accurately capture the technical and tactical features of each player and different expressions of each player facing different opponents, and the implicit vectors of the players are utilized to predict the game result, so that the prediction accuracy can be improved.

Description

Technical Field

[0001] One or more embodiments of the present specification relate to the field of machine learning technology, and more particularly, to a method for predicting the results of competitive games and related equipment. Background Art

[0002] In sports competitions, predicting the results of a game (such as the outcome of a win or loss) before the game starts can not only provide data support for game arrangements, increase the viewing appeal of the game and improve commercial opportunities, but also provide a reference for coaches and players to help develop more effective training plans and game strategies, thereby enhancing the competitiveness of players.

[0003] However, existing methods for predicting match results still have many shortcomings. They can only score players' strengths based on their historical match results (such as wins, draws, and losses), and then predict the outcome of the current match based on this single score. They are unable to capture the strengths and weaknesses of each player, let alone predict the different performances of players when facing different opponents, resulting in inaccurate and unreliable prediction results, and thus unable to provide effective pre-match tactical guidance for coaches and players. Summary of the invention

[0004] In view of this, one or more embodiments of the present specification provide a method and related devices for predicting the results of a competitive game to address the deficiencies in the related art.

[0005] In a first aspect, this specification provides a method for predicting the results of a competitive game, the method comprising: Identify multiple target players participating in a target match; Acquire feature data of the multiple target players; wherein the feature data of each target player includes a latent vector obtained by training based on game data of historical games in which the target player has participated, and used to represent the technical and tactical features of the target player; the game data includes game results of historical games, and game indicators of all contestants, and the game indicators are used to describe the application of the contestants' skills and tactics during the game; The latent vectors of the multiple target players are input into the trained prediction model to predict the game result of the target game.

[0006] In one illustrated embodiment, the prediction model includes a game indicator prediction module and a game result prediction module; Inputting the latent vectors of the plurality of target players into the trained prediction model to predict the result of the target game includes: Inputting the latent vectors of the plurality of target players into a game index prediction module to predict the game indexes of the plurality of target players in the target game; and The predicted game indicators of the multiple target players are further input into a game result prediction module to predict the game result of the target game.

[0007] In an illustrated embodiment, the competition indicator prediction module includes a latent vector interaction module and an intermediate feature decoding module; The step of inputting the latent vectors of the plurality of target players into a competition index prediction module to predict competition indexes of the plurality of target players in the target competition includes: Inputting the latent vectors of the multiple target players into a latent vector interaction module for latent vector interaction processing, and outputting intermediate features obtained after the interaction; and The intermediate features are further input into an intermediate feature decoding module for decoding processing, and the game indicators of the multiple target players in the target game obtained after decoding are output.

[0008] In one illustrated embodiment, the method further comprises: Acquire a training sample set, wherein the training sample set includes latent vectors of a plurality of sample players and game data of a plurality of historical games played between the plurality of sample players; The game data of each historical game is used as the sample label corresponding to the latent vector of the sample player in the historical game; Based on the latent vectors of the multiple sample players and their corresponding sample labels, supervised training is performed on the prediction model.

[0009] In an illustrated embodiment, supervised training is performed on the prediction model based on latent vectors of the plurality of sample players and their corresponding sample labels, including: Based on the latent vectors of the plurality of sample players and their corresponding game indicators as sample labels, supervised training is performed on the game indicator prediction module; and Based on the latent vectors of the multiple sample players and their corresponding game results as sample labels, supervised training is performed on the game result prediction module.

[0010] In an illustrated embodiment, obtaining a training sample set includes: Determine a plurality of first sample players to be studied, and further determine a plurality of second sample players who have played against at least N first sample players among the plurality of first players based on the game records of the plurality of first sample players; N is an integer greater than or equal to 1; Determining a plurality of historical matches between the plurality of first sample players and the plurality of second sample players; The game results and game videos of each historical game are obtained, and the game indicators of the first sample players and the second sample players in each historical game are marked according to the game videos.

[0011] In an illustrated embodiment, the latent vectors of the plurality of sample players are randomly initialized latent vectors; and the method further includes: Based on the latent vectors of the multiple sample players and their corresponding sample labels, supervised training is performed on the latent vectors of the multiple sample players to obtain latent vectors for representing the technical and tactical features of the multiple sample players.

[0012] In an illustrated embodiment, the target game includes a boxing match, and the match result of the target game includes a win, draw, or loss result of the boxers.

[0013] In a second aspect, this specification provides a latent vector training method, the method comprising: Acquire a training sample set, wherein the training sample set includes latent vectors of multiple sample players and game data of multiple historical games played between the multiple sample players; wherein the game data of each historical game includes a game result of the historical game and game indicators of the sample players participating in the historical game, wherein the game indicators are used to describe the application of skills and tactics of the sample players during the game; The game data of each historical game is used as a sample label corresponding to the latent vector of the sample player in the historical game, and the latent vector of the sample player in the historical game is input into the prediction model. Based on the difference between the output result of the prediction model and the sample label, the latent vector of the sample player is supervisedly trained to obtain the latent vector used to represent the technical and tactical characteristics of the sample player.

[0014] In a third aspect, this specification provides a device for predicting the results of a competitive match, the device comprising: A determination unit, used for determining a plurality of target players participating in a target game; an acquisition unit, configured to acquire feature data of the plurality of target players; wherein the feature data of each target player comprises a latent vector obtained by training based on game data of historical games in which the target player has participated, and used to represent the technical and tactical features of the target player; the game data comprises game results of historical games, and game indicators of all participating players, and the game indicators are used to describe the application of the skills and tactics of the participating players during the game; The prediction unit is used to input the latent vectors of the multiple target players into the trained prediction model to predict the game result of the target game.

[0015] In a fourth aspect, this specification provides a latent vector training device, the method comprising: an acquisition unit, configured to acquire a training sample set, wherein the training sample set includes latent vectors of a plurality of sample players and game data of a plurality of historical games played between the plurality of sample players; wherein the game data of each historical game includes a game result of the historical game and game indicators of the sample players participating in the historical game, wherein the game indicators are used to describe the application of techniques and tactics of the sample players during the game; A training unit is used to use the game data of each historical game as a sample label corresponding to the latent vector of the sample player in the historical game, and input the latent vector of the sample player in the historical game into the prediction model, and based on the difference between the output result of the prediction model and the sample label, perform supervised training on the latent vector of the sample player to obtain a latent vector for representing the technical and tactical characteristics of the sample player.

[0016] Accordingly, the present specification also provides a computing device, comprising: a memory and a processor; the memory stores a computer program / instruction that can be executed by the processor; when the processor executes the computer program / instruction, the method for predicting the results of the competitive game described in the first aspect or the latent vector training method described in the second aspect is executed.

[0017] Accordingly, the present specification also provides a computer-readable storage medium on which a computer program / instruction is stored. When the computer program / instruction is executed by a processor, the method for predicting the results of competitive games as described in the first aspect or the latent vector training method as described in the second aspect is executed.

[0018] Accordingly, the present specification also provides a computer program product, which includes a computer program / instruction. When the computer program / instruction is executed by a processor, it executes the method for predicting the results of competitive games as described in the first aspect or the latent vector training method as described in the second aspect.

[0019] In summary, the present application can be based on the game data of historical games in which the players have participated, and train to obtain latent vectors used to represent the technical and tactical characteristics of the players. These game data include not only the game results of historical games, but also the game performance of all contestants (i.e. the players themselves and their opponents) in historical games (such as the application of various technical and tactical techniques). In this way, the latent vectors obtained based on the training of historical game data can more accurately capture the technical and tactical characteristics of each player, as well as the different performances of each player when facing different opponents. Furthermore, for the current target game to be predicted, the present application can input the latent vectors of multiple target players participating in the target game into the prediction model to predict the game results, which is equivalent to simulating a game based on the technical and tactical characteristics of multiple target players, thereby improving the accuracy of the prediction of the final game results, and then providing players and coaches with more reliable pre-match data, so as to formulate more effective pre-match training plans and game strategies, etc.

[0020] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a flowchart of a method for predicting the results of a confrontational game provided by an exemplary embodiment; Figure 2 is a schematic diagram of a prediction model provided by an exemplary embodiment; Figure 3 It is a structural schematic diagram of a competition index prediction module provided by an exemplary embodiment; Figure 4 is a schematic diagram of a training process of a latent vector and a prediction model provided by an exemplary embodiment; Figure 5 is a flowchart of a latent vector training method provided by an exemplary embodiment; Figure 6 It is a schematic diagram of the structure of a device for predicting the results of a confrontational game provided by an exemplary embodiment; Figure 7 is a structural schematic diagram of a latent vector training device provided by an exemplary embodiment; Figure 8 It is a structural diagram of a computing device provided by an exemplary embodiment. DETAILED DESCRIPTION

[0022] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with one or more embodiments of this specification. Instead, they are merely examples of devices and methods consistent with some aspects of one or more embodiments of this specification as detailed in the appended claims.

[0023] It should be noted that: in other embodiments, the steps of the corresponding method are not necessarily performed in the order shown and described in this specification. In some other embodiments, the steps included in the method may be more or less than those described in this specification. In addition, a single step described in this specification may be decomposed into multiple steps for description in other embodiments; and multiple steps described in this specification may be combined into a single step for description in other embodiments.

[0024] It should be noted that the “plurality” mentioned in this application refers to two or more than two.

[0025] In addition, the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0026] In sports competitions, predicting the results of a game (such as the outcome of a win or loss) before the game starts can not only provide data support for game arrangements, increase the viewing appeal of the game and improve commercial opportunities, but also provide a reference for coaches and players to help develop more effective training plans and game strategies, thereby enhancing the competitiveness of players.

[0027] However, the existing match result prediction methods still have many shortcomings. They can only score the players' strength based on their historical match results (such as wins, draws, and losses), and then predict the outcome of the current match based on this single score. Specifically, taking boxing matches as an example, the existing match result prediction methods will score the boxers' strength based on the boxers' historical match results by minimizing the distance between the predicted win and loss and the actual win and loss, and the degree of change in the strength of a single boxer, so as to make a pre-match prediction of the outcome by comparing the scores of the two boxers. This single-scoring prediction method cannot capture the style, strengths and weaknesses of each player, and is even more unable to predict the different performances of players when facing different opponents, resulting in inaccurate and unreliable prediction results, and thus unable to provide effective pre-match tactical guidance for coaches and players.

[0028] Based on this, this specification provides a method for predicting the results of confrontational games to solve the deficiencies in the related art. The method can be applied to a computing device, which can be, for example, a smart wearable device, a smart phone, a tablet computer, a laptop computer, a desktop computer, a server, a server cluster composed of multiple servers, or a cloud computing service center, etc., and this specification does not specifically limit this.

[0029] In implementation, the present application may first determine multiple target players participating in the target game, and then obtain the feature data of the multiple target players; wherein the feature data of each target player may include a latent vector trained based on the game data of historical games in which the target player has participated, which is used to represent the technical and tactical features of the target player; wherein the game data of historical games may include the game results of historical games, and the game indicators of all participating players, which may be used to describe the application of the skills and tactics of the participating players during the game. Furthermore, the latent vectors of multiple target players may be input into the pre-trained prediction model to predict the game results of the target game.

[0030] In the above technical scheme, the present application can be based on the game data of historical games in which the players have participated, and train to obtain latent vectors used to represent the technical and tactical characteristics of the players. These game data include not only the game results of historical games, but also the game performance of all contestants (i.e., the players themselves and their opponents) in historical games (such as the application of various technical and tactical techniques). In this way, the latent vectors obtained based on the training of historical game data can more accurately capture the technical and tactical characteristics of each player, as well as the different performances of each player when facing different opponents. Furthermore, for the current target game to be predicted, the present application can input the latent vectors of multiple target players participating in the target game into the prediction model to predict the game results, which is equivalent to simulating a game based on the technical and tactical characteristics of multiple target players, thereby improving the accuracy of the prediction of the final game results, and then providing players and coaches with more reliable pre-match data, so as to formulate more effective pre-match training plans and game strategies, etc.

[0031] In one illustrated embodiment, see Figure 1 , Figure 1 FIG. 1 is a flow chart of a method for predicting the results of a competitive match provided by an exemplary embodiment. Figure 1 As shown, the method may specifically include the following steps S101 to S103.

[0032] Step S101, determining a plurality of target players participating in a target game.

[0033] First, for the current target game to be predicted, multiple target players participating in the target game can be determined. In an embodiment shown, multiple target players participating in the target game can be determined based on the schedule or other information, and this specification does not specifically limit this. Exemplarily, the target game can be any game in the schedule of this season, such as the quarterfinals, semifinals or finals, etc., and this specification does not specifically limit this.

[0034] It should be noted that the present specification does not specifically limit the type of the target game. In one embodiment, the target game can be any type of sports competition, such as boxing, table tennis, tennis, badminton and other confrontational games, and the present specification does not specifically limit this. Accordingly, the multiple target players participating in the target game can be boxers, table tennis players, tennis players, badminton players, etc. of the two opposing teams.

[0035] Step S102, obtaining feature data of the multiple target players; wherein the feature data of each target player includes a latent vector obtained by training based on game data of historical games in which the target player has participated, and used to represent the technical and tactical features of the target player; the game data includes the game results of the historical games, and the game indicators of all the players, and the game indicators are used to describe the application of the players' skills and tactics during the game.

[0036] Furthermore, after a plurality of target players participating in the target game are determined, feature data of the plurality of target players may be obtained.

[0037] In an illustrated embodiment, the feature data of each target player may include a latent vector for representing the technical and tactical features of the target player. The so-called technical and tactical features may include various features for describing the technical and tactical features that the player is good at or used to use, as well as some strengths or weaknesses of the player in certain technical and tactical features, and this specification does not specifically limit this.

[0038] In an illustrated embodiment, the latent vector of each target player can be trained based on the game data of one or more historical games in which the target player has participated. The training process of the latent vector is as follows: Figure 4 The description in the corresponding embodiment will not be repeated here.

[0039] In an illustrated embodiment, the game data of each historical game may include the game result of the historical game, and the game indicators (or technical and tactical indicators) of all players in the historical game.

[0040] The game results of historical games may be win / loss results or win / draw / loss results of the contestants, or specific scores of the contestants, etc. This specification does not make any specific limitation on this.

[0041] The competition index of the contestant can be used to describe the application of the contestant's skills and tactics during the game. The application of skills and tactics can include which skills and tactics the contestant used during the game, and the specific application of each skill and tactic. In an embodiment shown, the competition index of each contestant can be analyzed and marked based on the game video of the historical game, and this specification does not specifically limit this.

[0042] For example, taking a boxing match as an example, the competition indicators of a boxer may include: the punches used by the boxer in the match (such as straight punches, left hooks and right hooks), combination punches, number of effective punches, hitting parts and punch speed, etc. This specification does not make specific limitations on this.

[0043] For example, taking table tennis as an example, the competition indicators of table tennis players may include: serving skills (such as spin balls), receiving skills, long and short ball changes, backhand and forehand changes, etc. used by table tennis players in the game, and this specification does not make specific limitations on this.

[0044] In one illustrated embodiment, latent vectors of multiple target players participating in a target game may be pre-trained. For example, latent vectors of all active players may be trained in advance based on historical game data. For example, after the list of players for this season is announced, latent vectors of all participating players on the list may be trained, and so on. This specification does not make any specific limitation on this.

[0045] Alternatively, in an illustrated embodiment, after determining multiple target players participating in the current target game, latent vectors of the multiple target players may be trained based on game data of historical games in which the target players have previously participated. This specification does not specifically limit this.

[0046] Step S103: input the latent vectors of the multiple target players into the trained prediction model to predict the result of the target game.

[0047] Furthermore, after obtaining the latent vectors of multiple target players participating in the target game, the result of the target game can be predicted based on the latent vectors of the multiple target players. For example, the result of the game can be the win or loss result of the target player (e.g., a boxer), or a win, draw or loss result, or a specific score result of the player, etc., which is not specifically limited in this specification.

[0048] In one embodiment, the latent vectors of multiple target players can be input into a pre-trained prediction model, and the prediction model predicts the result of the target game based on the latent vectors of the multiple target players. Specifically, the game indicators of the multiple target players in the target game can be predicted based on the latent vectors of the multiple target players, and then the game result of the target game can be further predicted based on the predicted game indicators. The training process of the prediction model can be referred to as follows Figure 4 The description in the corresponding embodiment will not be repeated here.

[0049] In this way, this specification uses latent vectors of multiple target players in the game (such as boxers from both sides) to predict the results of the game, fully considering the different strengths and performances of the multiple players in the game, effectively improving the accuracy of the prediction of the game results, and solving the defects of the single historical indicator data in the prior art.

[0050] It should be noted that this specification does not specifically limit the specific structure of the prediction model. Figure 2 , Figure 2 FIG. 1 is a schematic diagram of a prediction model provided by an exemplary embodiment. Figure 2 As shown, the prediction model may include a game indicator prediction module and a game result prediction module.

[0051] Accordingly, based on Figure 2 The structure shown in the figure, when predicting the game results through the prediction model, the present application may specifically include: First, if Figure 2 As shown, multiple target players (for example, including Figure 2 The latent vectors of the target players 1 and 2 shown in FIG. 1 are input into a competition index prediction module, and the competition index prediction module predicts the competition indexes of the target players in the target competition based on the latent vectors of the target players, for example, including Figure 2 The game indicators of target player 1 and target player 2 are shown. The game indicators of target player 1 and target player 2 may include the predicted possible technical and tactical applications of target player 1 and target player 2 during the target game, which is equivalent to simulating a game process and predicting possible situations and respective game performances of the two parties when they compete.

[0052] Then, if Figure 2 As shown, the predicted competition indicators of the multiple target players are further input into a competition result prediction module, and the competition result prediction module further predicts the competition result of the target competition based on the predicted competition indicators of the multiple target players.

[0053] It should be noted that the present specification does not specifically limit the specific structure of the competition index prediction module. Figure 3 , Figure 3 FIG. 1 is a schematic diagram of a structure of a competition index prediction module provided by an exemplary embodiment. Figure 3 As shown, the competition indicator prediction module includes a latent vector interaction module and an intermediate feature decoding module.

[0054] Accordingly, based on Figure 3 The structure shown in the figure, when predicting the competition index through the competition index prediction module, the present application may specifically include: First, if Figure 3 As shown in FIG. 1 , the latent vectors of multiple target players are input into the latent vector interaction module, which performs latent vector interaction processing (or feature interaction processing) on ​​the latent vectors of multiple target players and outputs the intermediate features obtained after the interaction. It is not difficult to understand that the latent vector interaction module is equivalent to simulating the competition process between multiple target players based on the latent vectors of multiple target players.

[0055] Then, if Figure 3 As shown, the intermediate features output by the latent vector interaction module are further input into the intermediate feature decoding module for decoding processing, and the competition indicators of multiple target players in the target game obtained after decoding are output.

[0056] In an illustrated embodiment, the intermediate features output by the latent vector interaction module may include multiple intermediate features corresponding to multiple target players. Accordingly, the intermediate feature decoding module may decode the multiple intermediate features respectively, thereby predicting the competition indicators of the multiple target players.

[0057] In this way, this manual uses the latent vector interaction module to simulate the fighting process between players (such as the fighting process between boxers) on the one hand, and improves the accuracy of prediction through the interaction of latent vectors. On the other hand, it uses the intermediate feature decoding module to decode the intermediate features output after the latent vector interaction, so as to obtain the predicted game indicators and enhance the learning ability of the model.

[0058] In addition, it should be noted that this manual Figure 2 and Figure 3 The specific implementation methods of the latent vector interaction module, the intermediate feature decoding module and the game result prediction module shown are not particularly limited.

[0059] In one embodiment shown, the latent vector interaction module can be a multilayer perceptron (MLP), such as a multilayer perceptron with 2 layers, which is not specifically limited in this specification. Among them, MLP is a feedforward artificial neural network, which consists of multiple levels of nodes (or "neurons"), each node is connected to all nodes in the next layer. MLP usually includes an input layer, one or more hidden layers, and an output layer. This type of neural network can learn complex nonlinear functions and is used for classification and regression tasks, etc., which will not be described in detail here.

[0060] In an illustrated embodiment, the intermediate feature decoding module may be a linear projection layer (LinearProjection Layer), which is not specifically limited in this specification. The linear projection layer is mainly used to map input data from one representation space to another representation space, which is usually a lower dimensional space or a different feature space, and will not be described in detail here.

[0061] In one illustrated embodiment, the game result prediction module may also be a multi-layer perceptron, such as a multi-layer perceptron with 3 layers, etc. This specification does not specifically limit this.

[0062] Next, the training process of the latent vector and the prediction model will be described, which may specifically include the following steps S401-S402.

[0063] Step S401, obtaining a training sample set.

[0064] First, it is necessary to obtain a training sample set. The training sample set may include latent vectors of multiple sample players and game data of multiple historical games between multiple sample players. The set consisting of multiple sample players can be denoted as F, each sample player can be denoted as b∈F, the set consisting of multiple historical games can be denoted as M, and each historical game can be denoted as m∈M.

[0065] Among them, this application can construct a learnable latent vector for each sample player, denoted as v b ∈ , where d is the dimension of the latent vector. It should be understood that before the training begins, the latent vectors of the multiple sample players may be randomly initialized latent vectors, and this specification does not specifically limit this.

[0066] The match data of each historical match m may include the match result of each historical match (which can be recorded as y m ), and the game indicators of the sample players participating in the historical game (which can be recorded as I m In this way, each historical game can be used as a training sample, recorded as ( , ,y m , , ).in, is the latent vector of player i in historical game m, is the latent vector of player j in historical game m, y m is the result of the historical match m, is the competition index of player i in historical game m, is the game index of player j in historical game m. In an illustrated embodiment, the game index of player i and player j can be analyzed and marked based on the game video of historical game m. and , this manual does not make any specific limitation on this.

[0067] In an illustrated embodiment, the multiple sample players in the training sample set may be athletes to be studied. For example, boxers may be all active boxers, all boxers in the list of participants this season, or boxers who finally entered the top eight in the competition, etc. This specification does not specifically limit this. In an illustrated embodiment, the multiple sample players may include multiple target players who participated in the target competition.

[0068] It should be noted that this specification does not specifically limit the specific implementation method of how to determine the above-mentioned multiple sample players and the corresponding multiple historical games.

[0069] In an illustrated embodiment, the present specification may first determine a plurality of first sample players to be studied, and further determine a plurality of second sample players who have played against at least N first sample players among the plurality of first players based on the game records of the plurality of first sample players. Wherein, N may be an integer greater than or equal to 1, for example, N may be equal to 1, 2, or 4, etc., and the present specification does not specifically limit this.

[0070] Furthermore, based on the history of matches between the first sample players and the second sample players, multiple historical matches between the first sample players and the second sample players can be determined. Furthermore, the game results and game videos of each of the multiple historical matches can be obtained, and the game indicators of the first sample players and the second sample players in each historical match can be annotated based on the game videos.

[0071] For example, taking boxing matches as an example, we can first determine the set of boxers to be studied B (i.e., the multiple first sample players mentioned above), and then use the match records of each boxer in the set of boxers to be studied B to collect other boxers who have played against at least two of these boxers as expanded boxers, and obtain the expanded boxer set O (i.e., the multiple second sample players mentioned above). Then, the set of boxers to be studied B and the expanded boxer set O are merged to obtain the set of all boxers of interest F=B∪O (i.e., the multiple sample players mentioned above). Then, using the history of all boxers of interest, select the matches between all boxers of interest as the matches of interest, and obtain the set of matches of interest (i.e., the historical match set M mentioned above).

[0072] In this way, this manual selects multiple historical games as focus games based on the game records of the players to be studied, constructs learnable latent vectors for the players, and constructs training samples by combining the game data of multiple historical games, ensuring the diversity and representativeness of the training samples.

[0073] Step S402: Use each historical game as a training sample to perform supervised training on the prediction model and latent vector.

[0074] Furthermore, each historical game in the above historical game set can be used as a training sample to perform supervised training on the prediction model and latent vector.

[0075] In one illustrated embodiment, the game data of each historical game (including game results and game indicators) can be used as a sample label (i.e., true value) corresponding to a latent vector of a sample player in the historical game. Then, based on the latent vector of the sample player and its corresponding sample label, supervised training is performed on the prediction model and the latent vector.

[0076] Specifically, based on the latent vectors of multiple sample players and their corresponding game indicators as sample labels, a game indicator prediction module (including a latent vector interaction module and an intermediate feature decoding module) can be supervised trained; and based on the latent vectors of multiple sample players and their corresponding game results as sample labels, a game result prediction module can be supervised trained. This specification does not make any specific limitations on this.

[0077] In one illustrated embodiment, see Figure 4 , Figure 4 It is a schematic diagram of a training process of a latent vector and a prediction model provided by an exemplary embodiment.

[0078] First, if Figure 4 As shown, each historical game m is taken as a training sample and recorded as ( , ,y m , , ), then the latent vectors of the sample players in the historical game (e.g. Figure 4 The latent vector of player i and the latent vector of player j shown in FIG. 1 are input into the latent vector interaction module for feature interaction, thereby simulating the game process between player i and player j, and outputting the intermediate features obtained after the interaction. Exemplarily, the calculation logic involved in the latent vector interaction module can be shown in the following formula (1).

[0079] h i ,h j =f interact ( , ) (1) Among them, h i is the intermediate feature corresponding to player i, h j is the intermediate feature corresponding to player j.

[0080] Furthermore, if Figure 4As shown, the intermediate features output by the latent vector interaction module are further input into the intermediate feature decoding module, which decodes the intermediate features corresponding to the players i and j, respectively, so as to predict the competition indicators corresponding to the players i and j, respectively. Exemplarily, the calculation logic involved in the intermediate feature decoding module can be shown in the following formula (2).

[0081] =f decode (h i ), =f decode (h j ) (2) in, is the predicted competition index of player i, is the predicted competition index of player j.

[0082] like Figure 4 As shown, we can use the training samples ( , ,y m , , ) The actual competition indicators of players i and j marked in and Supervise the output of the current intermediate feature decoding module. Specifically, the predicted competition indicators can be used to and With real game indicators and The difference between them is used to optimize the latent vectors of players i and j as well as the parameters of the latent vector interaction module and the intermediate feature decoding module. For details, please refer to the following loss function.

[0083] Furthermore, if Figure 4 As shown, the predicted competition indicators of players i and j are further input into the competition result prediction module, and the competition result prediction module further predicts the competition result based on the predicted competition indicators of players i and j. Exemplarily, the calculation logic involved in the competition result prediction module can be shown in the following formula (3).

[0084] =f p ( || ) (3) in, The predicted results of the game.

[0085] like Figure 4 As shown, we can use the training samples ( , ,ym , , ) in the real game result y m (For example, player i wins) monitors the output of the current game result prediction module. Specifically, the predicted game result can be With the real game results m The difference between them is used to optimize the parameters of the game result prediction module. For details, please refer to the following loss function.

[0086] In an illustrated embodiment, the loss function defined in the present application may be expressed as the following formula (4).

[0087]

[0088] in, is the binary cross entropy loss function, is the mean square error loss function, is the regularization parameter.

[0089] In one illustrated embodiment, when training the player's latent vector and prediction model, the present application can select a batch of training samples (i.e., a batch of historical games) each time, perform gradient backpropagation on the loss function after the forward process, and use the Adam optimizer to simultaneously optimize the player's latent vector and model parameters, thereby training to obtain a latent vector for representing the player's technical and tactical characteristics, and a prediction model for predicting the game results based on the player's latent vector.

[0090] In summary, the latent vector-based prediction model provided in this description can more accurately capture the interaction characteristics and style differences between players, and improve the accuracy of match win / loss prediction. At the same time, by using historical game data to supervise the learning of latent vectors, the model not only reduces the reliance on a single historical indicator data, but also can reflect the performance of players when facing different opponents. This improvement makes the prediction results more comprehensive and reliable, thus providing more valuable tactical guidance for coaches and athletes.

[0091] In addition, in one illustrated embodiment, see Figure 5 , Figure 5 FIG. 1 is a flowchart of a latent vector training method provided by an exemplary embodiment. Figure 5 As shown, the method may specifically include the following steps S501-S502.

[0092] Step S501, obtaining a training sample set, wherein the training sample set includes latent vectors of multiple sample players and game data of multiple historical games played between the multiple sample players; wherein the game data of each historical game includes the game result of the historical game and game indicators of the sample players participating in the historical game, and the game indicators are used to describe the technical and tactical application of the sample players during the game.

[0093] Step S502: Use the game data of each historical game as a sample label corresponding to the latent vector of the sample player in the historical game, and input the latent vector of the sample player in the historical game into the prediction model. Based on the difference between the output result of the prediction model and the sample label, supervised training is performed on the latent vector of the sample player to obtain a latent vector for representing the technical and tactical characteristics of the sample player.

[0094] In one illustrated embodiment, Figure 5 The hidden vector training method shown in the figure can be specifically referred to above Figure 4 The description in the corresponding embodiment will not be repeated here.

[0095] Corresponding to the implementation of the above method flow, the embodiment of this specification also provides a device for predicting the results of a confrontational game, which can be applied to a computing device. Figure 6 , Figure 6 FIG. 1 is a schematic diagram of a device for predicting the results of a competitive match provided by an exemplary embodiment. Figure 6 As shown, the device 60 includes: A determination unit 601 is used to determine a plurality of target players participating in a target game; The acquisition unit 602 is used to acquire the feature data of the plurality of target players; wherein the feature data of each target player includes a latent vector obtained by training based on the game data of historical games in which the target player has participated, and used to represent the technical and tactical features of the target player; the game data includes the game results of the historical games, and the game indicators of all the players, and the game indicators are used to describe the application of the players' skills and tactics during the game; The prediction unit 603 is used to input the latent vectors of the multiple target players into the trained prediction model to predict the game result of the target game.

[0096] In one illustrated embodiment, the prediction model includes a game indicator prediction module and a game result prediction module; The prediction unit 603 is specifically used for: Inputting the latent vectors of the plurality of target players into a game index prediction module to predict the game indexes of the plurality of target players in the target game; and The predicted game indicators of the multiple target players are further input into a game result prediction module to predict the game result of the target game.

[0097] In an illustrated embodiment, the competition indicator prediction module includes a latent vector interaction module and an intermediate feature decoding module; The prediction unit 603 is specifically used for: Inputting the latent vectors of the multiple target players into a latent vector interaction module for latent vector interaction processing, and outputting intermediate features obtained after the interaction; and The intermediate features are further input into an intermediate feature decoding module for decoding processing, and the game indicators of the multiple target players in the target game obtained after decoding are output.

[0098] In an illustrated embodiment, the apparatus 60 further includes a model training unit 604, which is used to: Acquire a training sample set, wherein the training sample set includes latent vectors of a plurality of sample players and game data of a plurality of historical games played between the plurality of sample players; The game data of each historical game is used as the sample label corresponding to the latent vector of the sample player in the historical game; Based on the latent vectors of the multiple sample players and their corresponding sample labels, supervised training is performed on the prediction model.

[0099] In an illustrated embodiment, the model training unit 604 is specifically used to: Based on the latent vectors of the plurality of sample players and their corresponding game indicators as sample labels, supervised training is performed on the game indicator prediction module; and Based on the latent vectors of the multiple sample players and their corresponding game results as sample labels, supervised training is performed on the game result prediction module.

[0100] In an illustrated embodiment, the model training unit 604 is specifically used to: Determine a plurality of first sample players to be studied, and further determine a plurality of second sample players who have played against at least N first sample players among the plurality of first players based on the game records of the plurality of first sample players; N is an integer greater than or equal to 1; Determining a plurality of historical matches between the plurality of first sample players and the plurality of second sample players; The game results and game videos of each historical game are obtained, and the game indicators of the first sample players and the second sample players in each historical game are marked according to the game videos.

[0101] In an illustrated embodiment, the latent vectors of the plurality of sample players are randomly initialized latent vectors; the apparatus 60 further includes a latent vector training unit 605, which is used to: Based on the latent vectors of the multiple sample players and their corresponding sample labels, supervised training is performed on the latent vectors of the multiple sample players to obtain latent vectors for representing the technical and tactical features of the multiple sample players.

[0102] In an illustrated embodiment, the target game includes a boxing match, and the match result of the target game includes a win, draw, or loss result of the boxers.

[0103] In addition, the embodiments of this specification also provide a latent vector training device, which can be applied to a computing device. Figure 7 , Figure 7 FIG. 1 is a schematic diagram of a latent vector training device provided by an exemplary embodiment. Figure 7 As shown, the device 70 includes: The acquisition unit 701 is used to acquire a training sample set, wherein the training sample set includes latent vectors of multiple sample players and game data of multiple historical games played between the multiple sample players; wherein the game data of each historical game includes the game result of the historical game and the game index of the sample players participating in the historical game, wherein the game index is used to describe the application of skills and tactics of the sample players during the game; The training unit 702 is used to use the game data of each historical game as a sample label corresponding to the latent vector of the sample player in the historical game, and input the latent vector of the sample player in the historical game into the prediction model, and based on the difference between the output result of the prediction model and the sample label, perform supervised training on the latent vector of the sample player to obtain a latent vector for representing the technical and tactical characteristics of the sample player.

[0104] The implementation process of the functions and effects of each unit in the above-mentioned device 60 and device 70 is specifically described in the above-mentioned embodiment, and will not be repeated here. It should be understood that the above-mentioned device 60 and device 70 can be implemented by software, or by hardware or a combination of software and hardware. Taking software implementation as an example, as a device in a logical sense, it is formed by the processor (CPU) of the device where it is located reading the corresponding computer program instructions into the memory and running them. From the hardware level, in addition to the CPU and memory, the device where the above-mentioned device is located usually also includes other hardware such as chips for wireless signal transmission and reception, and / or other hardware such as boards and cards for realizing network communication functions.

[0105] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical modules, that is, they may be located in one place, or they may be distributed on multiple network modules. Some or all of the units or modules may be selected according to actual needs to achieve the purpose of the scheme of this specification. Those of ordinary skill in the art may understand and implement it without creative work.

[0106] The devices, units, and modules described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer, which may be in the form of a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email transceiver, a game console, a tablet computer, a wearable device, a vehicle-mounted computer, or a combination of any of these devices.

[0107] Corresponding to the above method embodiment, the embodiment of this specification also provides a computing device. Figure 8 , Figure 8 FIG. 1 is a schematic diagram of a computing device provided by an exemplary embodiment. Figure 8 As shown, the computing device includes a processor 1001 and a memory 1002, and may further include an input device 1004 (such as a keyboard, etc.) and an output device 1005 (such as a display, etc.). The processor 1001, the memory 1002, the input device 1004, and the output device 1005 may be connected via a bus or other means. Figure 8 As shown, the memory 1002 includes a computer-readable storage medium 1003, which stores a computer program that can be run by the processor 1001. The processor 1001 can be a CPU, a microprocessor, or an integrated circuit for controlling the execution of the above method embodiment. When the processor 1001 runs the stored computer program, it can execute the various steps of the match result prediction method in the embodiment of this specification, including: determining multiple target players participating in the target game; obtaining the feature data of the multiple target players; wherein the feature data of each target player includes a latent vector for representing the technical and tactical characteristics of the target player obtained by training based on the game data of the historical games in which the target player has participated; the game data includes the game results of the historical games, and the game indicators of all the contestants, and the game indicators are used to describe the technical and tactical application of the contestants during the game; the latent vectors of the multiple target players are input into the prediction model obtained by training to predict the game results of the target game, and so on.

[0108] For a detailed description of each step of the above-mentioned match result prediction method, please refer to the previous content and will not be repeated here.

[0109] Corresponding to the above method embodiment, the embodiment of this specification also provides a computer-readable storage medium, on which a computer program is stored, and when these computer programs are executed by a processor, the various steps of the match result prediction method in the embodiment of this specification are executed. Please refer to the description of the above embodiment for details, which will not be repeated here.

[0110] The above description is only a preferred embodiment of this specification and is not intended to limit this specification. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of this specification should be included in the scope of protection of this specification.

[0111] In a typical configuration, a terminal device includes one or more CPUs, input / output interfaces, network interfaces, and memory.

[0112] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0113] Computer readable media include permanent and non-permanent, removable and non-removable media and can store information by any method or technology. The information can be computer readable instructions, data structures, program modules or other data.

[0114] Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory media such as modulated data signals and carrier waves.

[0115] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0116] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems or computer program products. Therefore, the embodiments of this specification may take the form of complete hardware embodiments, complete software embodiments or embodiments combining software and hardware. Moreover, the embodiments of this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

Claims

1. A method for predicting the results of a competitive match, characterized in that: The method comprises: Identify multiple target players participating in a target match; Acquire feature data of the multiple target players; wherein the feature data of each target player includes a latent vector obtained by training based on game data of historical games in which the target player has participated, and used to represent the technical and tactical features of the target player; the game data includes game results of historical games, and game indicators of all contestants, and the game indicators are used to describe the application of the contestants' skills and tactics during the game; The latent vectors of the multiple target players are input into the trained prediction model to predict the game result of the target game.

2. The method according to claim 1, characterized in that The prediction model includes a competition index prediction module and a competition result prediction module; Inputting the latent vectors of the plurality of target players into the trained prediction model to predict the result of the target game includes: Inputting the latent vectors of the plurality of target players into a game index prediction module to predict the game indexes of the plurality of target players in the target game; and The predicted game indicators of the multiple target players are further input into a game result prediction module to predict the game result of the target game.

3. The method according to claim 2, characterized in that The competition index prediction module includes a latent vector interaction module and an intermediate feature decoding module; The step of inputting the latent vectors of the plurality of target players into a competition index prediction module to predict competition indexes of the plurality of target players in the target competition includes: Inputting the latent vectors of the multiple target players into a latent vector interaction module for latent vector interaction processing, and outputting intermediate features obtained after the interaction; and The intermediate features are further input into an intermediate feature decoding module for decoding processing, and the game indicators of the multiple target players in the target game obtained after decoding are output.

4. The method according to claim 3, characterized in that The method further comprises: Acquire a training sample set, wherein the training sample set includes latent vectors of a plurality of sample players and game data of a plurality of historical games played between the plurality of sample players; The game data of each historical game is used as the sample label corresponding to the latent vector of the sample player in the historical game; Based on the latent vectors of the multiple sample players and their corresponding sample labels, supervised training is performed on the prediction model.

5. The method according to claim 4, characterized in that Based on the latent vectors of the plurality of sample players and their corresponding sample labels, supervised training is performed on the prediction model, including: Based on the latent vectors of the plurality of sample players and their corresponding game indicators as sample labels, supervised training is performed on the game indicator prediction module; and Based on the latent vectors of the multiple sample players and their corresponding game results as sample labels, supervised training is performed on the game result prediction module.

6. The method according to claim 5, characterized in that The step of obtaining a training sample set includes: Determine a plurality of first sample players to be studied, and further determine a plurality of second sample players who have played against at least N first sample players among the plurality of first players based on the game records of the plurality of first sample players; N is an integer greater than or equal to 1; Determining a plurality of historical matches between the plurality of first sample players and the plurality of second sample players; The game results and game videos of each historical game are obtained, and the game indicators of the first sample players and the second sample players in each historical game are marked according to the game videos.

7. The method according to claim 4, characterized in that The latent vectors of the plurality of sample players are randomly initialized latent vectors; the method further comprises: Based on the latent vectors of the multiple sample players and their corresponding sample labels, supervised training is performed on the latent vectors of the multiple sample players to obtain latent vectors for representing the technical and tactical features of the multiple sample players.

8. The method according to any one of claims 1 to 7, characterized in that: The target game includes a boxing game, and the game result of the target game includes the win, draw and loss results of the boxers.

9. A latent vector training method, characterized in that: The method comprises: Acquire a training sample set, wherein the training sample set includes latent vectors of multiple sample players and game data of multiple historical games played between the multiple sample players; wherein the game data of each historical game includes a game result of the historical game and game indicators of the sample players participating in the historical game, wherein the game indicators are used to describe the application of skills and tactics of the sample players during the game; The game data of each historical game is used as a sample label corresponding to the latent vector of the sample player in the historical game, and the latent vector of the sample player in the historical game is input into the prediction model. Based on the difference between the output result of the prediction model and the sample label, the latent vector of the sample player is supervisedly trained to obtain the latent vector used to represent the technical and tactical characteristics of the sample player.

10. A device for predicting the results of a competitive match, characterized in that: The device comprises: A determination unit, used for determining a plurality of target players participating in a target game; an acquisition unit, configured to acquire feature data of the plurality of target players; wherein the feature data of each target player comprises a latent vector obtained by training based on game data of historical games in which the target player has participated, and used to represent the technical and tactical features of the target player; the game data comprises game results of historical games, and game indicators of all participating players, and the game indicators are used to describe the application of the skills and tactics of the participating players during the game; The prediction unit is used to input the latent vectors of the multiple target players into the trained prediction model to predict the game result of the target game.

11. A latent vector training device, characterized in that: The device comprises: an acquisition unit, configured to acquire a training sample set, wherein the training sample set includes latent vectors of a plurality of sample players and game data of a plurality of historical games played between the plurality of sample players; wherein the game data of each historical game includes a game result of the historical game and game indicators of the sample players participating in the historical game, wherein the game indicators are used to describe the application of techniques and tactics of the sample players during the game; A training unit is used to use the game data of each historical game as a sample label corresponding to the latent vector of the sample player in the historical game, and input the latent vector of the sample player in the historical game into the prediction model, and based on the difference between the output result of the prediction model and the sample label, perform supervised training on the latent vector of the sample player to obtain a latent vector for representing the technical and tactical characteristics of the sample player.

12. A computing device, characterized in that: include: Memory and processor; The memory has stored thereon computer programs / instructions executable by the processor; When the processor runs the computer program / instructions, the method according to any one of claims 1-8 or 9 is performed.

13. A computer-readable storage medium, characterized in that: A computer program / instruction is stored thereon, and when the computer program / instruction is executed by a processor, the method according to any one of claims 1-8 or 9 is implemented.

14. A computer program product, characterized in that The computer program product comprises a computer program / instruction, and when the computer program / instruction is executed by a processor, the method according to any one of claims 1 to 8 or 9 is implemented.

Images