Method and System for Interactive, Interpretable, and Improved Match and Player Performance Predictions in Team Sports

Abstract

A method of generating an outcome for a sporting event is disclosed herein. A computing system retrieves tracking data from a data store. The computing system generates a predictive model using a deep neural network. The one or more neural networks of the deep neural network generates one or more embeddings comprising team-specific information and agent-specific information based on the tracking data. The computing system selects, from the tracking data, one or more features related to a current context of the sporting event. The computing system learns, by the deep neural network, one or more likely outcomes of one or more sporting events. The computing system receives a pre-match lineup for the sporting event. The computing system generates, via the predictive model, a likely outcome of the sporting event based on historical information of each agent for the home team, each agent for the away team, and team-specific features.

What is claimed: 
     
         1 . A method of generating an outcome for a sporting event, comprising:
 accessing, by a computing system, event data for a sporting event;   identifying, by the computing system, a home team and an away team for the sporting event;   accessing, by the computing system, historical event data from a data store, the historical event data comprising play-by-play information for a plurality of historical sporting events across a plurality of seasons for the home team and the away team;   generating, by the computing system via a first neural network, team history embeddings for the home team and the away team based on the historical event data;   generating, by the computing system via a second neural network, a first set of player embeddings for the home team and a second set of player embeddings for the away team based on the historical event data;   identifying, by the computing system from the event data, a pre-match lineup for the sporting event, the pre-match lineup comprising a first plurality of players for the home team and a second plurality of players for the away team; and   generating, by the computing system via a predictive model, a likely outcome of the sporting event based on the team history embeddings, the first set of player embeddings, the second set of player embeddings, and the pre-match lineup.   
     
     
         2 . The method of  claim 1 , wherein generating, by the computing system via the first neural network, the team history embeddings for the home team and the away team based on the historical event data comprises:
 generating team context data for the home team and the away team based on the historical event data, the team context data comprising a plurality of features directed to characteristics of the home team and the away team;   inputting the team context data into the first neural network; and   generating the team history embeddings for the home team and the away team based on the team context data.   
     
     
         3 . The method of  claim 1 , wherein generating, by the computing system via the second neural network, the first set of player embeddings for the home team and the second set of player embeddings for the away team based on the historical event data comprises:
 generating player context data for each player associated with the home team and each player associated with the away team based on the historical event data;   inputting the player context data into the second neural network; and   generating the first set of player embeddings for the home team and the second set of player embeddings for the away team based on the player context data.   
     
     
         4 . The method of  claim 3 , wherein generating the player context data for each player associated with the home team and each player associated with the away team based on the historical event data comprises:
 for each player associated with the home team, aggregating all values associated with the player across the plurality of historical sporting events; and   for each player associated with the away team, aggregating all values associated with the player across the plurality of historical sporting events.   
     
     
         5 . The method of  claim 1 , further comprising:
 generating, by the computing system via third neural network, a first set of recent player embeddings for the home team and a second set of recent player embeddings for the away team based on a subset of the historical event data over a predefined time range.   
     
     
         6 . The method of  claim 1 , further comprising:
 identifying a game context corresponding to the sporting event, the game context comprising information related to a relationship between the home team and the away team across the plurality of historical sporting events.   
     
     
         7 . The method of  claim 6 , wherein the predictive model generates the likely outcome based on the game context. 
     
     
         8 . A non-transitory computer readable medium comprising one or more sequences of instructions, which, when executed by one or more processors, causes a computing system to perform operations comprising:
 accessing, by the computing system, event data for a sporting event;   identifying, by the computing system, a home team and an away team for the sporting event;   accessing, by the computing system, historical event data from a data store, the historical event data comprising play-by-play information for a plurality of historical sporting events across a plurality of seasons for the home team and the away team;   generating, by the computing system via a first neural network, team history embeddings for the home team and the away team based on the historical event data;   generating, by the computing system via a second neural network, a first set of player embeddings for the home team and a second set of player embeddings for the away team based on the historical event data;   identifying, by the computing system from the event data, a pre-match lineup for the sporting event, the pre-match lineup comprising a first plurality of players for the home team and a second plurality of players for the away team; and   generating, by the computing system via a predictive model, a likely outcome of the sporting event based on the team history embeddings, the first set of player embeddings, the second set of player embeddings, and the pre-match lineup.   
     
     
         9 . The non-transitory computer readable medium of  claim 8 , wherein generating, by the computing system via the first neural network, the team history embeddings for the home team and the away team based on the historical event data comprises:
 generating team context data for the home team and the away team based on the historical event data, the team context data comprising a plurality of features directed to characteristics of the home team and the away team;   inputting the team context data into the first neural network; and   generating the team history embeddings for the home team and the away team based on the team context data.   
     
     
         10 . The non-transitory computer readable medium of  claim 8 , wherein generating, by the computing system via the second neural network, the first set of player embeddings for the home team and the second set of player embeddings for the away team based on the historical event data comprises:
 generating player context data for each player associated with the home team and each player associated with the away team based on the historical event data;   inputting the player context data into the second neural network; and   generating the first set of player embeddings for the home team and the second set of player embeddings for the away team based on the player context data.   
     
     
         11 . The non-transitory computer readable medium of  claim 10 , wherein generating the player context data for each player associated with the home team and each player associated with the away team based on the historical event data comprises:
 for each player associated with the home team, aggregating all values associated with the player across the plurality of historical sporting events; and   for each player associated with the away team, aggregating all values associated with the player across the plurality of historical sporting events.   
     
     
         12 . The non-transitory computer readable medium of  claim 8 , further comprising:
 generating, by the computing system via third neural network, a first set of recent player embeddings for the home team and a second set of recent player embeddings for the away team based on a subset of the historical event data over a predefined time range.   
     
     
         13 . The non-transitory computer readable medium of  claim 8 , further comprising:
 identifying a game context corresponding to the sporting event, the game context comprising information related to a relationship between the home team and the away team across the plurality of historical sporting events.   
     
     
         14 . The non-transitory computer readable medium of  claim 13 , wherein the predictive model generates the likely outcome based on the game context 
     
     
         15 . A system comprising:
 a processor; and   a memory having programming instructions stored thereon, which, when executed by the processor, causes a computing system to perform operations comprising:   accessing event data for a sporting event;   identifying a home team and an away team for the sporting event;   accessing historical event data from a data store, the historical event data comprising play-by-play information for a plurality of historical sporting events across a plurality of seasons for the home team and the away team;   generating, via a first neural network, team history embeddings for the home team and the away team based on the historical event data;   generating, via a second neural network, a first set of player embeddings for the home team and a second set of player embeddings for the away team based on the historical event data;   identifying, from the event data, a pre-match lineup for the sporting event, the pre-match lineup comprising a first plurality of players for the home team and a second plurality of players for the away team; and   generating, via a predictive model, a likely outcome of the sporting event based on the team history embeddings, the first set of player embeddings, the second set of player embeddings, and the pre-match lineup.   
     
     
         16 . The system of  claim 15 , wherein generating, via the first neural network, the team history embeddings for the home team and the away team based on the historical event data comprises:
 generating team context data for the home team and the away team based on the historical event data, the team context data comprising a plurality of features directed to characteristics of the home team and the away team;   inputting the team context data into the first neural network; and   generating the team history embeddings for the home team and the away team based on the team context data.   
     
     
         17 . The system of  claim 15 , wherein generating, via the second neural network, the first set of player embeddings for the home team and the second set of player embeddings for the away team based on the historical event data comprises:
 generating player context data for each player associated with the home team and each player associated with the away team based on the historical event data;   inputting the player context data into the second neural network; and   generating the first set of player embeddings for the home team and the second set of player embeddings for the away team based on the player context data.   
     
     
         18 . The system of  claim 17 , wherein generating the player context data for each player associated with the home team and each player associated with the away team based on the historical event data comprises:
 for each player associated with the home team, aggregating all values associated with the player across the plurality of historical sporting events; and   for each player associated with the away team, aggregating all values associated with the player across the plurality of historical sporting events.   
     
     
         19 . The system of  claim 15 , wherein the operations further comprise:
 generating, via third neural network, a first set of recent player embeddings for the home team and a second set of recent player embeddings for the away team based on a subset of the historical event data over a predefined time range.   
     
     
         20 . The system of  claim 15 , wherein the operations further comprise:
 identifying a game context corresponding to the sporting event, the game context comprising information related to a relationship between the home team and the away team across the plurality of historical sporting events.