System and Method for Predicting Formation in Sports

Patent metadata

Field	Value
Publication number	US-2021383123-A1
Application number	US-202117303361-A
Country	US
Kind code	A1
Filing date	May 27, 2021
Priority date	Jun 5, 2020
Publication date	Dec 9, 2021
Grant date	—

Abstract

A system and method of predicting a team's formation on a playing surface are disclosed herein. A computing system retrieves one or more sets of event data for a plurality of events. Each set of event data corresponds to a segment of the event. A deep neural network, such as a mixture density network, learns to predict an optimal permutation of players in each segment of the event based on the one or more sets of event data. The deep neural network learns a distribution of players for each segment based on the corresponding event data and optimal permutation of players. The computing system generates a fully trained prediction model based on the learning. The computing system receives target event data corresponding to a target event. The computing system generates, via the trained prediction model, an expected position of each player based on the target event data.

First claim

1 . A method of predicting a team's formation on a playing surface, comprising:
retrieving, by a computing system, one or more sets of event data for a plurality of events, wherein each set of event data corresponds to a segment of a respective event; learning, by a deep neural network, to predict an optimal permutation of players in each segment of each respective event based on the one or more sets of event data; learning, by the deep neural network, a distribution of players for each segment based on the corresponding event data retrieved from data store and optimal permutation of players; generating, by the computing system, a fully trained prediction model based on the learning; receiving, by the computing system, target event data corresponding to a target event, the target event data comprising information directed to a team comprising a plurality of players on a target playing surface; and generating, by the trained prediction model, an expected position of each player of the plurality of players on the target playing surface based on the target event data.

2 . The method of claim 1 , further comprising:
for each set of event data, determining, by the computing system, a number of players on the playing surface; and parameterizing, by the computing system, the deep neural network based on the number of players on the playing surface.

3 . The method of claim 1 , wherein learning, by the deep neural network, to predict the optimal permutation of players in each segment of the event based on the one or more sets of event data comprises:
softly-assigning each player to a role based on the event information to generate a plurality of possible permutations.

4 . The method of claim 1 , wherein learning, by the deep neural network, the distribution of players for each segment based on the corresponding event data retrieved from the data store and the optimal permutation of players comprises:
learning to predict an underlying distribution of players in the optimal permutation of players.

5 . The method of claim 1 , wherein generating, by the trained prediction model, the expected position of each player of the plurality of players on the playing surface based on the target event data, comprises:
generating, as output, a 2p-dimensional distribution, that describes each player's positioning on the playing surface, wherein p represents a number of players of the plurality of players.

6 . The method of claim 1 , further comprising:
generating, by the computing system, a semantic label of a formation corresponding to the expected position of each player of the plurality of players on the target playing surface. The method of claim 1 , further comprising: receiving, by the computing system, a broadcast stream of the target event; determining, by the computing system, that at least one target player is absent from a video frame of the broadcast stream; and identifying, by the computing system, an inferred position of the at least one target player based on the expected position of each player of the plurality of players on the target playing surface.

8 . A system for predicting a team's formation on a playing surface, comprising:
a processor; and a memory having programming instructions stored thereon, which, when executed by the processor, performs one or more operations, comprising:
retrieving one or more sets of event data for a plurality of events, wherein each set of event data corresponds to a segment of each respective event;
learning, by a deep neural network, to predict an optimal permutation of players in each segment of each respective event based on the one or more sets of event data;
learning, by the deep neural network, a distribution of players for each segment based on the corresponding event data retrieved from data store and optimal permutation of players;
generating a fully trained prediction model based on the learning;
receiving target event data corresponding to a target event, the target event data comprising information directed to a team comprising a plurality of players on a target playing surface; and
generating, by the trained prediction model, an expected position of each player of the plurality of players on the target playing surface based on the target event data.

9 . The system of claim 8 , wherein the one or more operations further comprise:
for each set of event data, determining a number of players on a playing surface; and parameterizing the deep neural network based on the number of players on the playing surface.

10 . The system of claim 8 , wherein learning, by the deep neural network, to predict the optimal permutation of players in each segment of the event based on the one or more sets of event data comprises:
softly-assigning each player to a role based on the event information to generate a plurality of possible permutations.

11 . The system of claim 8 , wherein learning, by the deep neural network, the distribution of players for each segment based on the corresponding event data retrieved from data store and the optimal permutation of players comprises:
learning to predict an underlying distribution of players in the optimal permutation of players.

12 . The system of claim 8 , wherein generating, by the trained prediction model, an expected position of each player of the plurality of players on the playing surface based on the target event data, comprises:
generating, as output, a 2p-dimensional distribution, that describes each player's positioning on the playing surface, wherein p represents a number of players of the plurality of players.

13 . The system of claim 8 , wherein the one or more operations further comprise:
generating a semantic label of a formation corresponding to the expected position of each player of the plurality of players on the target playing surface.

14 . The system of claim 8 , wherein the one or more operations further comprise:
receiving a broadcast stream of the target event; determining that at least one target player is absent from a video frame of the broadcast stream; and identifying an inferred position of the at least one target player based on the expected position of each player of the plurality of players on the target playing surface.

15 . A non-transitory computer readable medium including one or more sequences of instructions that, when executed by one or more processors, causes a computing system to perform operations, comprising:
retrieving, by the computing system, one or more sets of event data for a plurality of events, wherein each set of event data corresponds to a segment of each respective event; learning, by a deep neural network, to predict an optimal permutation of players in each segment of each respective event based on the one or more sets of event data; learning, by the deep neural network, a distribution of players for each segment based on the corresponding event data retrieved from data store and optimal permutation of players; generating, by the computing system, a fully trained prediction model based on the learning; receiving, by the computing system, target event data corresponding to a target event, the target event data comprising information directed to a team comprising a plurality of players on a playing surface; and generating, by the trained prediction model, an expected position of each player of the plurality of players on the playing surface based on the target event data.

16 . The non-transitory computer readable medium of claim 15 , further comprising:
for each set of event data, determining, by the computing system, a number of p

Assignees

• Stats Llc

Inventors

• Hobbs Jennifer
• Ganguly Sujoy
• Lucey Patrick Joseph

Classifications

• G06N3/084 Primary

  Backpropagation, e.g. using gradient descent · CPC title

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G06N3/0499

  Feedforward networks · CPC title

• G06N3/09

  Supervised learning · CPC title