Personalizing prediction of performance using data and body-pose for analysis of sporting performance

Patent metadata

Field	Value
Publication number	US-11679299-B2
Application number	US-202016804964-A
Country	US
Kind code	B2
Filing date	Feb 28, 2020
Priority date	Mar 1, 2019
Publication date	Jun 20, 2023
Grant date	Jun 20, 2023

Abstract

A method of generating a player prediction is disclosed herein. A computing system retrieves data from a data store. The computing system generates a predictive model using an artificial neural network. The artificial neural network generates one or more personalized embeddings that include player-specific information based on historical performance. The computing system selects, from the data, one or more features related to each shot attempt captured in the data. The artificial neural network learns an outcome of each shot attempt based at least on the one or more personalized embeddings and the one or more features related to each shot attempt.

First claim

What is claimed:

1. A method of generating a player prediction, comprising:
retrieving, by a computing system, data from a data store, the data comprising information for a plurality of events across a plurality of seasons;
generating, by the computing system, a predictive model using an artificial neural network, by:
identifying a plurality of goalkeepers from the data;
for each goalkeeper of the plurality of goalkeepers, generating, by the artificial neural network, personalized embeddings based on the information, the personalized embeddings capturing an influence of the goalkeeper on a respective scoring event attempt;
selecting, from the data, a set of features related to each scoring event attempt captured in the data; and
learning, by the artificial neural network, an outcome of each scoring event attempt based at least on the personalized embeddings and the set of features related to each scoring event attempt;

receiving, by the computing system, a set of data directed to a target scoring event attempt, the set of data comprising an indication of at least a target goalkeeper involved in the target scoring event attempt and one or more features related to the target scoring event attempt, the one or more features related to the target scoring event attempt comprising a first set of location coordinates cooresponding to an origination location of an offensive player initiating the target scoring event attempt and a second set of location coordinates corresponding to an initial position of the target goalkeeper when the offensive player initiated the group scoring even attempt; and
generating, by the computing system via the predictive model, a likely outcome of the target scoring event attempt based on target personalized embeddings of the target goalkeeper and the one or more features related to the target scoring event attempt.

2. The method of claim 1 , wherein selecting, from the data, the one or more features related to each scoring event attempt captured in the data, comprises:
for each scoring event attempt, identifying at least one or more of scoring event start location information, goalkeeper location, and one or more geometric features of a corresponding scoring event attempt.

3. The method of claim 2 , wherein the one or more geometric features of the corresponding scoring event attempt comprises at least one or more of an angle between a respective offensive player and a respective goalkeeper, a first distance from the respective offensive player to the center of a goal, and a second distance from the respective goalkeeper to the center of the goal.

4. The method of claim 2 , further comprising:
for each scoring event attempt, identifying body pose information related to a respective offensive player of the corresponding scoring event attempt.

5. The method of claim 1 , further comprising:
identifying, by the computing system, a set of scoring event attempts over a first duration;
simulating, by the computing system, a number of scoring event attempts an average goalkeeper would concede based on one or more parameters associated with the set of scoring event attempts;
identifying, by the computing system, a set of goalkeepers, each goalkeeper associated with a respective set of embeddings;
for each goalkeeper in the set of goalkeepers, simulating a number of scoring event attempts a corresponding goalkeeper would concede based on the one or more parameters associated with the set of scoring event attempts and a respective set of embeddings; and
generating, by the computing system, a graphical representation ranking each goalkeeper of the set of goalkeepers based on expected scoring events conceded compared to the average goalkeeper.

6. The method of claim 1 , further comprising:
identifying, by the computing system, a first goalkeeper and one or more scoring event attempts defended by the first goalkeeper over a first duration;
generating, by the computing system, a data set corresponding to one or more parameters associated with the one or more scoring event attempts defended by the first player over the first duration;
identifying, by the computing system, a second goalkeeper, wherein the second goalkeeper is associated with a set of embeddings;
simulating, by the computing system, a number of goals the second goalkeeper would concede based on the one or more parameters associated with the one or more scoring event attempts defended by the first goalkeeper and the one or more personalized embeddings; and
generating, by the computing system, a graphical representation comparing the number of goals the second goalkeeper would concede compared to a number of goals the first goalkeeper conceded.

7. The method of claim 1 , further comprising:
identifying, by the computing system, a goalkeeper and one or more scoring event attempts defended by the goalkeeper over a first duration;
generating, by the computing system, a data set corresponding to one or more parameters associated with the one or more scoring event attempts defended by the goalkeeper over the first duration;
identifying, by the computing system, one or more embeddings associated with the goalkeeper, wherein the set of embeddings correspond to attributes of the goalkeeper over a second duration;
simulating, by the computing system, a number of goals the goalkeeper would concede based on the one or more parameters associated with the one or more scoring event attempts defended by the goalkeeper and the set of embeddings corresponding to the attributes of the goalkeeper over the second duration; and
generating, by the computing system, a graphical representation comparing the number of goals the goalkeeper would concede based on the attributes over the second duration compared to a number of goals the goalkeeper conceded in the first duration.

8. A system for generating a goalkeeper prediction, comprising:
a processor; and
a memory having programming instructions stored thereon, which, when executed by the processor, performs one or more operations, comprising:
retrieving data from a data store, the data comprising information for a plurality of events across a plurality of seasons;
generating a predictive model using an artificial neural network, the predictive model trained to predict an outcome of a shot attempt, by:
identifying a plurality of goalkeepers from the data;
for each goalkeeper, generating, by the artificial neural network, personalized embeddings based on the information, the personalized embeddings capturing an influence of the goalkeeper on a respective shot attempt;
selecting, from the data, a set of features related to each shot attempt captured in the data; and
learning, by the artificial neural network, an outcome of each shot attempt based at least on the one or more personalized embeddings and the set of features related to each shot attempt;

receiving a set of data directed to a target shot attempt, the set of data comprising an indication of a target goalkeeper involved in the target shot attempt and one or more features related to the target shot attempt, the one or more features related to the target shot attempt comprising a first set of location coordinates corresponding to an origination location of an offensive goalkeeper initiating the target shot attempt and a second set of location coordinates corresponding to an initial position of the target goalkeeper when the offensive goalkeeper initiated the target shot attempt; and
generating, via the predictive model, a likely outcome of the target shot attempt based on target personalized embeddings of the target goalkeeper and the one or more features related to the target shot attempt.

9. The system of claim 8 , wherein selecting, from th

Assignees

• Stats Llc

Inventors

• Power Paul David
• Cherukumudi Aditya
• Ganguly Sujoy
• Wei Xinyu
• Sha Long
• Hobbs Jennifer
• Ruiz Hector
• Lucey Patrick Joseph

Classifications

• A63B24/0006 Primary

  Computerised comparison for qualitative assessment of motion sequences or the course of a movement · CPC title

• A63B24/0087

  Electric or electronic controls for exercising apparatus of groups A63B21/00 - A63B23/00, e.g. controlling load · CPC title

• G06N20/00

  Machine learning · CPC title

• A63B24/0021

  Tracking a path or terminating locations · CPC title

• G06N3/08 Primary

  Learning methods · CPC title