Dynamically predicting shot type using a personalized deep neural network

1. A method for predicting a shot type, comprising:

retrieving, by a computing system, ball-by-ball data for a plurality of sporting events;

generating, by the computing system, a trained neural network, by:

generating a plurality of training data sets based on the ball-by-ball data by supplementing the ball-by-ball data with ball-by-ball match context features;

generating, from the ball-by-ball data, personalized embeddings based on a batsman and a bowler for each delivery; and

learning, by a neural network associated with the computing system, to predict a shot type based on the ball-by-ball data and the personalized embeddings;

receiving, by the computing system, a target batsman and a target bowler for a pitch to be delivered in a target event;

identifying, by the computing system, target ball-by-ball data for a window of pitches preceding the pitch to be delivered;

retrieving, by the computing system, historical ball-by-ball data for each of the target batsman and the target bowler;

generating, by the computing system, personalized embeddings for both the target batsman and the target bowler based on the historical ball-by-ball data; and

predicting, by the computing system using the trained neural network, a shot type for the pitch to be delivered based on the target ball-by-ball data and the personalized embeddings.

2. The method of claim 1 , wherein the neural network comprises:

a long term-short term memory (LSTM) network; and

a feed forward neural network.

3. The method of claim 2 , wherein learning, by the neural network, to predict the shot type based on the ball-by-ball data and the personalized embeddings comprises:

inputting the ball-by-ball data into the LSTM network; and

generating, by the LSTM network, an output based on the ball-by-ball data.

4. The method of claim 3 , further comprising:

concatenating the output of the LSTM network with the personalized embeddings to generate concatenated data; and

inputting the concatenated data into the feed forward neural network.

5. The method of claim 1 , wherein the shot type prediction comprises a predicted location on a field in which a hit pitch will land.

6. The method of claim 1 , wherein identifying, by the computing system, the target ball-by-ball data for the window of pitches preceding the to be delivered pitch comprises:

identifying a last X pitches bowled by a team of the target bowler, where X is a positive integer.

7. The method of claim 1 , further comprising:

generating a graphical representation of a location on a field in which a hit pitch is likely to land.

8. A system for predicting a shot type, comprising:

a processor; and

a memory having programming instructions stored thereon, which, when executed by the processor, causes the system to perform one or more operations, comprising:

retrieving ball-by-ball data for a plurality of sporting events;

generating a trained neural network, by:

generating a plurality of training data sets based on the ball-by-ball data by supplementing ball-by-ball data with ball-by-ball match context features;

generating, from the ball-by-ball data, personalized embeddings based on a batsman and a bowler for each delivery; and

learning, by a neural network, to predict a shot type based on the ball-by-ball data and the personalized embeddings;

receiving a target batsman and a target bowler for a pitch to be delivered in a target event;

identifying target ball-by-ball data for a window of pitches preceding the pitch to be delivered;

retrieving historical ball-by-ball data for each of the target batsman and the target bowler;

generating personalized embeddings for both the target batsman and the target bowler based on the historical ball-by-ball data; and

predicting a shot type for pitch the to be delivered based on the target ball-by-ball data and the personalized embeddings.

9. The system of claim 8 , wherein the neural network comprises:

a long term-short term memory (LSTM) network; and

a feed forward neural network.

10. The system of claim 9 , wherein learning, by the neural network, to predict the shot type based on the ball-by-ball data and the personalized embeddings comprises:

inputting the ball-by-ball data into the LSTM network; and

generating, by the LSTM network, an output based on the ball-by-ball data.

11. The system of claim 10 , wherein the one or more operations further comprise:

concatenating the output of the LSTM network with the personalized embeddings to generate concatenated data; and

inputting the concatenated data into the feed forward neural network.

12. The system of claim 8 , wherein the shot type prediction comprises a predicted location on a field in which a hit pitch will land.

13. The system of claim 8 , wherein identifying the target ball-by-ball data for the window of pitches preceding the to be delivered pitch comprises:

identifying a last X pitches bowled by a team of the target bowler, where X is a positive integer.

14. The system of claim 8 , wherein the one or more operations further comprise:

generating a graphical representation of a location on a field in which a hit pitch is likely to land.

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, ball-by-ball data for a plurality of sporting events;

generating, by the computing system, a trained neural network, by:

generating a plurality of training data sets based on the ball-by-ball data by supplementing ball-by-ball data with ball-by-ball match context features;

generating, from the ball-by-ball data, personalized embeddings based on a batsman and a bowler for each delivery; and

learning, by a neural network, to predict a shot type based on the ball-by-ball data and the personalized embeddings;

receiving, by the computing system, a target batsman and a target bowler for a to be delivered pitch in a target event;

identifying, by the computing system, target ball-by-ball data for a window of pitches preceding the to be delivered pitch;

retrieving, by the computing system, historical ball-by-ball data for each of the target batsman and the target bowler;

generating, by the computing system, personalized embeddings for both the target batsman and the target bowler based on the historical ball-by-ball data; and

predicting, by the computing system, a shot type for the to be delivered pitch based on the target ball-by-ball data and the personalized embeddings.

16. The non-transitory computer readable medium of claim 15 , wherein the neural network comprises:

a long term-short term memory (LSTM) network; and

a feed forward neural network.

17. The non-transitory computer readable medium of claim 16 , wherein learning, by the neural network, to predict the shot type based on the ball-by-ball data and the personalized embeddings comprises:

inputting the ball-by-ball data into the LSTM network; and

generating, by the LSTM network, an output based on the ball-by-ball data.

18. The non-transitory computer readable medium of claim 17 , further comprising:

concatenating the output of the LSTM network with the personalized embeddings to generate concatenated data; and

inputting the concatenated data into the feed forward neural network.

19. The non-transitory computer readable medium of claim 15 , wherein the shot type prediction comprises a predicted location on a field in which a hit pitch will land.

20. The non-transitory computer readable medium of claim 15 , wherein identifying, by the computing system, the target ball-by-ball data for the window of pitches preceding the to be delivered pitch comprises:

identifying a last X pitches bowled by a team of the target bowler, where X is a positive integer.