Graph Based Method of Next Pitch Prediction
A system and method for predicting next pitch are disclosed herein. A computing system retrieves pitch-by-pitch information for a plurality of events and game context information associated with each pitch in the pitch-by-pitch information. The computing system converts the pitch-by-pitch information and the game context information into a plurality of graph-based representation. A graph neural network learns to generate a pitch type prediction for each pitch based on the plurality of graph-based representations. The computing system generates a trained graph neural network based on the learning. The computing system receives a current graph-based representation of current pitch-by-pitch information for a current pitcher and current game context information. The computing system predicts, via the trained graph neural network, a pitch type for the next pitch to be delivered from the current pitcher.
1 . A method for predicting a next pitch, comprising:
retrieving, by a computing system, pitch-by-pitch information for a plurality of events and game context information associated with each pitch in the pitch-by-pitch information;
converting, by the computing system, the pitch-by-pitch information and the game context information into a plurality of graph-based representations;
learning, by a graph neural network, to generate a pitch type prediction for each pitch based on the plurality of graph-based representations;
generating, by the computing system, a trained graph neural network based on the learning;
receiving, by the computing system, a current graph-based representation of current pitch-by-pitch information for a current pitcher and current game context information; and
predicting, by the computing system, via the trained graph neural network, a pitch type for the next pitch to be delivered from the current pitcher.
2 . The method of claim 1 , wherein the pitch-by-pitch information comprises pitch type labels for each pitch in the pitch-by-pitch information.
3 . The method of claim 1 , wherein the game context information associated with each pitch in the pitch-by-pitch information comprises a pitch count, an inning, a score, runners on base, and result of the pitch.
4 . The method of claim 1 , wherein converting, by the computing system, the pitch-by-pitch information and the game context information into the plurality of graph-based representations, comprises:
generating, for each pitcher in each event, a graph-based representation of pitches delivered, wherein each node in the graph-based representation corresponds to a respective pitch delivered by the respective pitcher and wherein each node comprises at least a type of pitch corresponding thereto.
5 . The method of claim 4 , further comprising:
connecting each node in the graph-based representation with an edge that denotes a sequential ordering of the pitches delivered.
6 . The method of claim 5 , further comprising:
generating a global variable data set for each node based on the pitch-by-pitch information and the game context information, wherein the global variable data set does not include a pitch type corresponding to the pitch associated with each node.
7 . The method of claim 6 , wherein the graph neural network includes a first multilayer perceptron layer, a second multilayer perceptron layer, and a third multilayer perceptron layer, and wherein learning, by the graph neural network, to generate the pitch type prediction for each pitch based on the plurality of graph-based representations, comprises:
identifying a leading node and a trailing node for each edge, and inputting node information associated with the leading node and the trailing node into the first multilayer perceptron layer;
outputting, by the first multilayer perceptron layer, an edge embedding corresponding to each edge;
providing, as input to the second multilayer perceptron layer, edge embedding information for each edge and node information for each node; and
providing, as input to the third multilayer perceptron layer, the global variable data set and an aggregation of the embeddings corresponding to the edges and nodes in the graph-based representation.
8 . A system for predicting a next pitch, comprising:
a processor; and
a memory having programming instructions stored thereon, which, when executed by the processor, performs one or more operations, comprising:
retrieving pitch-by-pitch information for a plurality of events and game context information associated with each pitch in the pitch-by-pitch information;
converting the pitch-by-pitch information and the game context information into a plurality of graph-based representations;
learning, by a graph neural network, to generate a pitch type prediction for each pitch based on the plurality of graph-based representations;
generating a trained graph neural network based on the learning;
receiving a current graph-based representation of current pitch-by-pitch information for a current pitcher and current game context information; and
predicting via the trained graph neural network, a pitch type for the next pitch to be delivered from the current pitcher.
9 . The system of claim 8 , wherein the pitch-by-pitch information comprises pitch type labels for each pitch in the pitch-by-pitch information.
10 . The system of claim 8 , wherein the game context information associated with each pitch in the pitch-by-pitch information comprises a pitch count, an inning, a score, runners on base, and result of the pitch.
11 . The system of claim 8 , wherein converting the pitch-by-pitch information and the game context information into the plurality of graph-based representations, comprises:
generating, for each pitcher in each event, a graph-based representation of pitches delivered, wherein each node in the graph-based representation corresponds to a respective pitch delivered by the respective pitcher and wherein each node comprises at least a type of pitch corresponding thereto.
12 . The system of claim 11 , further comprising:
connecting each node in the graph-based representation with an edge that denotes a sequential ordering of the pitches delivered.
13 . The system of claim 12 , further comprising:
generating a global variable data set based on the game context information, wherein the global variable data set does not include a pitch type corresponding to the pitch.
14 . The system of claim 13 , wherein the graph neural network includes a first multilayer perceptron layer, a second multilayer perceptron layer, and a third multilayer perceptron layer, and wherein learning, by the graph neural network, to generate the pitch type prediction for each pitch based on the plurality of graph-based representations, comprises:
identifying a leading node and a trailing node for each edge, and inputting node information associated with the leading node and the trailing node into the first multilayer perceptron layer;
outputting, by the first multilayer perceptron layer, an edge embedding corresponding to each edge;
providing, as input to the second multilayer perceptron layer, edge embedding information for each edge and node information for each node; and
providing, as input to the third multilayer perceptron layer, the global variable data set and an aggregation of the embeddings corresponding to the edges and nodes.
15 . A non-transitory computer readable medium including one or more sequences of instructions that, when executed by one or more processors, causes:
retrieving, by a computing system, pitch-by-pitch information for a plurality of events and game context information associated with each pitch in the pitch-by-pitch information;
converting, by the computing system, the pitch-by-pitch information and the game context information into a plurality of graph-based representations;
learning, by a graph neural network, to generate a pitch type prediction for each pitch based on the plurality of graph-based representations;
generating, by the computing system, a trained graph neural network based on the learning;
receiving, by the computing system, a current graph-based representation of current pitch-by-pitch information for a current pitcher and current game context information; and
predicting, by the computing system, via the trained graph neural network, a pitch type for a next pitch to be delivered from the current pitcher.
16 . The non-transitory computer readable medium of claim 15 , wherein the pitch-by-pitch information comprises pitch type labels for each pitch in the pitch-by-pitch information.
17 . The non-transitory computer readable medium of claim 15 , wherein the game context information associated with each pitch in the pitch-by-pitch information comprises a pitch count, an inning, a score, runners on base, and result of the pitch.
18 . The non-transitory computer readable medium of claim 15 , wherein converting, by the computing system, the pitch-by-pitch information and the game context information into the plurality of graph-based representations, comprises:
generating, for each pitcher in each event, a graph-based representation of pitches delivered, wherein each node in the graph-based representation corresponds to a respective pitch delivered by the respective pitcher and wherein each node comprises at least a type of pitch corresponding thereto.
19 . The non-transitory computer readable medium of claim 18 , further comprising:
connecting each node in the graph-based representation with an edge that denotes a sequential ordering of the pitches delivered.
20 . The non-transitory computer readable medium of claim 19 , further comprising:
generating a global variable data set for the game context information for the pitch, wherein the global variable data set does not include a pitch type corresponding to the pitch.