SYSTEMS AND METHODS FOR A TRANSFORMER NEURAL NETWORK FOR PREDICTIONS IN POSSESSION-BASED SPORTING EVENTS

1 . A method of generating a set of predictions associated with a possession-based sporting event using an axial transformer neural network, the method comprising:

receiving an input tuple, including a set of tensors representing game context, team strength, player strength, live team features, live player features, game events, and a super feature, wherein the super feature includes a current lineup based on a current possession in a possession-based sporting event at a particular time during the sporting event;

inputting the input tuple into an axial transformer neural network by inputting each tensor from the set of tensors within a corresponding initial embedding layer;

concatenating the initial embedding layers to form a single tensor;

applying self-attention to the single tensor through axial transformer layers of the axial transformer neural network;

mapping output embeddings from the axial transformer layers to target layers, each of the output embeddings being of a dimension of a target metric; and

generating a set of target metric predictions for each of a set of players, one or more teams, and a match, based on the output embeddings from the target layers.

2 . The method of claim 1 , wherein the super feature includes a playbook embedding that defines performance for certain types of players during a particular player in the possession-based sporting event.

3 . The method of claim 1 , wherein possession-based sporting events include football games, hockey games, and basketball games.

4 . The method of claim 1 , wherein the axial transformer neural network is configured to accept inputs with different modalities.

5 . The method of claim 1 , wherein the super feature is determined based on broadcast data.

6 . The method of claim 1 , wherein the applying self-attention includes applying an autoregressive attention mask to a row in each layer of the single tensor.

7 . The method of claim 1 , wherein the target layers map the output embedding of final transformer layers to a required feature dimension of each target metric.

8 . A system for generating a set of predictions associated with a possession-based sporting event using an axial transformer neural network, the system comprising:

a memory configured to store processor-readable instructions; and

a processor operatively connected to the memory, and configured to execute the instructions to perform operations comprising:

receiving an input tuple, including a set of tensors representing game context, team strength, player strength, live team features, live player features, game events, and a super feature, wherein the super feature includes a current lineup based on a current possession in a possession-based sporting event at a particular time during the sporting event;

inputting the input tuple into an axial transformer neural network by inputting each tensor from the set of tensors within a corresponding initial embedding layer;

concatenating the initial embedding layers to form a single tensor;

applying self-attention to the single tensor through axial transformer layers of the axial transformer neural network;

mapping output embeddings from the axial transformer layers to target layers, each of the output embeddings being of a dimension of a target metric; and

generating a set of target metric predictions for each of a set of players, one or more teams, and a match, based on the output embeddings from the target layers.

9 . The system of claim 8 , wherein the super feature includes a playbook embedding that defines performance for certain types of players during a particular player in the possession-based sporting event.

10 . The system of claim 8 , wherein possession-based sporting events include football games, hockey games, and basketball games.

11 . The system of claim 8 , wherein the axial transformer neural network is configured to accept inputs with different modalities.

12 . The system of claim 8 , wherein the super feature is determined based on broadcast data.

13 . The system of claim 8 , wherein the applying self-attention includes applying an autoregressive attention mask to a row in each layer of the single tensor.

14 . The system of claim 8 , wherein the target layers map the output embedding of final transformer layers to a required feature dimension of each target metric.

15 . A non-transitory computer readable medium configured to store processor-readable instructions, wherein when executed by a processor, the instructions perform operations comprising:

receiving an input tuple, including a set of tensors representing game context, team strength, player strength, live team features, live player features, game events, and a super feature, wherein the super feature includes a current lineup based on a current possession in a possession-based sporting event at a particular time during the sporting event;

inputting the input tuple into an axial transformer neural network by inputting each tensor from the set of tensors within a corresponding initial embedding layer;

concatenating the initial embedding layers to form a single tensor;

applying self-attention to the single tensor through axial transformer layers of the axial transformer neural network;

mapping output embeddings from the axial transformer layers to target layers, each of the output embeddings being of a dimension of a target metric; and

generating a set of target metric predictions for each of a set of players, one or more teams, and a match, based on the output embeddings from the target layers.

16 . The non-transitory computer readable medium of claim 15 , wherein the super feature includes a playbook embedding that defines performance for certain types of players during a particular player in the possession-based sporting event.

17 . The non-transitory computer readable medium of claim 15 , wherein possession-based sporting events include football games, hockey games, and basketball games.

18 . The non-transitory computer readable medium of claim 15 , wherein the axial transformer neural network is configured to accept inputs with different modalities.

19 . The non-transitory computer readable medium of claim 15 , wherein the super feature is determined based on broadcast data.

20 . The non-transitory computer readable medium of claim 15 , wherein the applying self-attention includes applying an autoregressive attention mask to a row in each layer of the single tensor.