Systems and methods for model evaluation using prior data
Systems and methods for model evaluation using prior data are disclosed. A system can store a set of inputs provided to a first model to generate first data points for a first live event. Each input of the set of inputs can include a respective state of the first live event. The system can initiate an execution environment for a second model configured to generate second data points for the first live event. The system can execute the second model within the execution environment using the set of inputs to generate candidate data points for the first live event. The system can generate a score based on the first data points, the candidate data points, and one or more corresponding outcomes of the first live event. The system can set a flag to replace the first model with the second model responsive to the score satisfying a threshold.
1. A system, comprising:
one or more processors coupled to non-transitory memory, the one or more processors configured to:
store a set of inputs provided to a first model to generate first data points for a first live event, each input of the set of inputs comprising a respective state of the first live event;
initiate an execution environment for a second model configured to generate second data points for the first live event;
execute the second model within the execution environment using the set of inputs to generate candidate data points for the first live event;
generate a score based on the first data points, the candidate data points, and one or more corresponding outcomes of the first live event; and
set a flag to replace the first model with the second model responsive to the score satisfying a threshold.
2. The system of claim 1 , wherein the one or more processors are further configured to:
detect a change in the first live event during the first live event; and
generate a first input of the set of inputs responsive to the change.
3. The system of claim 1 , wherein the one or more processors are further configured to:
store the set of inputs in a time-series data structure according to an order that the set of inputs were provided as input to the first model.
4. The system of claim 1 , wherein the one or more processors are further configured to:
intercept the candidate predictions generated by the second model via the execution environment.
5. The system of claim 1 , wherein the one or more processors are further configured to:
provide each input of the set of inputs to the second model in a same order as the set of inputs were previously provided to the first model.
6. The system of claim 1 , wherein the one or more processors are further configured to:
determine that third predictions are to be generated for a second live event; and
retrieve a second set of inputs provided to the first model to generate second predictions for the second live event.
7. The system of claim 6 , wherein the one or more processors are further configured to:
execute the second model within the execution environment using the second set of inputs to generate second candidate predictions for the second live event; and
flag the first model for replacement with the second model further based on the second candidate predictions and one or more corresponding second outcomes of the second live event.
8. The system of claim 1 , wherein the one or more processors are further configured to:
generate a report based on the first predictions for a first live event generated via the first model and the candidate predictions for the first live event generated via the second model.
9. The system of claim 1 , wherein the set of inputs comprises a state of the first live event corresponding to a respective timestamp.
10. The system of claim 1 , wherein the first predictions for the first live event comprise predictions for one or more conditional events to occur during the first live event.
11. A method, comprising:
storing, by one or more processors coupled to non-transitory memory, a set of inputs provided to a first model to generate first data points for a first live event, each input of the set of inputs comprising a respective state of the first live event;
initiating, by one or more processors coupled to non-transitory memory, an execution environment for a second model configured to generate second data points for the first live event;
executing, by one or more processors coupled to non-transitory memory, the second model within the execution environment using the set of inputs to generate candidate data points for the first live event;
generating, by one or more processors coupled to non-transitory memory, a score based on the first data points, the candidate data points, and one or more corresponding outcomes of the first live event; and
setting, by one or more processors coupled to non-transitory memory, a flag to replace the first model with the second model responsive to the score satisfying a threshold.
12. The method of claim 11 , further comprising:
detecting, by the one or more processors, a change in the first live event during the first live event; and
generating, by the one or more processors, a first input of the set of inputs responsive to the change.
13. The method of claim 11 , further comprising:
storing, by the one or more processors, the set of inputs in a time-series data structure according to an order that the set of inputs were provided as input to the first model.
14. The method of claim 11 , further comprising:
intercepting, by the one or more processors, the candidate predictions generated by the second model via the execution environment.
15. The method of claim 11 , further comprising:
providing, by the one or more processors, each input of the set of inputs to the second model in a same order as the set of inputs were previously provided to the first model.
16. The method of claim 11 , further comprising:
determining, by the one or more processors, that third predictions are to be generated for a second live event; and
retrieving, by the one or more processors, a second set of inputs provided to the first model to generate second predictions for the second live event.
17. The method of claim 11 , further comprising:
executing, by the one or more processors, the second model within the execution environment using the second set of inputs to generate second candidate predictions for the second live event; and
flagging, by the one or more processors, the first model for replacement with the second model further based on the second candidate predictions and one or more corresponding second outcomes of the second live event.
18. The method of claim 11 , further comprising:
generating, by the one or more processors, a report based on the first predictions for a first live event generated via the first model and the candidate predictions for the first live event generated via the second model.
19. The method of claim 11 , wherein the set of inputs comprises a state of the first live event corresponding to a respective timestamp.
20. The method of claim 11 , wherein the first predictions for the first live event comprise predictions for one or more conditional events to occur during the first live event.