Query on how best to perform model comparisons #6
Description
First of all - thank you so much for your excellent work with TTT!
I was wondering if you could provide a few hints about how you performed the model comparison in your paper. In particular how would one go about repeating this: "To predict the outcome of one observation at time t, we use all the data up to the day preceding t (in both training and test sets)."
I understand why it is important to avoid using future information that was unavailable at the time of the game being played. But I am struggling to think of a good way of repeating what you did without creating an updated History class and rerunning convergence prior to every time t. Given the test and training datasets are tens of thousands of games long this seems computationally prohibitive. Is there a clever way of doing this that I am missing?