A follow up post to my one earlier this morning (which I promptly closed when I saw starting_price as a field in historic_runners_beta)...
My historic model runs through 1000s of days of historic race data and provides model parameters going forward when choosing a future pick. A big issue for me is odds. Prior to my post this morning, I was using forecast odds for both historic and future picks and the reality was that the historic PnL was inflated because most of the starting prices seem to be lower than the forecast odds. Come race day, I was picking the same horses but the returns were lower (much lower in some cases).
I'm wondering if anyone has any ideas how I can circumvent this issue. Short of running my model minutes before each race with amended market odds, I'm not sure how this can be improved. Running the historic model with starting price seems to give accurate historic returns but I can't use them on 'today's races.
Thanks