I'm working on my first strategy, and as part of it I'd like to model slippage with ML. My plan to do this is to execute a bunch of trades with a paper trading account, and use the data from that to train a model that predicts slippage. While there are definitely cases where the price it fills at is different than the last bid/ask, it seems to be the exception rather than the rule. From reading posts here I've gotten the sense that slippage can be a death sentence to an otherwise profitable model, so I'm a bit worried that maybe slippage is worse when live trading. As a note, I'm executing these trades as market orders.
Any thoughts one way or another?
Submitted November 12, 2020 at 01:46PM by EdvardDashD