Prediction Hedging to Improve TC Volatility

I am curious if anyone has given this thought, but if TC is the gradient of the returns with respect to my stake, is there a way to adjust my predictions to reduce the noise in that gradient? For example, for assets with volatile future returns, predict those at 0.5 but for stable positive returns predict with 1.0 and stable negative returns predict with 0.0

Your predictions are still uniformly rank normalized for TC – never the raw values – so any adjustments you make come down to reshuffling the order of your predictions. Ensembling predictions that get uncorrelated TC may be the way to go, but it is hard to get a handle on empirically when we can’t test directly and have limited data to make guesses.

1 Like