Quantopian's community platform is shutting down. Please read this post for more information and download your code.
Back to Community
Alphalens tearsheet great but algorithm backtest bad with same exact strategy?

My Alphalens tear sheet performance looks really good. But when I take the same strategy and put it into the algorithm, the backtest performance is less than the risk-free rate. The factor is based on factset consensus EPS quarterly released data for next quarter. The algorithm re-balances weekly. Dunno if different periodicity of data releases versus the algorithm rebalance timing is the reason? Trying to figure this out...

1 response

I think I found the solution. My alphalens mean_information_coefficient showed the IC decay at its maximum 125 days into the future indicating the predictive strength of my factor was highest 125 days after the underlying data released. My algorithm was initially set to rebalance weekly which is just 5 days. I had to change my algorithm to rebalance semi-annually in order to match what my IC decay was telling me about the predictive timing of my factor. Now my algorithm backtest performance results line up with what I expected.