I've been working on Quantopian for around half a year now and still have yet to find a strategy that I would actually be comfortable putting money behind. Whether it be an idea that I developed on my own, or a seemingly solid algorithm that I've cloned from another, it appears that nearly everything flops in an out of sample period, or just flat out will not work in a live market environment (perhaps due to liquidity, slippage, commisions, etc.). I am wondering if someone that has been at this longer than myself could outline their workflow for developing new ideas and share some standard practices for avoiding over-optimization.
Thanks,
Maddox