[Python] I built an open-source trading bot with a walk-forward optimizer… lessons learned

Been working on a momentum breakout bot for Coinbase for the past few months and finally got around to adding a walk-forward optimizer. Spent way too long on it for the results I got.

The idea was simple: every Sunday (in simulation), retune the strategy params using 4 weeks of training data and 1 week of validation. 108 combinations across breakout period, RSI thresholds and volume multiplier.

Static params: +46.4% in 2024. Adaptive: +45.0%. Profit factor identical at 2.11.

I genuinely don't know if this means the strategy is robust or if my parameter grid is just too narrow to make a difference. Probably both.

Still not fully happy with the live setup. The bot is scanning 8 assets on Coinbase and I'm not 100% sure the position sizing math holds up with a small account. Haven't seen a real entry yet either… either the market conditions haven't been right or there's something else I'm missing.

Code is here if anyone wants to poke holes in it: https://github.com/pecintra/crypto-bot

Genuinely curious if anyone's found walk-forward optimization useful for momentum strategies, or if it's more of a mean-reversion thing.