4 ms·
Built a SAAS app that automatically generates trading strategies from price/volume data. The strategies were optimized to achieve high Sharpe. However, it turns
by cg94301 6y ago
Built a SAAS app that automatically generates trading strategies from price/volume data. The strategies were optimized to achieve high Sharpe. However, it turns out that high Sharpe has little value in predicting out of sample performance. Stopped development short before trying to monetize. The app is still up at turboquant.com
- maest 6y agoA historical good Sharpe is usually a prerequisite, but not a guarantee of having good live performance. I'm sure you're aware of overfitting backtestes and it sounds like that's what you were doing.
- cg94301 6y agoYou can find patterns without necessarily overfitting to them. It's more a question of correlation vs causation. Looking for patterns reported in research papers would be more promising. But cannot automate that.