I've been exploring some of the big open-source collections of trading strategies, like FreqTrade, StockSharp, QuantTrading, and Alorse. What really impresses is how easily AI can now convert strategies from one language or platform to another. Personally, I prefer my own custom Python engine. But when it comes to actually finding “edge” or evaluating which strategies are worth trading, I still don’t really trust AI. For now, I’m leaning on a mass-testing approach: just batch-running lots of strategies and reviewing the results myself. Anyone else using this kind of bulk-testing workflow? Would be interested to hear about other people’s setups or tips.
The Holy Grail is timing, understanding and process. It's not some magical fixed, set it and forget it, market mathematical, formula that has yet to be discovered.
I used to "bulk test" strategies focussed on index futures...I got some that backtested well but they blew up. I find it better to start with an observation or hypothesis and figure out how to test it.
The only reason why I would "bulk test" strategies is to see correlation between them to make sure that combining them gives better result than each of them separately. The problem arises when one of them makes much more money than the others and then any equity swing on this strategy will dominate "combined" equity. I'm not quite sure yet how to deal with this, but I'm thinking something like using ln of equity before adding them up. It seems like it does the trick but more tests are needed.