Hi, i only feel comfy running strategies that have proven themselves in “real time” or however u wanna put it.
Lets say i make a strategy that on avg takes 1 trade pr day, i wanna see some good and some bad trades in demo just to make sure its actually doing what i want them to do. So theres no funky stop loss going crazy or sellsignal has a wrong > < or whatever… u just never know.
When u have a 10 year backtest that looks good, and a systematic approach to making the system so its not curvefit (hopefully) and u see 10~ good and 10~ bad trades and they all behave like they should, i would run live.
But it depends on how many trades per day/week it should take. If its down to like 0.3 i might not wait for 10 good and 10 bad trades, i might just wait a month or 2 and if it all looks good then, its go time.
I like to think of it like this:
- I make my strategy in a way tries to avoid too much curvefitting
- I place strategy in demo to see that its not 100% curvefit and that there is no code-issues/bugs
- Lets say my algo is Long only: If market is up, and algo is up then thats good. If market is up but algo is down, then thats a red flag that u might have overfit it. Or maybe market was just extra volatile or extra whatever.. if the market is up, and ur system is down, then it MIGHT not be a bad system, just bad luck in timing. Then again it MIGHT mean ur system is overfit and shit.
I wanna be confident in my system so when i put it in demo its just basicly a final check/test to see that everything that i expect to happen, does indeed happen, and so theres no funky surprises that might destroy my account. Its all about confidence in urself and ur systems
.
