more of trading question than a wealth lab question:
Say I have a set basket of 50 ETF's that I backtest a strategy on and I find a set of parameters I like. When I observe how each ETF performs, lets just say 10 of them perform poorly and are dragging down the final result. If I decided to live trade only the best 40 out of the 50 ETF basket, would that be over-fitting?
Say I have a set basket of 50 ETF's that I backtest a strategy on and I find a set of parameters I like. When I observe how each ETF performs, lets just say 10 of them perform poorly and are dragging down the final result. If I decided to live trade only the best 40 out of the 50 ETF basket, would that be over-fitting?
Rename
QUOTE:
more of trading question
I feel this is more than welcome. I find too many "Bug Reports" in this forum. Would be interested to see much more "trading questions".
QUOTE:
would that be over-fitting?
Yes.
Every change to a strategy which is based on a backtest on a fixed Interval of backtest data will lead to a system which works on exactly that interval.
It will perform worse on future/unseen data.
One basic method to avoid (the worst examples of) overfitting works like this:
1. divide your backtest interval into two parts, labeled "Insample / IS" and "Out of Sample / OS" today I would suggest:
IS: 6 years 2012 - 2017
OS: 4 years 2018 - 2021
Leave 2022 out for final verification.
2. Make a change to your trading strategy and run it on the IS interval to see if it works. Repeat until you are happy.
3. See if the change from step 2 improves results for the OS interval. Reject your change if it does not.
4. repeat steps 2 and 3 until you run out of ideas.
5. Do a final check on the verification interval (2022 - 2023). If you are not happy with what you see, throw everything away (it will not make you any money anyway). Take a rest. Start over with step 1 if new ideas come up.
Your Response
Post
Edit Post
Login is required