tl;dr: How hard should you look for the best settings when backtesting indicators?
When I figured out how to backtest indicators automatically with scripts I thought it would put my system building on warp speed! “an INSTANT backtest, on any indicator, any setting, any timeframe, any sl/tp rules? sign me the f up!” I thought.
The problem is now I’m spending as much or MORE time on each indicator testing each possible setting to see which shows the best winning percentage and net profit, especially on indi’s that have, like, three Moving Averages with six types you can choose from.
Am I doing it wrong?
On one hand, many indicators have an exact setting or approx setting that seems to work a lot better than the default setting, at least with my the trading style I’m trying to test for. I figure it’s worth seeking those out, right?
On the other hand settings that work best on some pairs are not nearly the best on other pairs. And we can assume that settings that work the best in the past may not work the best in the near future, even if you test a large sample size.
Does anyone else use scripts for backtesting? How do you ascertain which indicators and settings test the best? Perhaps more importantly, how do you log them?
At first I was just writing down the setting that gives the indi the best winning %, but some of the highest only gave a few signals per year. I started going by which settings shows the highest net profit instead, but that seems faulty too because right now I’m only testing individual indicators and combing them with others will change those results.
What’s more, the net profit numbers seems to bounce around wildly just by a small change in the settings. Can an indicator really return that much more profit/loss if one of it’s moving averages is two candles faster, even when the win percentage is about the same?
Regards Greg
MQL5 website have big codebase, check it