Backtest Depth - Long-term Robustness (10 years) vs. Recent Adaptation (5 years)? - page 6
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
This topic became something different, is thera a possibility to change the name?
If you don't get the gist of the conversation @Enrique Dangeroux and I had, well, you'll probably be stuck here for a few more years. Someday it will all make sense to you.
If you don't get the gist of the conversation @Enrique Dangeroux and I had, well, you'll probably be stuck here for a few more years. Someday it will all make sense to you.
I mean, as I saw your conversation was about reducing the testing time with Enrique's method.
Let me know if I missed something, pls!
Didn't mean to be rude, my bad.
It's fine, don't worry, I'm not offended, I'm just pointing out that...
Ok!
I'll read it deeply. :)
If I have questions, could I ask?
Drawdown Control
Fixed Risk
Survival First
Ok!
I'll read it deeply. :)
If I have questions, could I ask?
I guess so, that's what forums are for. But as for the technique used, I don't know the first thing about it yet, I haven't had time to check it out; you'll have better luck with Enrique.
Drawdown Control
Fixed Risk
Survival First
A practical way to compare 10y vs 5y is to treat the older window as a stress test and the recent window as the live candidate. If performance only survives when spread, commission, and slippage are held unrealistically low, the edge is probably fragile.
I’d run walk-forward on the last 2–3 folds, then compare trade distribution, drawdown shape, and average adverse excursion between the two windows. If the recent 5y behaves materially differently, that usually points to regime dependence rather than better robustness.
Do you see the same direction of drift across all symbols/timeframes, or only around certain session hours?
For me, the question is not simply 10 years or 5 years. I use the two periods for different purposes. The older data is a robustness stress test, while the recent period shows whether the edge is still relevant to current market behaviour.
For XAUUSD, I believe data and execution validation can be as important as the test duration. In my current work, I am comparing individual trades between MT5 and Dukascopy/JForex data after normalising the timestamps to GMT. I check direction, entry and exit times, prices, spread, commission, swap, and any missing or additional trades. I also compare how the same strategy behaves under different broker feeds and slippage assumptions.
After that, I prefer dividing the history into distinct volatility regimes and testing the same rules and parameters without re-optimising each period. If the strategy works only with one data provider, during one market regime, or with unrealistically low trading costs, I would consider the edge fragile.
Therefore, my preference is: use the 10-year test for survival and regime robustness, then use the recent years as out-of-sample or walk-forward confirmation. I would select a stable parameter region rather than the single highest-profit result.