Every tick based on real ticks vs 1-minute OHLC: What is your backtesting workflow?
Do you optimize on 1-minute OHLC first and only validate the best parameter sets using Real Ticks?
For swing trading EA's, I initially work with OHLC data because that tends to expedite experimentation with tweaking, adding, and/or removing specific functions/features early on. Once the majority of the design dust has settled, I switch over to real ticks for the final run. For a scalping EA based on say, intra-bar VSA logic, OHLC data is largely irrelevant so I skip it.
Do you rely on built-in Forward Testing or custom criteria ( OnTester ) to avoid overfitting?
I actually avoid overfitting by avoiding "selective" optimization. I really prefer to run fresh single backtests instead. Some users have reported differences between optimization results and single run results. If and when I optimize anything, I optimize based on all quality historic data that I have. Usually, I "optimize" a custom chart structure, itself, using MT5 and Python. My last such case involved 26 years of historic M1 data.
How long do you typically run a live demo forward test before trusting the backtest results?
As long as I have solid stats regarding the average/consistency of my broker-dealer's slippage, spreads, commissions, swaps, and liquidity, I don't really worry about forward testing. A demo account doesn't precisely mimic a live account, so there's no point in that. I only use demo accounts for testing the basic functionality/operability of new code.
Hi everyone,
I'm curious about how most developers and algo traders here structure their testing workflow in the MT5 Strategy Tester.
When optimizing an EA—especially on volatile symbols like XAUUSD—running full optimization with "Every tick based on real ticks" can be very slow. On the other hand, "1-minute OHLC" is much faster, but it can miss intrabar spread spikes and exact execution behavior.
What does your usual workflow look like before moving an EA to a demo or live account?
Do you optimize on 1-minute OHLC first and only validate the best parameter sets using Real Ticks?
Do you rely on built-in Forward Testing or custom criteria ( OnTester ) to avoid overfitting?
How long do you typically run a live demo forward test before trusting the backtest results?
Would appreciate hearing how others balance testing speed and accuracy. Thanks!