EA #4 – XAUUSD M15 | 20-Month Backtest Results & Strategy Discussion
I’m sharing the latest backtest results of EA #4, tested on XAUUSD M15 from 1 January 2025 to August 2026.
Key results:
Initial Deposit: $1,000Net Profit: $767.66Return: 76.77%Profit Factor: 1.56Total Trades: 103Max Equity Drawdown: 18.14%Sharpe Ratio: 5.29The objective is not simply to show a profitable backtest, but to understand the strength and weaknesses of the strategy.
I’m currently analyzing the losing trades and the periods of consecutive losses to identify whether there are specific market conditions where the EA loses its edge.
I’d appreciate feedback from experienced traders on:
Entry qualityDrawdown and consecutive lossesMarket-regime filteringRobustness of the strategyWhat you would investigate before moving toward live testingBacktest results are historical and do not guarantee future performance.
Interesting results. First thing I would check is how the 5.29 Sharpe was calculated, because it looks quite high next to PF 1.56 and 18% equity drawdown.
With only 103 trades I would not add more filters yet, as it is easy to overfit this sample. Better freeze the settings, test an untouched out-of-sample period, increase spread and slippage, and vary the main parameters by around 10–20%.
I would also split the results by year, trading session and long/short direction. A Monte Carlo shuffle of the trade sequence can give a more realistic idea of possible drawdown. If the profit stays stable across these checks, a demo forward test would be the logical next step.
fxrobustix posted:Interesting results. First thing I would check is how the 5.29 Sharpe was calculated, because it looks quite high next to PF 1.56 and 18% equity drawdown.
With only 103 trades I would not add more filters yet, as it is easy to overfit this sample. Better freeze the settings, test an untouched out-of-sample period, increase spread and slippage, and vary the main parameters by around 10–20%.
I would also split the results by year, trading session and long/short direction. A Monte Carlo shuffle of the trade sequence can give a more realistic idea of possible drawdown. If the profit stays stable across these checks, a demo forward test would be the logical next step.
Thanks for the detailed feedback. I agree — with 103 trades, adding more filters at this stage could easily lead to overfitting.
I’m currently analysing the losing periods and market conditions first. The next step will be an untouched out-of-sample test, followed by robustness checks before making any further changes.
Appreciate the input.
