Robustness lab for MetaTrader 5
Is your edge real, or did the optimizer just get lucky?
Every optimizer finds something. This is the landscape it searched — every setting a point, height is profit. Scroll to fly down to the one it reported.
The pick
The optimizer always finds something.
MetaTrader tested 110 settings and reported the best: +15,148. The question is whether that number belongs to the strategy — or to the luck of one history.
The neighbourhood
Its neighbours tell on it.
Move one step in either input and the profit falls to 1% of the peak. A real edge is a region, not a point — live fills never land exactly on the cell you optimized.
The forward period
What the backtest implied — and what happened.
Resampling the trades 2,000 times gives the range the forward period should have landed in. It landed below it: +972 against an implied +4,249.
The checks
Eight ways a backtest can lie.
Each check asks a different question of the same data; no single one carries the verdict. Weights are calibrated against the study, none above 20%.
- Monte Carlo16%
- Stability14%
- Deflated Sharpe14%
- Concentration14%
- Rank consistency14%
- Walk-forward12%
- Permutation8%
- Sample8%
The evidence
We tested the score first.
5,000 simulated optimizations, scored exactly as the app scores yours, then traded forward on fresh data. Rated ROBUST had a genuine edge 79% of the time; FRAGILE, 10%. Synthetic worlds are not real markets — the full study, including where the score is weakest, ships inside the app.
The plateau
Trade the plateau, not the peak.
The robust demo's best setting sits where every neighbour also profits. It scores 86 and kept 65% of its in-sample pace through the forward period. Plateau reads the exports MetaTrader 5 already makes — no code needed.
Private beta
Stop trusting backtests that flatter you.
Plateau runs in your browser; your reports, trades and parameters are never uploaded. The only thing that leaves your machine is the email you type here.