MT5 optimization · overfitting audit
Your backtest looked perfect. Then it died live.
MT5 sorts thousands of configurations by profit and hands you a winner. Orometra finds the one that survives when the market shifts — the stable plateau, not the lucky peak.
Your files never leave your browser. No accounts. No upload to a server.
Includes a synthetic demo with a planted plateau — open See a full analysis to inspect a complete audit without uploading files.
From MT5 chaos to inspectable evidence
Isolated peak
Hover — peaks collapse into the plateau
Stable plateau
MODERATE EVIDENCE
#1 rejected
Plateau center · maximin · not the profit peak
Illustrative example. Your audit explains why a configuration was selected or rejected — not just a score.
What you get
-
Stable regions, not luck
Finds connected neighborhoods where neighboring parameters also work, shown as a real 3D surface — not this page's illustration. An isolated spike is rejected, even if it topped your MT5 table.
-
Top 3 with reasons
Three concrete configurations, neighbor support, out-of-sample behaviour, and the risks that still apply.
-
Evidence strength
Strong, moderate, weak, or insufficient. Orometra measures what your data can support — it does not promise future profits.
-
Robustness under real costs
Drop your backtest report on the unseen period and get a Monte Carlo over your actual trades, a sample-size check, and a cost-stress table — no install, straight from the trade list MT5 already gives you.
Three mistakes it prevents
-
Trusting the #1 row
The top of an MT5 table is often an isolated peak. Neighbors and out-of-sample behaviour stay invisible until you audit the surface.
-
Averaging away risk
Orometra scores by the worse period and by maximin neighborhoods — not by an average that hides a fragile edge.
-
Confusing evidence with a GO
A strong grade means the data can support a stable region. It is not permission to trade or a forecast of future profit.
How it works
Optimize. Export. Audit.
- Optimize with Forward on. Split in-sample and out-of-sample in the Strategy Tester before you run the optimization, not after — it's the only way the forward period is genuinely one you never used to choose. Export both reports (XML).
- Drop the files in the app. In-sample and forward are detected by content.
- Read the verdict, plateaus, and rejected peaks. Export a .set to refine in MT5, or drop a single backtest report to see Monte Carlo and cost stress on your actual trades.
How the engine decides — gates, plateaus, maximin — is documented in the methodology.
What Orometra does not do
- It does not predict the market or guarantee returns.
- It does not pick the highest-profit pass by default.
- It does not turn a weak EA into a strong one.
Ready to audit an optimization?
Open the app, load your MT5 results, and see whether the best rows deserve trust.
Audit my optimization