Your backtest passed.
Will your strategy survive live?
Upload your backtest. Get a 7-layer validation in under 5 minutes. Find out if your edge is real before you risk capital.
The problem
Why backtests lie
You spent weeks optimizing. The backtest showed 40% annual returns. You deployed it. Month one: down 8%.
That's not bad luck. It's overfitting. The strategy captured noise in the training data and mistook it for signal. A 2.5 Sharpe in backtesting becomes 0.4 once you account for spreads, slippage, market impact, and regime shifts.
Walk-forward analysis catches about 30% of overfit strategies. The other 70% look fine until they bleed money in live trading. The gap between paper performance and live performance is where real capital disappears.
Process
How it works
Upload, validate, decide. The full process takes under six minutes.
Upload your strategy
Pine Script, Python, MQL5, or CSV backtest results. Drag and drop. The system auto-detects your format and column structure. No configuration, no API keys.
Seven layers validate
Each layer runs independently. Overfitting detection finishes first. Then causal inference, stress scenarios, regime analysis, execution costs, and integrity checks. You watch results arrive as each layer completes.
Get a clear verdict
Validated, Conditional, Weak, or Rejected. No ambiguity. The composite score (0 to 100) and per-layer breakdown tell you exactly where your strategy is strong and where it's vulnerable. Deploy with data, not hope.
Detection layers
What we catch
Four of seven layers, focused on the failure modes that cost individual traders the most.
Statistical Core: Is the edge real?
Cross-validates your strategy across multiple time periods and parameter variations. If performance only exists in one specific slice of history, this layer flags it.
CPCV (Combinatorially Purged Cross-Validation) tests every possible combination of training and testing periods while removing contaminated data points. It catches over 80% of overfitting that standard walk-forward analysis misses.
Scenario Lab: What hasn't happened yet?
Your backtest only covers historical conditions. This layer generates thousands of synthetic scenarios your strategy has never seen: fat-tail events, liquidity shocks, correlation breakdowns.
The chart shows the core problem. A strategy that looks great in backtesting (the ascending curve) can follow two very different paths after deployment. Validated strategies continue to compound. Overfit strategies give back everything. This layer reveals which path yours is more likely to follow.
Regime Intelligence: Will it survive the next shift?
Detects which market regime your strategy was optimized for and tests whether it degrades when conditions change. A strategy built for 2017's low-volatility calm can fall apart when the VIX spikes.
The framework identifies three distinct regime types using Hidden Markov Models on market data. Your strategy is tested against each regime independently. If it only works in one, you'll know before the market shifts.
Execution Realism: What are the real costs?
Models spreads, market impact, slippage, and fees for your specific strategy. The backtest assumes zero friction. This layer shows what +15.2% annual returns look like after the market takes its cut. Often: +3.1%. Sometimes: negative.
These are 4 of 7 layers. The full platform includes causal inference, live monitoring, and cryptographic integrity. See all seven layers
Crisis validation
Proven against real crises
We ran 61 strategies through the pipeline against three decades of market crises. The framework correctly separated genuine strategies from overfit ones, and its rankings predicted which strategies would actually perform in the years after each crisis. Read the full case studies
Comparison
Compare your options
Sigmentic | Manual Review | No Validation | |
|---|---|---|---|
| Time to result | Under 5 minutes | 40+ hours per strategy | N/A |
| Overfitting detection | CPCV + PBO (catches 80%+) | Walk-forward only (~30%) | None |
| Regime testing | HMM regime detection + stress | Rarely done | None |
| Execution costs | Full cost model | Estimated or ignored | Assumed zero |
| Output | Portable Validation Passport | Internal notes | Nothing |
Get started
Start validating. Free.
3 validations per month. No credit card required.