Strategy validation that catches what backtesting misses

Your backtest passed.
Would it survive live trading?

Upload your strategy. Get a clear verdict in under five minutes, backed by seven independent statistical tests.

Failure modes

Four classes of failure between backtest and live

IN-SAMPLEOUT-OF-SAMPLE
L1Overfitting

Performance looks perfect in-sample but collapses on unseen data. The strategy captured noise, not signal.

r = 0.87NO CAUSAL LINK
L2Spurious correlations

Two series move together in-sample but share no causal link. The relationship is coincidence. It won't persist.

BULLBEARTRANSITION
L4Regime blindness

Optimized for bull markets, gives back all gains during a bear regime. When conditions shift, the model's assumptions break.

BACKTEST: +15.2%AFTER COSTS: +3.1%+15.2+3.1
L5Execution gap

+15.2% backtest returns erode to +3.1% after spreads, impact, slippage, and fees. Most backtests assume zero friction.

Validation engine

Seven layers, independent statistical methods, one composite score

L1Statistical Core82

Tests whether returns are real or mined from noise. Catches the overfitting that standard cross-validation misses.

CPCV + PBO + Sharpe Inference + Multiple Testing Correction
84
ValidatedComposite ScoreWeighted aggregation across seven layers with floor constraints
Try for Free →
7Independent layers

Each layer produces its own score from 0 to 100. They test different failure modes using different statistical methods. A strategy that passes one layer can fail another.

1Weighted composite

Layer scores are combined into a single composite. Weights reflect each layer's contribution to real-world predictive validity, calibrated against three decades of crisis data.

4Verdict categories

The composite maps to a clear label: Validated, Conditional, Weak, or Rejected. Each verdict carries a specific recommendation. No ambiguity.

Process

Upload, validate, receive a signed Passport

STEP 01

Upload your strategy

Pine Script, Python, MQL5, or API. All inputs converted to canonical format. Your logic is never stored.

STEP 02

Seven-layer validation

Statistical, causal, regime, and execution tests run independently. Results in under five minutes.

STEP 03

Get your Passport

Portable, cryptographically verifiable report with layer scores, composite verdict, and statistical evidence.

Comparison

What you're probably doing now, and where it falls short

Catches ~30% of overfitting

Manual walk-forward analysis

Walk-forward testing splits data into in-sample and out-of-sample windows. It only tests one failure mode. It won't detect regime vulnerability, spurious correlations, or execution cost erosion.

Catches nothing reliably

Sharpe ratio thresholds

A Sharpe ratio above 2.0 doesn't mean the strategy works. Harvey and Liu showed that Sharpe ratios above 2.5 almost always reflect data mining, not genuine alpha.

Tests historical fit, not predictive validity

Commercial backtesting platforms

QuantConnect, Zipline, and similar platforms are good at running backtests. They're not designed to validate whether those backtests represent real edges.

Confidentiality

Zero data retention, TEE enclaves, cryptographic audit trail

Zero data retention

Your strategy logic is never stored, logged, or shared. Inputs are discarded after validation completes.

TEE confidential validation

Your strategy runs in a secure enclave. No one can access it, including Sigmentic.

Cryptographic audit trail

Every Passport is cryptographically signed. Investors can verify without seeing your strategy.

The Validation Passport is the deliverable: a portable certificate of your strategy's validation status. Share it with allocators or compliance teams without revealing your logic.

Get started

Validate your strategy before you risk capital

3 free validations. No credit card required.

Under 5 min to verdictZero data retentionPeer-reviewed methodology