Robustness verdicts for trading strategies
Snapshots flatter.
Full history decides.
A strategy's best four weeks can show an 81% win rate. The same rules can lose money for six years. Edge Court runs your strategy on years of real market data, puts the best snapshot next to the full record, and tests whether the profit survives time, market regimes, parameter changes, resampling and costs. Every verdict is Ed25519-signed.
What every audit checks
| Test | Question it answers |
|---|---|
| Snapshot vs full history | How good did the best window look, and what share of windows looked like a monster? |
| Per-year and 5-fold stability | Was the profit earned across time or in one stretch? |
| Probabilistic & Deflated Sharpe, MinTRL | Is the average trade distinguishable from zero after skew, fat tails and the number of variants tried? |
| Market regimes | Does it make money in bull, bear and range markets, and in low and high volatility? |
| Parameter sweep (Pine) | Do nearby settings (±10–20%) still make money, or is this an isolated peak? |
| Walk-forward (Pine) | Would settings picked on the past have made money on the data that came next? |
| Block bootstrap | How likely is a losing outcome given the same trades in a different order? |
| Cost stress | How much extra cost per trade wipes the edge out? |
Pricing
| Audit | Web | API / MCP (x402, USDC on Base) |
|---|---|---|
| Pine strategy audit | 3 free per day | $0.25 per audit |
| Trade-list audit | 15 free per day | $0.05 per audit |
| Preview, receipt verification, methodology | free | free |
Payment settles only after the signed report is produced. A strategy that fails to run is never charged. API & MCP docs
FAQ
Why can a strategy show an 80% win rate and still lose money?
Win rate ignores the size of losses and the period tested. A short window can show a high win rate and profit factor while the same rules lose over several years.
What data does Edge Court use?
Real Binance spot candles (closed bars only, SHA-256 hashed into every receipt), or price history you upload. Trade-list audits use your exported trades exactly as exported.
Is this financial advice?
No. A verdict describes how a set of rules behaved on past data. It does not predict future results.