Skip to main content
A backtest is not a research finding — it’s one path through one history. Gecko grades every strategy with selection-bias-aware rigor before it earns a deploy.

​
The verdict block

STRATEGY: trend_breakout
  CPCV median Sharpe:      <value>  (5–95%: [<low>, <high>])
  CPCV % paths Sharpe < 0: <value>
  Deflated Sharpe Ratio:   <value>  (DEPLOY needs ≥ 0.95)
  PBO:                     <value>  (DEPLOY needs < 0.20)
VERDICT: DEPLOY | PAPER ONLY | REJECT

​
Two gates, never confused

  • PAPER-CONTINUE (keep running on paper): DSR > 0, PBO < 0.5, fee-adjusted return CI excludes 0 on ≥3 of 5 symbols.
  • DEPLOY (real money): the stricter bar — DSR ≥ 0.95, PBO < 0.20, %CPCV-paths-Sharpe<0 < 25%.
A §5 paper pass is never a real-money green light. And a null — no edge after fees — is a cheap, valid, honest outcome. Most candidates reject. That’s the point.

​
Run it

curl -s localhost:8271/backtest -X POST -H 'Content-Type: application/json' \
  -d '{"strategy_id":"trend_breakout"}'