Backtest Evidence Review

Not sure it fits? Ask before buying

Secure checkout by Stripe. After verified payment, I email the intake and secure upload instructions.

• Builders of systematic or rule-based strategies who already have trades/results exported.
• People who saw green backtests but feel uneasy about costs, OOS, or “too good” metrics.
• Creators sharing backtests who want a second pair of eyes on methodology.

Not for: personalized investment advice, asset picks, parameter tuning, or “will this make money?” guarantees.

• Look-ahead / leakage and train-test contamination (risk screening, not certification)
• Unrealistic fees, slippage, or fill assumptions
• Weak IS/OOS separation or walk-forward design
• Retrospective parameter selection (implicit p-hacking)
• Multiple testing without correction
• Missing or weak null / permutation tests
• Low robustness across periods
• Gap between backtest narrative and plausible live execution

I do not re-run your full backtest or guarantee absence of leakage.

• Methodology scorecard
• Five primary evidence risks (ranked)
• One recommended next experiment (scope only)
• Evidence verdict: insufficient · exploratory · candidate for prospective validation
• One round of clarifications (async)

Full scope, limits, and data policy →

• Financial advice, signals, or “buy/sell/deposit” guidance
• Strategy fixes, rewrites, or parameter optimization
• Running untrusted code, API keys, or exchange access
• Legal, tax, or regulatory sign-off
• Full backtest reproduction or exhaustive error search

Xavier Gisbert Blanquer — independent reviewer

I built and pre-registered six hypothesis tests (H001–H006) in a personal crypto-strategy research lab, using out-of-sample, null and minimum-sample gates. None passed the gates for live capital.

Public methodology sample: H001 evidence review

I pre-register hypotheses in a personal research lab (H001–H006). Under pre-registered gates, none passed for live capital — including H001, which showed positive OOS PnL on paper but failed null and sample gates. The public sample summarizes H001 only.

Abbreviated summary of pre-registered experiment H001 that failed evidence gates despite positive OOS PnL on paper.

Secure checkout by Stripe. After verified payment, I email the intake and secure upload instructions.