FIELD GUIDE / 01
PRE-LIVE STRATEGY REVIEW
Twelve checks before real money touches the strategy.
A profitable backtest is a claim. These checks ask whether the claim survives missing trades, trading costs, statistical noise and a bad week.
Use the list as a gate. A “no” is not a reason to abandon the strategy; it is the next thing to measure before scaling it.
THE RECORD
Can the sample be trusted?
- Every closed trade is present.
No cherry-picked dates, symbols or “test” trades silently removed after seeing the result.
- The period was fixed before review.
Keep a sealed holdout period. Repeatedly inspecting the same sample turns research into curve-fitting.
- Timestamps and units reconcile.
Gross PnL, fees and funding must use one currency and satisfy gross − fees + funding = net.
THE ECONOMICS
Does the edge survive contact with the market?
- Fees are measured, not assumed away.
Use the actual maker/taker mix and round-trip cost for the venue and account tier.
- Funding and slippage are included.
A signal can be directionally right and still lose money after carrying and execution costs.
- Net edge is stated per turnover.
Compare strategies in basis points on traded notional, not only in dollars or headline return.
THE EVIDENCE
Is the result larger than the noise?
- The uncertainty interval clears zero.
A positive average is not established when its 95% interval still spans losses.
- Dependence is respected.
Use a block or stationary bootstrap when trades cluster in time; ordinary resampling can overstate confidence.
- Multiple experiments are counted.
If twenty variants were tried, the best result needs a higher bar than a single pre-registered test.
THE FAILURE MODE
What breaks first?
- One symbol cannot explain the result.
Remove the largest contributor and recalculate. A portfolio edge should not be one lucky market in disguise.
- Drawdown fits the real risk budget.
Size from the observed adverse path plus a safety margin—not from the return you hope to earn.
- Paper execution runs before live capital.
Verify fills, rejected orders, stale data, disconnects and kill switches under the exact production path.
THE MINIMUM EVIDENCE PACKAGE
Bring the record. VAILE does the arithmetic.
Export closed trades with PnL. Add fees, funding, notional, symbol and close time when available. The free diagnosis checks economics, sample quality and concentration without storing the raw CSV.