Winners that still closed red
POSITION LIFECYCLE81% of closed positions were up at least +100bp at some point; half of those still closed at a loss.
From a real forensic: 17 reports on my own live system — $80.46 lost in six weeks, and $4 of it was execution.
Most backtests don’t fail live because the idea was bad. They fail because of what the backtest couldn’t show: costs modeled at a fraction of the real ones, a lookahead leak, a universe that quietly excludes the delisted losers, or a result that only exists because it was the best of 400 tries. I know, because I ran the full forensic on my own live system after it lost money for six straight weeks.
Before this audit was a product, it was an autopsy. I ran the full forensic on my own live system after six straight weeks of losses — 17 reports, every fill reconciled against the exchange. The verdict surprised me: exits were innocent, execution cost about $4 of an $80.46 loss. This is what the method found.
81% of closed positions were up at least +100bp at some point; half of those still closed at a loss.
Across 236 exit-rule variants replayed over 5 years, only one nominally beat doing nothing; the median destroyed 63% of terminal wealth.
Execution cost ≈ $4 total — acquitted. The usual suspect didn’t do it.
The killers: 3× long exposure into a −22% alt month, and holds over 10 days (−$73.95).
Every fill reconciled against the exchange; residual 0.37 USDT on the whole window.
Exits were innocent and execution was ≈$4 of an $80.46 loss — the killers were 3× long exposure in a falling market and symbol picks that scored worse than random (z −3.2). That audit is the template for yours.
A written forensic report on your backtest — the same six checks, in the same order, that produced the findings above.
Do your trades actually reproduce your claimed equity curve, fill by fill?
Your cost model vs what your instruments really charge; what 2× and 5× costs do to the result.
Signal timing, bar leakage, corporate-action and data-age traps.
Who is missing from your test, and whether they were the losers.
Trial count, selection effects, deflated expectations; is the edge distinguishable from luck?
With the fixes ranked by measured effect, not opinion.
Any platform — if it can export a trade list, it can be audited.
Your trade list (CSV or platform export), your equity curve, a short strategy description, and the cost assumptions you used.
Your report lands within 7 days of a complete submission — all six checks, with the evidence behind every finding.
One written clarification round is included — ask me anything about the findings.
The first 3 audits are priced to build the public track record. In exchange, I ask (a) a short testimonial if the audit earns it, and (b) your written permission to publish an anonymized teardown.
Both are your choice; the audit is yours either way. After the founding cohort, the price moves to $249–$299.
Backtest Honesty Checklist buyers: your $14 is credited toward the price.
The same honesty method at three depths: check it yourself for free, run the full framework yourself, or have the forensic run for you.
A one-page starter for checking four common backtesting problems — the fastest way to see whether your backtest is lying to you.
The method, in your hands: a structured way to document your assumptions and find reasons to stop or investigate further — before committing real capital.
A forensic audit of your backtest, done for you. The checklist is the method in your hands; the Autopsy is me running the full forensic on your numbers.
Stated plainly, before you pay anything.
A method and cost audit of your backtest.
Not financial advice. Not a strategy redesign — I audit the test, I don’t rewrite your system. No return predictions, ever.
Your data stays private. Nothing about your strategy is published without your written permission.
Start with the free one-page starter, or put your backtest under the full forensic while the founding cohort is open.