Outsample Research
We stress-test trading strategies with the tests most backtests never face, and we build indicators that pass the same gauntlet before they ship. No signals. No promises. Evidence.
01 · The problem
A strategy tuned to its own history will look brilliant on that history. The equity curve is smooth, the ratios are spectacular, and none of it survives contact with data it has not seen.
The industry calls this a track record. Statistics calls it overfitting. The difference between the two is exactly one test, and almost nobody runs it before selling you the result.
We run it. It is the whole company.
02 · The method
Every audit we deliver and every verdict we publish runs the same battery. It is designed to kill strategies. The ones that survive have earned it.
The strategy as submitted. Full history, honest costs, no tuning. The floor everything else is measured from.
50,000 reorderings of the trade sequence. Ruin probability, abandonment probability, terminal-loss probability.
One continuous run, sliced by market regime. A bear-year failure is a verdict, not a footnote.
If the settings are a lucky peak, the neighborhood collapses. Robust settings live on plateaus.
The strategy re-run on thousands of shuffled histories. If randomness explains the returns, we say so, with the p-value and the seed.
Every report states its limitations, every assumption is flagged, and Pine ports ship with an exact parity proof. The full methodology page publishes with our first public verdict. Honesty is not proprietary.
03 · Services
Send your strategy. Get a verdict: robust, fragile, or overfit, with every test shown.
3 to 5 business days. If we cannot test it honestly, we decline it before you pay.
Start an audit intakeYour idea, built properly in Pine or Python, delivered privately with honest limitations stated.
5 to 10 business days. One revision round included. Your code stays yours.
Start a build intake04 · The Verdict series
The same engine customers pay for, run in public on strategies traders actually use. First target, in production now: SuperTrend. The verdict will be whatever the evidence says. That is the point.
05 · Questions
No, and we never will. We sell evidence about strategies, and tools that were tested before they shipped. If someone promises you entries and exits with a win rate, that is the product we exist to check.
We will tell you what the evidence says. Verdicts come out robust, fragile, or overfit, and we publish the same battery against popular public strategies, so you can see we mean it.
Tier 1 needs a trade export from any platform. Tier 2 needs your Pine script, shared read-only. Deep work is scoped person to person. Every intake states exactly what we assumed and what we could not verify.
Because we show the work. Pine ports include a parity proof against the original, permutation runs disclose their seeds, and the methodology is public. You are not asked to trust us, you are shown the evidence.
Your code is read privately, never published, never reused, and deleted on request after delivery. The report is yours. We keep only what we need to stand behind the verdict.
The same as any other verdict. You are paying for the truth about a strategy before the market charges you more for the same lesson.