POC
🎰 TOOL · Is my edge real?

Find a fake edge

This market is pure chance — a coin-flip with no edge hidden in it anywhere. Now play the game every strategy-hunter plays: keep trying strategies and keep the best one. Watch a convincing "winner" appear out of thin air… then watch it do nothing on the part of the market it never saw.

How it works
1 We build a market out of pure chanceRandom ups and downs with any lucky drift taken out, so there is genuinely no edge in it to find.
2 We try 60 strategies and keep the bestSixty random trading rules run on the first stretch of the market. We keep whichever one made the most — exactly what strategy-hunting does.
3 We run that winner on a stretch it never sawThe last part of the market was held back. Testing on data the search never touched is called out-of-sample — it is the check that exposes a fit to noise.
What the two big numbers are: both are the total return of that one winning strategy across a stretch of this market — not per year, and not a real market. The searched stretch is the first 60% of the data and the held-back stretch is the last 40%, so they are not the same length. The point is not which number is bigger; it is whether the profit survives at all once the strategy meets data it was not picked on.

What this can’t tell you

Worth reading before you take the lesson too far in either direction.

  • It does not prove that backtests are worthless. It shows what happens when you search a market that has no edge. Real markets may contain real edges; the point is that a good-looking result alone cannot tell you which case you are in.
  • One hunt proves nothing on its own. Sometimes luck carries into the held-back stretch too. The lesson lives in the pattern across many hunts, not in any single run — which is why the button is there.
  • Passing an out-of-sample test is not a guarantee. It is one check that a result was not simply fitted to noise. It cannot tell you the edge will continue, survive costs, or suit you.
  • Data stops being out-of-sample once you keep looking at it. Test twenty variations against your held-back stretch and you are mining that too — only slower. It is a check you get to spend roughly once.
  • The 60 strategies here are random rules, not real ideas. They stand in for the searching a person does. An idea with a genuine reason behind it starts from a better place than a blind search — but it still has to clear the same test.
  • The percentages are specific to this made-up market. Their size means nothing on its own; only the gap between the two, seen repeatedly, carries the lesson.
The thinking behind it
06
A pretty backtest proves almost nothing
READ →
Next

That was a market with no edge in it at all. Now put a real record through the same scrutiny — does yours beat luck once we allow for every variation you tried?

Real or Noise? →