THE AUDIT · MARKET BY MARKET
Check your own market.
Then check the error bar.
You came for one market, not an average. So here is every one of them, for both rules we pre-registered and published as failures. What you will find is that almost nothing here separates itself from zero · and that the few results which do are all on the losing side.
READ THIS FIRST · IT IS THE WHOLE PAGE
Twenty-five trades is not enough to know anything.
Flip a fair coin 25 times. You will not get 12 heads and 13 tails. You will get 16 and 9, or 10 and 15 · pure luck, no skill. Now put 44 people in a room and have them all do it. About half will finish ahead. The best one is not a better flipper. They are just the luckiest person in the room.
Every market below holds about that many trades. So its average is a coin-flip average: it wobbles. The bar shows how far it could have wobbled on luck alone.
HOW TO READ ONE ROW
- The dot is what the rule actually returned in that market.
- The line through it is everywhere the dot could have landed if the same rule had simply met a different run of luck.
- The faint vertical line down the middle is zero · no edge, nothing gained, nothing lost.
- If the line crosses the middle, the result means nothing. It is labelled INSIDE NOISE. A market can show +0.18R and still be worth exactly zero.
- Only when the whole line sits clear of the middle has anything been measured at all.
Across both rule sheets and 84 measurable market-tests, 9 clear the middle. All 9 of them are losses. None is a win. That is the finding, and it is why there is no ranking on this page: sorting these by return would put the luckiest coin-flipper at the top and call it a recommendation.
Blind fading of the extreme
Wait for a positioning extreme, wait for price to break its own 10-day range against the crowd, then enter the next open and fade the crowd. Stop and target fixed in advance from the market's own volatility.
ALL MARKETS TOGETHER
Across every market at once, 1,234 trades, this rule returned −0.03R per trade · anywhere from −0.08R to +0.01R once you allow for luck. That range covers zero, so the honest reading is: no measurable edge.
Split into single markets it gets weaker, not stronger. Each market holds a median of 29 trades, which is few enough that the average wobbles by about ±0.144R on luck alone. 17 of 43 markets come out positive · if every one of them were worthless you would still expect about 22.
1 market separates itself from zero once luck is accounted for, and it loses money. Zero markets clear the bar on the winning side.
alphabetical, never ranked by return · 95% interval · every number reproducible from research/strategy.py · metric: expectancy per trade, in R (risk units), 0.10% costs
Our own protocol, mechanised
Our own published checklist, run as a machine with no judgement: arm on the extreme, require the turn fingerprint, size the stop from this market's own historical shakeout.
ALL MARKETS TOGETHER
Across every market at once, 1,050 trades, this rule returned −0.10R per trade · anywhere from −0.26R to +0.07R once you allow for luck. That range covers zero, so the honest reading is: no measurable edge.
Split into single markets it gets weaker, not stronger. Each market holds a median of 26 trades, which is few enough that the average wobbles by about ±0.53R on luck alone. 20 of 41 markets come out positive · if every one of them were worthless you would still expect about 21.
8 markets separate themselves from zero once luck is accounted for, and every one of them loses money. Zero markets clear the bar on the winning side.
alphabetical, never ranked by return · 95% interval · every number reproducible from research/protocol_test.py · metric: mean R per trade, stop = max(median shakeout, 1 ATR)
What this does not say
It does not say your market is untradeable. It says that these two mechanical rules, applied blind, did not make money there · and that on this much data most of the differences between markets are luck rather than character.
Your own read was never in this test. Neither was your sizing, nor the weeks you decide to stay out. We cannot measure those and we are not trying to sell you a replacement for them. What we can hand you is the ground underneath: where the crowd sits today, how far past turns ran against the trade before they worked, and what the worst case on file says about size.
The Friday Letter
Every Friday 21:30 CET the new report lands and we regrade all 1,206 cases overnight. Confirm and your crowd-map snapshot arrives right away. Then every Friday: what moved, where the crowd is stretched, and which fades are running in public. And the moment we open one, you get the ping. Free.
Every Friday: where the crowd is stretched, what is running, what it cost. Free.
Confirmation by e-mail, one-click unsubscribe in every issue. Your address is never shared.