…
Rules decide every entry and every exit. In the book’s first week, before a single trade had closed, we fixed a probability envelope for where it could plausibly land. Everything since is measured against it.
The bands were fixed on 19 July 2026, before the first trade closed, and have not been touched since.
The fan is the prediction. On 19 July 2026, the book's first week, before any trade had closed, we simulated the ruleset's own outcome distribution and fixed the percentile bands: median, quartiles and 5-95th tails. They have not been redrawn, re-fitted or re-scaled since.
The dark line is what actually happened, one point per closed trade. Hover or tap it for any trade's exact position inside the bands.
When the line broke below the 5th percentile in August, that was precisely the condition we said in advance would trigger a review rather than a change of story. The review was held on 19 August; no parameter was changed.
Rules decide, humans don't. Entries, exits, sizing and stand-downs are computed from price. There is no discretionary override, no "this one feels different", and no way to quietly skip a trade the system asked for.
Every idea gets a pass bar, required edge, minimum sample, and the control tests it must survive, written down before it sees the data. Goalposts that move aren't evidence, they're marketing.
Sixteen rejected ideas are published, each with its hypothesis and the reason it failed; survivors wait in staging until a scheduled review. Anyone can show you their winners. This is the part that tells you whether to believe them.
Five longs stopped inside three days, minus 5.31R, as the book turned from short to long after nine stopped shorts. The record: +12.11R after 91 trades, one band lower in the envelope. Plus five ideas tested and rejected in 48 hours, and what they had in common.
Read issue 04 →One letter a week: what the regime did, what the machines did, and one idea we rejected. No signals, no positions, nothing to buy.
Unsubscribe in one click. We don't share the list.