The lab's original rules
Every challenger is scored against these rules on the same flags.
The lab’s lessons, forward evidence, and rules in test.
Every closed paper trade becomes a lesson. Lessons become hypotheses. A hypothesis only changes the lab's rules after it wins on trades it has never seen.
A replay of its last run. Not a live execution. Connections show the research process, not individual lesson attribution.
The lab's original rules
Every challenger is scored against these rules on the same flags.
Tests a profit target together with a calmer stop
Practice score on past flags (not counted): +0.99 pts per flag vs current rules · hindsight
In test since 18 Nov 2026
Tests taking profit at a fixed gain
Practice score on past flags (not counted): +1.26 pts per flag vs current rules · hindsight
In test since 18 Nov 2026
Tests taking profit at a larger fixed gain
Practice score on past flags (not counted): +1.20 pts per flag vs current rules · hindsight
In test since 18 Nov 2026
Tests letting winners run further before taking profit
Practice score on past flags (not counted): +1.48 pts per flag vs current rules · hindsight
In test since 18 Nov 2026
Tests protecting a gain once it appears
Practice score on past flags (not counted): +0.06 pts per flag vs current rules · hindsight
In test since 18 Nov 2026
Tests giving the stop more room below the line
Practice score on past flags (not counted): −0.45 pts per flag vs current rules · hindsight
In test since 18 Nov 2026
Tests judging the stop on the day's close instead of every 15 minutes
Practice score on past flags (not counted): −1.22 pts per flag vs current rules · hindsight
In test since 18 Nov 2026
Tests requiring two closes below the line before stopping
Practice score on past flags (not counted): −0.31 pts per flag vs current rules · hindsight
In test since 18 Nov 2026
Challengers and the current rules are both replayed on the same chart-feed bars, so their comparison is like-for-like; the book on the home page is the lab's own record.
Nothing has been ruled out yet. Rejected challengers will stay listed here.
A challenger takes over only when all of these hold on forward trades: at least 20 flags traded by either version, at least 90% confidence it does better on the same flags, a positive average, and a drawdown no worse than the current rules' by more than a quarter. Then the version number moves up, the change is logged in Lab updates, and every earlier trade keeps the version it ran under.
Engine reproduced 39 of 45 lab trades exactly, 4 within one bar
Before trusting any challenger, the brain replays the lab's own rules on the same bars and checks it gets the same trades. Residual gaps come from the lab and the chart feed using different 15-minute data.
Forward evidence starts 18 Nov 2026. The chart fills in as new flags close.
Engine reproduced 39 of 45 lab trades exactly, 4 within one bar
Engine reproduced 34 of 45 lab trades exactly, 3 within one bar
Registered S-3: Tests requiring two closes below the line before stopping · motivated by 28 'noise_stop' lessons
The brain has now learned 75 lessons from closed paper trades and dead flags
The brain has now learned 50 lessons from closed paper trades and dead flags
The brain has now learned 25 lessons from closed paper trades and dead flags
Engine reproduced 37 of 45 lab trades exactly
Registered C-1: Tests a profit target together with a calmer stop · launch challenger
Registered E-1: Tests taking profit at a fixed gain · launch challenger
Registered E-2: Tests taking profit at a larger fixed gain · launch challenger
Registered E-3: Tests letting winners run further before taking profit · launch challenger
Registered E-4: Tests protecting a gain once it appears · launch challenger
Registered S-1: Tests giving the stop more room below the line · launch challenger
Registered S-2: Tests judging the stop on the day's close instead of every 15 minutes · launch challenger