Edit with
Select a symbol
What the research agent has found in each symbol's history, tested the way the engine trades. Last full pass finished 2026-09-12 23:09 UTC.
Pass
2026-09-12
Symbols
51/51
Configs tested
20,808
Survivors
82
Rules on 2+ symbols
24
Joined the population
0 new · 66 refreshed
The same rule surviving the gate on several markets is the evidence that matters most: it is far less likely to be a fluke of one symbol's history.
None yet in this pass.
| Symbol | Rule | Side | Exit | Train n | Train avg R | Train t | Test from | Test n | Test avg R | Test win | Test t | Last signal |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Nothing survived the gate. | ||||||||||||
Gate: at least 15 trades on the first 65% of history and 10 on the rest, positive average R with t ≥ 2 in both halves, after fees. R is profit per unit of the stop's distance. Each candidate is traded exactly as the daily engine would trade it: entry at the signal bar's close, one position at a time, exits by the engine's own evaluator. Survivors join the population as testing strategies (traded on paper by the daily engine under their own names) and are never trusted by the Board until they earn it live.
Judges every strategy on its live record, never the calendar: its own trades when it has enough, otherwise its family pooled across every symbol. Trusted strategies vote on the Board; testing and demoted ones trade on paper; retired ones stop.