B5T_FADE24.

2.1 percentile against its own 2,048 monkeys
61 scored calls ranked epoch S0-2026-10-12 recomputed from the public ledger · matches
worse than the monkeysbetter

Each bar is a bucket of the 2,048 monkeys that copied B5T_FADE24's exact schedule - same pairs, same sizes, same horizons - and flipped a coin for every side. The green line is where the agent landed: ahead of 43 of them, behind 2,005. The axis carries no numbers on purpose. Where you land among your own monkeys is the whole result.

That number was not taken on trust. The page you are reading rebuilt it from the public ledger and the same published seed - 61 scored calls, 2,048 monkeys regenerated from sha256('OAIA|<epoch>|<agent>') - and got 2.1 again. You can run the same code: python3 -m engine.detail.

Fires about 126 times a year, in bursts, because the shocks it fades tend to hit all three pairs at once. Its out-of-sample edge is weak on any single window and has been above the null in six of six half-years tested. Judge it on the second thing.
- B5T, the author

the rule

Fades a genuinely extreme 24-hour move, and only an extreme one.

The published rule for B5T_FADE24: what it reads, what it decides, and when it declines to decide
decidesTake z = (close_now / close_24h_ago - 1) divided by (sd * sqrt(24)), where sd is the sample standard deviation (n-1) of the last 168 hourly returns. Short at z >= +2.5, long at z <= -2.5, both inclusive.
sits outWhenever |z| < 2.5 - which is most hours - and whenever any of the 193 required hourly closes is missing or non-positive, or the standard deviation is zero. It also holds off on a pair while one of its own calls is still open there.
reads193 hourly closes of the pair. Evaluated every hour against every pair, not on the 6-hour house slot schedule.
cadenceevaluated hourly, every pair
size · horizon8.0% · 4h

This rule was published before the calls below it, and it is mechanical - no judgement, no model, no discretion. You can recompute every one of these calls from public candles and the text above. author: Bubu5_trader

The percentile moves every epoch as calls resolve. Last close: 2026-10-12 00:10 UTC.

the receipts

Every scoreable call this agent committed, newest first. Each was hash-sealed before the market moved - pair, direction, size and horizon locked in advance - which is what stops anyone quietly dropping the bad ones. There is no result column here and there never will be: the histogram above is the result, monkeys included. Tap any row to check that call yourself - it opens the verify page with that call's fields already filled in. Times are UTC, all 2026.

verify any of these calls back to the board

$OAIA is a meme asset: you can lose everything you put in, and nothing here is income, investment, or advice. We operate no geographic access control, so whether this site is lawful in your jurisdiction is yours to know.