OTR LEARNS · OTR-LEARN-2026-07-27-0016-SGT · OTR-QUALITY-35-2026-07-27-0016-SGT

It remembers the call, checks the result, then records the lesson.

OTR does not quietly rewrite a bad call. It keeps the original evidence, timing and rule visible, then adds what happened afterward. The active settlement keeps failure, uncertainty and limited changes visible.

See today's editorial →

How it learns

Four simple steps.

1

Read

Check sources and publish only what passes the evidence gate.

2

Lock

Freeze the claim, price, time and invalidation rule before the outcome is known.

3

Check

Compare the locked call with date-matched evidence. Mark PASS, FAIL, NO ENTRY, HELD or INVALID honestly.

4

Learn

Record one limited rule change only when the proof supports it. One outcome cannot rewrite the system.

What OTR has learned

Current, evidence-backed lessons.

These are operating lessons, not a claim that a prediction model has proven an advantage. Current settled sample: 3 outcomes.

26 Jul quality reset

12 published, 23 HOLD, 15 URLs
The refusal rule worked: weak lanes stayed empty.

Keep 0-5 as capacity. Never restore a quota.

PSQ paper call

ENTRY HIT and +0.86% PASS
The timing and result contract was checkable. One success does not prove an engine.

Preserve the result. Do not change weights.

Two Bitcoin paper calls

Both ENTRY HIT and later FAIL
A verified entry can still produce a failed result.

Keep failures public; require stronger breakout confirmation.

Ten expired U.S. addendum calls

Original 24 Jul windows passed without exact proof
Later quotes cannot prove an old entry.

Archive as HELD EVIDENCE / NO VERIFIABLE ENTRY.

Open the full proof and learning record →See past prediction results →Read the public learning data →Open the complete 27 July publication PDF →