Every call, and what happened next.
LAYA looks at the price again 1, 6 and 24 hours after every verdict and keeps score in public, including the misses and the ones still waiting. This is a record, not a promise.
After each verdict the outcome worker reads the curve again at the first confirmed block after 1, 6 and 24 hours. Every result counts, including pending, unavailable, migrated and failed windows and every contradiction. Returns are curve spot, before fees and slippage.
Totals
When each call landed.
How many verdicts LAYA made each hour, coloured by whether their 24 hour result is in yet.
Each bar is one hour of issued verdicts, stacked by the state of its 24 hour outcome window.
Action confidence against risk.
Each dot is one verdict. Dots in the shaded corner are the ones where the AI favored further research but also flagged high risk. We show them anyway.
x is risk.noul, y is the action top probability. The shaded corner holds research labels with risk at or above 0.7. They are published as they are.
Curve spot return by window.
How the price moved after each verdict. These are not profits anyone made: fees and slippage are not included.
Curve spot return from the reserve ratio at issuance to the first block after each window. Before fees and slippage, so not executable. Graduated curves count as unavailable, never as wins.
Research with elevated risk.
Verdicts where the AI favored research but also said risk is elevated. Open one to see both answers.
action.choice buy (v1) or research (v2) with risk.noul at or above the scoreboard risk_threshold. Not hidden, not re-scored.
The full ledger.
Every verdict for the selected question version, with the state of its 1, 6 and 24 hour check.
One row per result in the selected question version with each outcome window status, from the complete paged archive.