The Public Loss Ledger.
Outcomes now resolved for prediction signals recorded in the trailing 14 days. Wins and losses come from the stored public archive.
Losses render at the same size as wins. Counts cover signals fired in the full trailing 14-day window whose outcomes are now resolved. The table shows the newest up to 200 signal rows and can omit older signals when that window is busier.
Model validation snapshot
The outcome stream behind this ledger.
Resolved outcomes are the hard part of the claim: this panel links each ledger view to live evidence quality and backtest signal stability.
Telemetry is still loading.
Alert and threshold posture
Model tuning diagnostics
Public view of ML edge evidence and threshold-risk posture, with invalid or stale measurements withheld.
Loading model telemetry...
Resolved outcomes for signals fired in 14d
Counts for CORE and Research Lab super signals fired in the trailing 14 days whose outcomes are now resolved. Outcomes may resolve later than their signal time. The capped table below may contain fewer rows. No profit factor, win rate, or customer-PnL claim appears here.
Newest resolved signal rows
Each returned row links to its standalone prediction record. The default view is losses; switch to all to see wins and flats in the same capped table.
Counts cover signals fired in the trailing 14-day window whose outcomes are now resolved; those outcomes may resolve later. The table is newest by signal time and returns up to 200 prediction rows, so older signal rows can be omitted. Percentages are path-labeled outcomes from stored market bars with a modeled round-trip cost allowance, not exchange fills, customer trades, or customer PnL. Information only. Not financial advice.
Verify, then decide.
Read the ledger first, including the worst row and loss context. If it still earns your attention, connect your wallet.
Want to verify the ledger first? Open the loss ledger.
The evidence posture is losses kept in plain sight beside wins. The separate Glass Ledger includes both registrations appended before their stored resolution timestamp and historical rows harvested later. Verify that split and the chain boundary, read the wider evidence window, then decide if the desk is worth following.