We answer for our verdicts.

Every verdict is snapshotted the moment it's issued and re-checked against the wallet's actual forward performance after 7, 30 and 90 days. COPY should keep winning; AVOID should keep losing. Here is how often we're right.

HEADLINE (30D, COPY+AVOID)
COPY VERDICTS RIGHT
wallet stayed net-positive
AVOID VERDICTS RIGHT
wallet bled after we said so
SNAPSHOTS ON RECORD
every scan adds one
SIGNAL QUALITY: DOES THE SCORE ITSELF PREDICT?

Hit rate grades our verdicts. This grades the raw number: the Information Coefficient is the rank correlation between the Skill Score at scan time and the wallet's actual next-30-days ROI, the same metric quant desks use to grade their own signals. IC 0 = the score predicts nothing; even 0.05 is a real edge; ICIR > 0.5 = it predicts consistently, month after month.

IC (30d forward)
ICIR
RECENT VERIFIED CALLS (30-DAY CHECKS)

Snapshots mature after 7/30/90 days. The table fills as time passes; honest track records can't be rushed.

Methodology: forward performance = PnL of trades the wallet made AFTER the verdict, from the same cleaned-data engine. RISKY verdicts are conditional by design and excluded from the headline. IC = Spearman rank correlation of Skill Score vs forward 30-day ROI; ICIR = mean monthly IC / std of monthly ICs. Engine versions are stored with every snapshot. Research tool, not financial advice.