Lexicon · frozen definitions

One fixed meaning per term, each a statistical property with its threshold.

These are the words we use to read prediction-market wallet quality. Each is defined once here and used the same way on every surface, from the analyzer to the research notes. Every one is a statistical property with a threshold, not an endorsement and not an instruction to act. We describe behavior on the public record we read; a past read is not a forecast.

Realized entry edge: How much more often a wallet's entries resolved true than the price it paid implied, on the probability scale. We read it with a bias-corrected and accelerated (BCa) bootstrap 95 percent interval, never a point estimate on its own.
FDR-cleared: Clears our false-discovery-rate-corrected bar. When you test many wallets, some look skilled by chance alone (about one in forty at a 2.5 percent threshold). The correction controls for that, so FDR-cleared means the read survives being one test among many. It is a statistical property, not a certification, and never an instruction to copy a wallet.
Skilled (verdict): Renders only when a wallet has at least 30 resolved positions, at least 20 independent resolved event clusters, coverage floors met, its BCa 95 percent lower bound strictly above zero, AND (since 2026-08-12) proof that the result survived false-discovery correction across the family the wallet was selected from. Without that last gate, a lower bound above zero renders as 'Ahead of its entry prices' - a record state, not a verdict - and its mirror, an upper bound below zero, renders as 'Behind its entry prices'. Nothing currently threads the multiplicity gate, so no per-wallet Skilled verdict renders anywhere today. The floors win: a flattering point estimate alone never reaches this state. It describes the resolved record, retrospectively and in-sample; it is not a forecast.
Not separable from chance (the luck verdict): The BCa interval includes zero, so on the resolved record the edge cannot be told apart from chance. This is NOT a claim that the wallet has no skill. It is that the record cannot establish one either way at this sample size, and the point estimate should not be treated as reliable.
Insufficient: Fewer than 30 resolved positions, or fewer than 20 independent resolved events: too thin or too clustered to tell an edge from chance in either direction. A thin record is not a verdict, good or bad.
Flagged (reserved integrity hold): Reserved for a wallet-integrity screen (wash and Sybil activity) that is not yet live; the current machine never produces it. A held record reads as held for an integrity review, not as a skill verdict either way.
Concentrated (style label): A single event drives at least 60 percent of the net result. A descriptive style label on the verdict, never a demotion: the equal-weighted per-position edge can be clean while the dollar outcome rode on a few events, so a copier experiences the lumpy path, not the average. Concentrated records can also mask gamed activity, so the label always travels with that caveat.
Edge Score: A descriptive behavioral composite we use to rank a cohort. It is a ranking input, not a skill verdict: a high Edge Score does not mean a wallet is FDR-cleared. Our own published top-50 Edge Score cohort returned zero FDR-cleared wallets.

How the bar and the bootstrap interval are defined: methodology. The terms in action on our own published cohort: the top-50 skill scan. The tests that came back null: negative results.