Whale X-ray for prediction markets

CALIBRE's labels predict what a wallet does next

Before you chase a price move or copy a whale, see whether the money behind it comes from traders who reliably lose. Tested on 983 wallets, older bets to newer ones.

Accepts a Polymarket profile link, a wallet address, or a market or event link. No address handy? Try a sample wallet, or open the radar.

The extension labels the Polymarket wallet SpecStavok mostly sniping: 72% of its bets were on outcomes already decided. — The extension, on a real wallet

traded in the 4,587 closed markets we replayed, across 142 leagues: $1.78B; traded in the 4,587 closed markets we replayed, across 142 leagues
of money behind 11,975 price moves, traced to 10,113 wallets: $197.6M; of money behind 11,975 price moves, traced to 10,113 wallets
bets scored against the price paid, across 5,740 wallets: 790,185; bets scored against the price paid, across 5,740 wallets
verdicts sealed on Walrus Memory before the result was known: 149; verdicts sealed on Walrus Memory before the result was known

Market volume is Polymarket's own figure for each replayed market. Every number is a plain sum over the study outputs: method.

The test: a wallet's label from its old bets, and what it did next

983 wallets chosen without looking at profit, sports markets. Each wallet gets the label it would have had from its older bets; the chart shows how it did on its newer ones.

−3.8 pts Proven losing record: 81% of 57 wallets lost again.
+2.2 pts Proven winning record: rare (40 wallets), modest.
−1.1 pts All 983 wallets, the baseline.

Edge is how much more often bets won than the prices paid implied, per bet. The whiskers are 95% intervals. The result is mostly driven by in-play bets, and older and newer bets of one wallet share its style, so this is not a trading test. Read the method and limits.

The same numbers as a table, with every label

Label from the older half	Wallets	Older edge	Next edge (pts)	95% interval	Negative next half
Proven losing record	57	−9.1	−3.8	−5.1 to −2.4	81%
Proven winning record	40	+8.3	+2.2	+0.5 to +4.1	38%
No proven edge	801	+0.2	−1.0	−1.3 to −0.7	60%
Too early	85	+2.7	−1.9	−4.3 to +1.0	60%
All wallets	983	+0.2	−1.1	−1.5 to −0.7	61%

The leaderboard has no signal; ranking by raw edge only finds the losers

The same 983 wallets ranked by raw edge on older bets, split in thirds, then followed on newer ones: the best third breaks even, the worst third keeps losing.

≈ 0 Past winners do not stay ahead.
-2.1 pts Past losers lose about 2 pts per bet.
No signal in the profit leaderboard (p 0.57).

Edge is how much more often bets won than the prices paid implied. Already-decided bets (97 cents or more, 3 cents or less) are excluded. Wallets trade the same events, so the p-values are optimistic. Read the method and limits.

All five samples, with the same numbers

Sample	Best third	Worst third	Rank corr.
All bets (983)	-0.2 pts	-2.1 pts	0.17 p <0.0001
All bets, wallets with 100+ bets (840)	+0.3 pts	-2.1 pts	0.21 p <0.0001
Pre-game bets, 100+ bets (403)	-0.8 pts	-1.9 pts	0.07 p 0.09
In-play bets, 100+ bets (683)	-0.5 pts	-2.0 pts	0.14 p <0.0001
Top earners on the profit leaderboard (23)	-1.1 pts	+0.4 pts	-0.04 p 0.57

One real verdict, start to finish

This is how every move on the radar is handled, shown with an actual verdict from the record rather than a diagram. It was written to Walrus Memory with a Sui checkpoint before the market resolved, then graded when it did.

It scored better than the market price on this one. Most graded verdicts are level with the price: the value is knowing who moved it.

Open this verdict

The price moves
Rockets vs. Suns 32% → 69%
Up 37 points in 10 min.
Who pushed it
192 wallets, 443 fills
The fills behind the move, summed by wallet, biggest first.
X-ray the takers
70% of the pushing money judged
Each big taker is scored against the prices it paid on bets that had already resolved. Wallets with too short a record are left unjudged, not guessed.
Square money
3% from proven losers
A caution signal, not a trade: it describes who pushed the price.
Sealed before the result
2026-11-08 03:04 UTC
Sui checkpoint 342,858,785, blob I_o-YvzQ8C… on Walruscan.
Graded
Resolved SUNS: forecast Brier 0.443, market 0.469
The outcome is its own memory, QesAUJsCtQ…, pointing back at the verdict.

Sharps and squares, without the hype

A square is a bettor who reliably loses: A proven losing record is the strongest label: in testing, those wallets lost again (-3.8 pts on every next bet, 81% of them).
A sharp is rare, and the edge is modest: A proven winning record needs z of 2 or more over 100+ bets. In testing those wallets stayed ahead modestly (+2.2 pts next). Leaderboard rank carries no signal.
Square money is a caution, not a trade: On any market we show how much of the latest move came from proven losers, and how much of it we could judge. In a replay of 11,975 moves, loser-driven moves did 3.1 pts worse for followers (p = 0.07); early evidence, not conclusive.

X-ray Polymarket as you browse

The Chrome extension adds a small panel to polymarket.com. On a profile it shows the wallet's label and record; on a market or game page it shows who pushed the latest move and how much of that money came from proven losers. It reads only the page address and talks only to this site.

Download for Chrome Install stepszip, 11 KB, Chrome, Edge, Brave

1Unzip the download into a folder you will keep.
2Open chrome://extensions and turn on Developer mode.
3Click Load unpacked and pick that folder.
4Open any Polymarket profile or market.

Memories on mainnet: 627; Walrus Memory, Sui
Sealed verdicts: 149; Before the result
Graded: 105; Against the market price
Stamped via Tatum gRPC: 134; of 134 verdicts, Sui checkpoint via Tatum

Check them yourself: the radar, the record of every graded call, and the proof with every blob on Walruscan.