Open-source market forensics · Polymarket · top-volume watchlist connecting

The invisible hands.

Made by Henk van Ess · www.digitaldigging.org

A live screening of wallet-level trades on Polymarket's biggest bets, from the midterms to the World Cup, built entirely from open, unauthenticated APIs. Detectors hunt for accounts that trade as one. Click any account name to open its file. A flag is a lead, not a verdict.

Story 01 · Betting against themselves the hero: no opinion, just theatre

They bet Yes and No.

The same accounts bought both sides of the same question: thousands of dollars on Yes, thousands on No. Whatever happens on election night, these positions cancel out. Nobody does that out of conviction. It means the "betting boom" on the midterms is partly accounts betting against themselves, and the market signal they create is theatre.

How can betting on both sides make sense

A Yes ticket and a No ticket on the same question always pay out exactly $1 together: one of them wins. So if you can buy the pair for 98 cents combined, you pocket a guaranteed 2 cents per pair; no election-watching required. If you pay more than $1, you have locked in a guaranteed loss, and the only thing you bought is the appearance of trading volume.

Either way, these trades carry zero information about who will win. Every dollar of it inflates the headline number "$X million bet on the midterms".

If you use information from this database, credit Henk van Ess · digitaldigging.org

Story 02 · Temporal lockstep

 

New here? What am I looking at

Every bet on Polymarket is placed from a wallet, a blockchain account with a self-chosen nickname. All bets are public, timestamped to the second.

Read the chart like sheet music: each row is one account, time flows left to right, every dot is one bet. When dots line up vertically inside an amber stripe, those accounts pressed the button within 30 seconds of each other, on the same bet. Once is coincidence. Hundreds of times over months is a schedule.

Click any account name anywhere on this page to open its file: what it bets on, who it trades with, and links to its public profile.

two or more ring wallets inside one 30-second window
The ring's book · where their money actually went
All rings found in this corpus · click any name for its file

If you use information from this database, credit Henk van Ess · digitaldigging.org

Story 03 · Self-match

Both sides of their own trade

Wallets that appear as buyer and seller inside a single on-chain transaction. The bluntest wash-trade signal there is.

Why would anyone trade with themselves

A market is an auction: every buy needs a seller. Here the exchange matched an account with itself: it bought and sold the same bet in the same instant. You gain nothing and lose fees, unless the point is to make a market look busier than it is, or to nudge the price line. Traders call it a wash trade.

Once or twice can be an accident of fast trading. Fifty-two times is a habit.

If you use information from this database, credit Henk van Ess · digitaldigging.org

Story 04 · Counterparty concentration

Pairs that keep finding each other

On a liquid market your counterparties should be near-random. These pairs fill each other over and over.

Why repeated partners are suspicious

Thousands of accounts trade these markets, so who ends up on the other side of your bet is like bumping into strangers in a crowd: essentially random. These pairs keep bumping into each other.

The percentages show how much of each account's trading happens against that one single partner. When a third of your deals are with the same stranger, they are probably not a stranger.

If you use information from this database, credit Henk van Ess · digitaldigging.org

Story 05 · Conviction money the control group: what real betting looks like

The true believers.

Not everyone is faking it. These accounts put real, one-way money on an outcome and are stuck with it until election night. Betting big is legal and often smart; this list is your interview list: who knows something, or thinks they do? It is also the honest yardstick that makes the theatre in stories 01 to 04 visible.

Why show the honest bettors at all

Because a whale with an opinion is not a manipulator. Research on prediction markets keeps finding that most big one-sided positions are simply strong beliefs. If you only ever publish the suspicious accounts, you will smear the honest ones. The tell of a true believer: money flows one way, the position stays open, and the account would lose real dollars if it is wrong.

If you use information from this database, credit Henk van Ess · digitaldigging.org

Story 06 · The case register the memory: findings expire from the feed, never from the file

Nothing scrolls away.

Polymarket's public feed only shows the most recent trades of each bet, so evidence literally scrolls out of view within days. This register is the fix: every detector hit is written to our own database the first time it appears, and checked in on every later sweep. First seen, last seen, how many sweeps, at its peak. The cases that keep coming back are the ones worth a story.

Why keep a database instead of screenshots

A screenshot proves one moment; a register proves a pattern. When an account wash-trades once, it may be an accident. When the register shows the same case checking in sweep after sweep, week after week, with rising peaks, you can write "the behaviour persisted for N days" and back it with timestamps.

ACTIVE means the case was still visible in the latest sweep. DORMANT cases have stopped, or their trades have scrolled out of the public window. Both stay on file.

If you use information from this database, credit Henk van Ess · digitaldigging.org

Story 07 · Born yesterday the newcomers: fresh wallets, veteran confidence

New in town, betting big.

These accounts made their first visible trade within the last seven days and have already moved thousands of dollars. New bettors usually start small; operations that got flagged or banned start over with fresh wallets and pick up right where they left off. A newcomer that is already flagged by another detector is the strongest lead on this page.

Why newcomers matter, and one honest caveat

Creating a Polymarket wallet costs nothing and takes a minute. When a coordination ring gets noticed, the cheapest counter-move is to abandon those names and continue with new ones. The same person, the same strategy, a clean record.

The caveat: "first seen" means first inside our trade window (the most recent trades of each bet), not first on the platform. An old account returning after a long pause can look new. The wallet file's Polygonscan link settles it: the blockchain shows the real account age.

If you use information from this database, credit Henk van Ess · digitaldigging.org

Reference · The watchlist every bet this dashboard screens, refreshed daily

Every bet we watch.

The 200 biggest active bets on all of Polymarket, by lifetime volume. For each one we hold the most recent trades the public API exposes (up to 3,500) and re-screen them every five minutes. FLAGGED counts how many wallets from stories 01 to 04 are active in that bet; QUIET means none are.

If you use information from this database, credit Henk van Ess · digitaldigging.org

Forensics desk for investigators · every item is a lead requiring on-chain confirmation

Building the case.

Five tools aimed at an investigation, not a headline: how money is split, how it is cycled, whether two accounts share one pair of hands, and how to preserve what you find. Open any account file for a sealed evidence pack and a funding-trace worksheet.

If you use information from this database, credit Henk van Ess · digitaldigging.org

How this works

Every trade on Polymarket settles on Polygon and is published, wallet by wallet, through the public data-api. This dashboard tracks the top-volume bets on the whole platform (watchlist refreshed daily), holds their recent trades in DuckDB and re-screens everything every five minutes. No API keys, no scraping, no blockchain node.

What a flag means

A screening lead, nothing more. Market-maker bots legitimately match everyone. A whale with conviction is not manipulation. Confirmation requires a funding trace: do flagged wallets draw from the same source address on Polygonscan?

The research behind it

Counterparty-graph scoring follows the Columbia wash-trading study (SSRN 5714122). Volume caveats from Tsang & Yang (arXiv 2603.03136): naive on-chain volume overstates activity roughly 2.4x through mint/merge mechanics.

Small glossary

Wallet: a blockchain account, like an IBAN with a nickname. Share: one bet ticket that pays $1 if the outcome happens; its price (say 17 cents) is the market's probability. Fill: one matched deal between a buyer and a seller. Notional: the dollar value that changed hands. Ring: accounts that repeatedly bet in the same 30-second windows, mostly alongside each other. Churn: buying and selling the same ticket in near-equal amounts; it earns nothing but makes a market look busy.

A findings database by Henk van Ess · digitaldigging.org · credit required when using information from this database · built for the GIJN investigative workshop · data: gamma-api + data-api.polymarket.com · engine: DuckDB · usernames are self-chosen and unverified ·