bloomberg flagged 200 million of polymarket volume as possible insider trading. how much of that is real

_local_nerd_

_local_nerd_

Active member
Messages
43
Reaction score
104
Points
33
bloomberg businessweek ran an analysis last week thats worth reading properly because the methodology is the actual story, not the headline number.
they took roughly 34,000 transactions flagged by an analytics outfit called polysights, august 2025 through june 2026, all on polymarket's global exchange where everything settles on chain and is publicly visible. about $200m of h1 volume carries what they call characteristics associated with potential insider activity

what polysights actually scores on: how recently the wallet was created, size and timing of the position, the odds when the trader entered, and whether the account concentrates nearly everything in one or a handful of markets.
geopolitical markets dominate. iran strikes and ceasefire contracts alone account for around $45m of the flagged volume, venezuela the other big chunk

polymarket's response was that theyve referred nearly 100 wallets to law enforcement. kalshi separately said their enforcement team opened 150+ investigations in q1, blocked 100+ potential insider trades with new screening, made 20+ referrals.
and heres my problem with it. every one of those signals describes a new user making a confident concentrated bet. that also describes someone who read a good analysis and went in hard. the flag and the behaviour are not distinguishable from the data alone and bloomberg says so themselves

so how much of the 200 million is anything
 
this is the correct question and its the oldest problem in pattern work. the signal set polysights is using isn't wrong, its just not sufficient on its own.
new wallet, concentrated position, entered at long odds, resolved correct. thats four indicators and each one has an innocent explanation. together they raise the prior. they don't establish anything. what would actually distinguish it is repetition across uncorrelated events. one person hitting a maduro contract is luck or information. same wallet cluster hitting maduro, then iran, then something else with no common thread except the person, thats a pattern
whether the bloomberg set does that work or just counts flags is the thing id want to know before i formed a view 🤔
 
theres a polymarket analyst guy who went at the bloomberg piece pretty hard on X. said they misread the on chain data and overstated the winnings. his line was do bloomberg journalists think someone has insider info about argentina winning the world cup 💀
reckons the wallet they built the story round made a couple hundred grand not the 1.5m they printed
 
worth being careful in both directions there. the rebuttal came from someone on the platform side with an obvious interest in the number being smaller, same as bloomberg has an interest in it being big. neither is neutral. substantive point in the rebuttal is fair though. a leaderboard discrepancy is checkable. if polymarket's own top position on that contract was 1.1m and the wallet bloomberg tracked never appeared at that level, that's a factual problem with the reporting and not a matter of interpretation.
 
morgan_c said:
worth being careful in both directions there. the rebuttal came from someone on the platform side with an obvious interest in the number being smaller, same as bloomberg has an interest in it being big. neither is neutral. substantive point in the rebuttal is fair though. a leaderboard discrepancy is checkable. if polymarket's own top position on that contract was 1.1m and the wallet bloomberg tracked never appeared at that level, that's a factual problem with the reporting and not a matter of interpretation.
Click to expand...
@morgan_c thats where i landed too. the aggregate methodology critique is arguable either way. the specific wallet figure is either right or it isn't and its verifiable on chain by anyone who wants to spend an afternoon
 
part i think matters more than the number is the composition. geopolitical markets accounting for most of the flagged volume isn't a coincidence and it isn't about sophisticated traders.

sports outcomes have thousands of people with partial information and none with the whole picture. an election has millions of independent inputs. a military operation has a small,identifiable,cleared group who know the date.

informational structure of the market determines how exploitable it is, and prediction markets have quietly listed the most exploitable category of event there is because that's where the volume goes when something big is happening.
 
worth noting the platforms know this now and are moving on it faster than any regulator asked them to. kalshi put in guardrails in march stopping politicians trading their own races and athletes trading their own leagues. polymarket published market integrity rules the same week
what none of that touches is the category vogafox is describing. you can screen a politician trading his own election because you know who he is. you cannot screen a cleared officer trading a contract on an operation he was briefed into, because the platform has no idea who he is and no way to find out. guardrails address the cases that were embarrassing. like structural problem is untouched
 
so here's what nobody in the coverage will say out loud!

polymarket listed a contract on whether a specific head of state would be removed by a specific date, while a us military operation to do exactly that was in planning. they listed it because it was going to do enormous volume
then the flagged trading happened, which everyone is shocked about
you cannot design a market on a covert operation and then be surprisd that the people running the covert operation showed up. that isn't an integrity failure. thats the product working as spacified
 
HughRoller said:
so here's what nobody in the coverage will say out loud!

polymarket listed a contract on whether a specific head of state would be removed by a specific date, while a us military operation to do exactly that was in planning. they listed it because it was going to do enormous volume
then the flagged trading happened, which everyone is shocked about
you cannot design a market on a covert operation and then be surprisd that the people running the covert operation showed up. that isn't an integrity failure. thats the product working as spacified
Click to expand...
you've been making this exact point about every one of these stories for months and ur right every time and nobody ever picks it up
 
well if you missed that one back in may, that's the google engineer. spagnuolo. staff security engineer, twelve odd years there, had access to an internal tool with the year in search data before it went public and allegedly traded 23 of the year in search contracts on polymarket between october and december. roughly 1.2m profit on about 2.7m risked. handle was AlphaRaccoon.
whatever happened with that? two months and i haven't seen anything since the arrest
 
Still pending as far as I can find, which isn't unusual at this stage. What's more interesting is that his case and the Van Dyke case now both turn on the same question, and it isn't whether they did it.

Van Dyke is the Special Forces sergeant charged over the Maduro contracts. Bought 436,000-odd "Yes" shares on whether Maduro would be removed by 31 January, made north of $404,000 when it resolved. Signed an NDA in December over the operation he'd been briefed into. Pleaded not guilty.

His lawyers filed a pre-motion letter on 6 July previewing a motion to dismiss. Their argument is that the CFTC has no authority over him at all, because the contracts weren't swaps under the Commodity Exchange Act - they were, in the filing's words, geopolitical bets. If it isn't a swap, the CFTC's insider trading theory has nothing to attach to.

The CFTC's answer is that Maduro's removal carried financial and commercial consequences - they cite crude prices, Venezuelan sovereign bonds, the bolívar - which makes the contract a swap.

Same question sits under Spagnuolo. Same question sits under every state fight Kalshi is running. It's one narrow definitional issue holding up an entire sector.
 
and note which way the defendant needs it to go. van dyke needs event contracts to be bets rather than swaps. kalshi needs them to be swaps rather than bets, because that's the whole preemption argument against the states.

so a soldier trying to avoid a commodities fraud conviction is arguing the states' position for them, in federal court, for free.
 
esports version of this has existed for years and nobody ever cared. players and staff know roster changes and stand-ins before they're public, markets are live anyway and the amounts were never big enough for anyone to bother. whats changed isn't the behaviour. its that the numbers got big enough to be worth a prosecutor's time and the ledger makes the case build itself
 
@Hextech.40 thats consistently how enforcement arrives in any market. conduct precedes the interest by years what triggers attention is scale plus evidentiary convenience, and on chain settlement delivers the second one for free.
polysights creator said something along those lines to bloomberg, that insiders are getting more sophisticated about how they enter markets and accumulate positions. which if true means the flag rate should be falling as the actual activity rises, and the measured number becomes less meaningful over time rather than more.
thats the thing about a public ledger. it only catches the people who havent adapted to it yet
 
@goldenfinger 's right, and thats already visible in the countermeasure. split the position across several fresh wallets funded from different sources and every signal polysights uses degrades at once. wallet age stops being informative, concentration stops being informative, size per wallet drops below the threshold

Bloomberg's own note says the traceable subset was wallets funded directly by a centralised exchange. anything funded through a dex or wallet-to-wallet wasn't in the sample at all

so the 200m is by construction the volume from people who didn't bother hiding
 
thats the bit that should worry people and its the same shape as every screening system. measurable population is the careless one. and while number goes up, everyone congratulates themselves on detection improving and the actual sophisticated activity has moved a layer down and out of the dataset entirely.
then the reported figure gets used to justify the next round of controls, which are calibrated against the careless population,which is the only one you can see
 
You must log in or register to reply here.
Back
Top