Are Polymarket Traders Hedging on Trade.xyz?

Polymarket is increasingly read as an information source rather than only a place to bet. Its prices are quoted as forecasts, and Alex McCullough's dashboard tracks how close those forecasts land to the outcome. If traders take the prices that seriously, an obvious question follows: do they use Polymarket to offset risk they are carrying somewhere else?

Hyperliquid makes that question testable. Under HIP-3, builders can deploy their own perpetual markets, and Trade.xyz was the first to do it, listing NVDA, TSLA, AAPL, GOLD, SP500 and other real-world assets as on-chain perpetuals. Those are the same assets Polymarket writes its finance markets about. A trader long NVDA perpetuals who buys "NVIDIA misses earnings" on Polymarket is hedging, and because both venues settle on-chain, the whole thing is visible.

On the evidence available, they do not. That answer needs a boundary drawn around it, because this method can only see a hedge that a single wallet holds both halves of, and wallets are cheap and disposable. Someone running perpetuals from one address and placing bets from another is doing exactly the thing being looked for, and leaves nothing on-chain that connects the two.

What the search does settle is the base case. The most visible form of cross-platform hedging, one wallet holding one asset with positions that overlap in time and point opposite ways, is not happening at any scale between these two venues. Anything subtler than that stays open. This is a finding about what the chain can show, not a verdict on whether the use case is worth having.

Hedging means holding a position that pays when your main position loses. Nothing about that changes on-chain. What changes is that you cannot ask a trader what they intended, so the intent has to be inferred from what the wallet did, and the inference needs three things to line up at once.

The wallet has to hold a Trade.xyz position in an asset, and a Polymarket bet on a question about that same asset. The two have to overlap in time, and specifically during the window the trader personally held the bet, not merely while the market happened to be open. And the two positions have to point in opposite directions.

Direction is the condition that does the work, and it took two attempts to measure. My first pass used the Polymarket buy price as a proxy, which does not hold up: price separates favourite from longshot, not bullish from bearish. Reading YES and NO directly fails for the same reason, because YES is not always the bullish side. "Will MSFT dip to $465?" resolves YES when the stock falls. What the numbers below use instead is the outcome token each wallet actually bought, read against a classification of whether YES on that specific question means the asset rises or falls.

Without that step, doubling down and hedging are indistinguishable. A wallet long NVDA perpetuals that also buys "NVIDIA will be the largest company by market cap" holds two positions that win together and lose together, adding exposure rather than covering it. That is what most of these wallets turn out to be doing.

Polymarket's Gamma API returns finance-tagged markets, filtered to the ones naming an asset Trade.xyz actually lists. That gives 2,004 markets and 116,054 proxy addresses. Polymarket routes most users through magic-login proxy contracts, so those addresses had to be resolved back to the externally owned accounts behind them via Alchemy on Polygon. 63,540 resolved, 52,514 did not, and after removing duplicates the sample is 63,264 unique EOAs. Everything below refers to those.

The Trade.xyz side is the complete fill archive since launch, roughly 338 million fills across every tradeable asset over 250 days. Cross-matching against the resolved EOAs found about 1.76 million fills belonging to Polymarket participants, from 1,413 distinct wallets.

96 of the 118 wallets had enough trade history to establish direction on both sides. Sorting them by what they actually held:

Counted by episode rather than by wallet, 37 took a hedging position at least once and 13 of those also doubled down elsewhere. Of the 31 above, 24 never took a doubling-down position at all.

The doubling-down wallets cluster in market-cap questions, of the "will NVIDIA be the largest company in the world by market cap on March 31" shape. A long NVDA perpetual alongside a YES on that question is one directional view expressed twice.

Among the 24 wallets that only ever hedged, 20 did it exactly once. The other 4 hedged two or three times, in every case against a single Trade.xyz asset and different sub-markets of the same Polymarket event. Even at this scale, the behaviour does not repeat.

Across the 68 hedging episodes from those 37 wallets, the median gap between placing the Polymarket bet and the market resolving was 9 days, the mean 13, and the longest 55.

For comparison, Artemis puts the average gap between a Polymarket market being created and resolving at 14.77 days. Market-cap questions written against a distant fixed date, with no scheduled catalyst to force early resolution, sit well above that and pull the average up.

Half the episodes, 34 of 68, were entered when the Polymarket contract was already trading at 90 cents or higher, meaning the market had largely settled on the answer before the position was taken. Splitting those 34 by the size of the Polymarket bet shows the pattern underneath:

Bet size, entry timing and side move together across the three bands. The largest bets arrive latest and sit on the favourite; the smallest arrive earliest and sit against it.

Grouping the 37 wallets by what they put on Polymarket, 1 bet over $1,000, 5 bet between $100 and $999, 11 between $25 and $99, and 22 between $2 and $24. Wallets appear in more than one band where they placed differently sized bets, though none of the six that bet $100 or more also placed anything under $100.

Set against the Trade.xyz side, the mismatch is the whole story. The mean Polymarket-to-Trade.xyz ratio is 0.055 and the median 0.0024. For a typical wallet here, the Polymarket position is roughly a quarter of one percent of the perpetual position it supposedly hedges. A hedge that covers 0.24% of the exposure is not a hedge. It is a bet placed by someone who happened to also be trading the asset.

The wallets in the $100 to $999 band are the exception, and the only ones where the two sides are within sight of each other.

None of the five ran anything periodic. Each traded a single asset, and none placed a further offsetting trade after the one identified. Three are linked to public X accounts.

Trade.xyz is a noisy place to look. The platform has run aggressive trading incentives, and Arrakis has shown how much of the volume that produces. Wallets farming volume for a future airdrop generate fills that look like positions but carry no directional intent, which both inflates the denominator and makes any single match harder to interpret. A venue without incentives, a centralised exchange or an established perpetuals market, would be a cleaner test.

The search was also narrow by construction. Only Trade.xyz was covered, not the other HIP-3 deployments. 45% of Polymarket proxy wallets never resolved to an EOA, mostly dormant email accounts, and those traders are invisible here. Direction was taken from the first fill in each interval. Nineteen matches had no retrievable Polymarket trade data and 22 more had no usable outcome token.

Hedging inside Polymarket would not show up here at all. Traders do build positions across economically related markets on a single venue, including simple constructions put on as much for entertainment as for risk, and this method is not looking for them. Cross-platform was the hypothesis worth testing, because it is the version that requires treating Polymarket as a risk instrument rather than a place to have a view. At these parameters it is not there.

Prediction markets native to Hyperliquid may also be the better comparison. A trader who already has perpetuals and prediction markets in one ecosystem has little reason to bridge to Polymarket to build a hedge, though the liquidity on those venues is thinner and may not support size.