What is negative risk on Polymarket?

The short answer: it is a link between contracts, not a market type

The name is unhelpful, so it is worth clearing up first. Negative risk on Polymarket does not describe a risk level, a hedging product, or a category of question. It describes a contractual relationship between the legs of a multi-outcome event. When Polymarket lists a field where exactly one option can end up true, it can deploy those legs so that they are aware of each other at the contract layer rather than merely sitting next to each other in the interface.

The concrete operation is the one Polymarket's own documentation leads with: a No share in any market in the set can be converted into one Yes share in every other market in that set. The conversion runs through a contract Polymarket calls the Neg Risk Adapter. That single sentence is the whole idea, and almost everything else about how these markets behave falls out of it.

Why the conversion makes sense

Start from the logic rather than the contract. If a set of outcomes is genuinely exhaustive and mutually exclusive, then saying one option will not happen is the same statement as saying one of the others will. Those are not similar claims, they are the same claim written two ways. In a field of three candidates, No on the first is definitionally Yes on the union of the second and third.

Plain multi-outcome events do not encode that. Each leg is its own binary market with its own Yes and No tokens, and a No token on one leg has no contractual relationship to the others, even though the reasoning above still holds in the world. Negative risk closes that gap by making the equivalence executable: hold the No token, call the adapter, receive Yes tokens across the remaining legs. Because the position you already held was economically that basket, the conversion is a relabelling of the same exposure rather than a new position.

That is where the capital efficiency people cite comes from. It is not extra leverage and it is not a yield. It is that the collateral for a full basket does not have to be posted leg by leg when the contracts recognise that the legs cannot all pay out.

Binary, plain multi-outcome, and negative risk compared

The table sets the three structures against the attributes that change how their prices should be read together. Rows describing product details are marked as of 2026.

Stated as of July 2026. Contract behaviour and API fields change, so confirm current details against Polymarket's developer documentation. How a single market's Yes and No prices relate is covered in how to read implied probability on Polymarket.

What the sum looks like in live data

The sum-to-one property is checkable rather than theoretical, so here is a check against our own data. In the Cent Signals snapshot generated 2026-07-17, the next Prime Minister of Israel event appears with sixteen legs. Their Yes prices ran from 49.6 cents on the leading name down to 0.1 cents on several long-tail names, and the sixteen legs summed to 99.6 cents. That is about four tenths of a cent away from a dollar across a sixteen-way field, which is roughly what you would expect once quoted prices sit inside a spread and individual legs last traded at slightly different moments.

Two figures in that field are worth noticing beyond the sum. The distribution is extremely top-heavy: two names carried about 86 cents of the total dollar between them, and eleven of the sixteen legs sat at or below half a cent. A negative risk field is therefore mostly a two-horse question wearing a sixteen-row interface, and the long tail exists to make the set exhaustive rather than because those legs carry meaningful probability. Individual markets from fields like this, with their volume and liquidity, are listed on our markets page.

The partial-field trap, using our own data as the example

Here is the failure mode that matters most when reading these fields, and the honest way to show it is to point at our own dataset rather than someone else's. In the same 2026-07-17 snapshot, the legs we carry for the 2028 Democratic presidential nomination sum to about 41 cents, not a dollar. Read naively, a field summing to 41 cents looks like an enormous discrepancy in a market that is supposed to sum to one.

It is not. Cent Signals tracks a ranked subset of Polymarket rather than every leg of every event, so what our snapshot holds for that nomination is fifteen legs out of a field that is materially larger on Polymarket itself. The missing 59 cents is sitting in legs we did not collect. Nothing is mispriced; the sample is partial. How the subset is selected and what it deliberately excludes is written up in our methodology.

The general lesson generalises past this site. Any sum across a negative risk field is only interpretable if the field is complete, and completeness is an assumption about your data collection rather than a property of the market. Screenshots, third-party trackers, and API pulls with a page limit all drop legs quietly. A gap in a sum is far more often a gap in the data than a gap in the pricing.

Why a leg can move when nothing happened to it

Negative risk changes what price movement means, and this is the part that confuses people reading a single market page in isolation. In a standalone binary market, the price moves when the market's view of that question changes. In a negative risk field, the legs are competing for a fixed dollar, so a leg can move purely because a rival leg moved.

If a leading candidate falls twenty cents, that twenty cents has to land somewhere in the same field, and every other leg can drift up without a single piece of news attaching to any of them individually. Watching one row and concluding that sentiment about that specific name improved would be a misreading. The event, not the leg, is the unit of analysis. The broader question of what actually moves prices is covered in why do Polymarket prices move.

What an apparent gap in the sum does and does not represent

Sums slightly above or below a dollar in a complete field attract a lot of attention, and the arithmetic invites a conclusion the mechanics do not support. A field quoting to 104 cents does not describe four cents sitting available to whoever notices it first.

Several things stand between the quoted number and anything realisable. The quoted price is one side of a spread, and touching sixteen legs means crossing sixteen spreads. Polymarket applies taker fees on some markets, described in Polymarket fees explained. Depth on long-tail legs is often thin, so the size actually available at the quoted price can be small, which is the distinction drawn in volume vs liquidity on Polymarket. And the legs do not stand still while a multi-leg sequence executes. The published academic work on prediction-market arbitrage finds these gaps are typically small and short-lived, which is the pattern you would expect from a mechanism that many automated participants are watching continuously. We describe the arithmetic here; we do not suggest anyone act on it.

How to read a negative risk field

Pulling it together, three habits make these fields legible. Establish whether the field is complete before summing anything, because the negRisk flag on the Gamma API tells you the structure while your own collection determines whether you are seeing all of it. Treat the event rather than the individual leg as the unit, since the legs share one dollar and move against each other. And read a long tail of half-cent legs as the machinery that makes the set exhaustive rather than as sixteen independent forecasts.

Terms used above are defined in our glossary, and the general mechanics of how Polymarket works sit in what is Polymarket.

Frequently asked questions