There is a trade that looks obviously free the first time you see it on a prediction market. YES is bid at 22 cents. NO is bid at 69 cents. Buy both and you have paid 91 cents for a pair that pays exactly one dollar at settlement, whatever happens. Nine cents of free money, repeatable across hundreds of markets. It is one of the most common ideas in this space and it is worth understanding precisely why it is not what it appears to be.
It is not an arbitrage, it is a two-sided quote
The trade is only an arbitrage if both legs fill at the same instant. If you are placing resting limit orders and waiting, which is the whole point since crossing the spread would erase the gap, then you are not arbitraging anything. You are quoting.
Here is the translation that makes it obvious. On a binary market YES and NO always sum to one dollar, so buying NO at 69 cents is the same position as selling YES at 31 cents. Which means:
Buy YES at 0.22 + buy NO at 0.69
is exactly
Bid 0.22 / offer 0.31 on YES
You are a market maker with a nine-cent spread. That is a perfectly respectable thing to be. It is just not free money, and everything that goes wrong for market makers now applies to you.
Problem one: only the losing leg fills
This is the structural issue and it does not go away with better execution.
A resting buy order fills when someone sells into it, which happens when the price is moving down through your level. So your YES bid at 0.22 fills when YES is falling, and your NO bid at 0.69 fills when NO is falling. But NO falling is YES rising.
The two fills require opposite price movements. To get both, the market has to oscillate down through one level and then down through the other. If it simply trends in either direction, exactly one of your orders fills, and it is always the one on the wrong side of the trend.
Trending markets fill you on the loser. Oscillating markets fill you on both. Whether the pair trade works at all is therefore a bet on realised volatility being choppy rather than directional, which is a much more specific claim than "the numbers add to less than a dollar."
Concretely: on one weather market observed in July, a NO leg posted at 0.69 filled while the YES leg at 0.22 never did, because the market was rallying. Over the next 40 minutes the filled leg moved from 0.69 to 0.57. The pair gap was 9 cents. The adverse move on the single filled leg was 11 cents. The edge was smaller than the risk it was compensating.
Problem two: queue depth works backwards
Most guides tell you to look for deep order books, and for a taker that is right: depth means you can size up without moving the price.
For a maker resting at the touch it is the opposite. Your order joins the back of the queue at that price level, first in first out. Everything already sitting there gets filled before you do.
- Tight spread, deep book: 1 cent wide, 300 to 1,500 shares queued at the best bid. Your order may never reach the front. Looks liquid, is unusable.
- Wide spread, thin book: 4 to 8 cents wide, 5 to 100 shares queued. You are near the front and will actually trade.
So the markets that look best on a screener are frequently the ones where a passive order does nothing, and the ones that look neglected are where a maker actually gets filled. The screening variable is not depth, it is queue length at your price, and almost nobody displays that directly.
Problem three: fees are charged per leg, edge is earned per pair
Fee structures differ sharply across venues and this is where a lot of otherwise reasonable strategies quietly die.
Polymarket's central book does not charge trading fees, which is why the pair gaps there are as tight as they are. Some venues charge on both maker and taker sides. A venue taking half a percent on each side costs a full percent round trip on a strategy whose entire gross edge might be 3 percent.
The asymmetry that matters is this: you pay fees on each leg, but you earn the gap once per completed pair. Add an unwind and you are paying three or four legs of fees against a single pair of edge. Any strategy where the round-trip fee is a meaningful fraction of the gross gap is fighting a headwind that compounds every rotation.
This is also why the same idea can be sound on a zero-fee book and structurally negative on a fee-charging one. The trade did not change. The cost of expressing it did.
Problem four: the tails have no depth
Headline volume is misleading in a way that specifically punishes this strategy.
A daily temperature market might carry $2.9 million a day across the category, but that volume concentrates in the two or three buckets nearest the expected outcome, because that is where the genuine disagreement lives. The far buckets, which are also the ones that most often look obviously mispriced, are close to empty.
A measured example. On one temperature ladder in July, a tail bucket was offered at 4.5 cents against a plausible fair value several times higher. The entire visible offer at that price was 93 shares. Four dollars and change.
You could be completely right about that bucket and the position would still be too small to matter. Apparent edge and capturable edge are different quantities, and the gap between them is widest exactly where the apparent edge is largest.
If one leg fills anyway, do not panic-cross
When you end up holding one side, the instinct is to hit the bid immediately and flatten. That is the most expensive available exit.
Measured on the same weather position: crossing the spread to exit cost 12 cents per share. Resting an offer at the touch instead would have cost 8 cents. On 25 shares that is a dollar, which was a third of the total loss on the trade.
But patience is not free either, because the same trend that filled you is still running. The workable answer is a ladder rather than a choice: rest passively first, walk the order one tick more aggressive every few minutes, and cross only when a hard limit is hit, whether that is time elapsed, a loss cap, or approaching settlement when liquidity evaporates entirely.
The general principle is worth stating on its own. Removing exposure is urgent. Paying the full spread to remove it is not. Those two things get conflated constantly and the difference is real money.
So when does providing liquidity actually work?
The conditions are narrower than they look, and they are checkable in advance.
- A real edge source you can name in one sentence. "The sum is less than a dollar" is not an edge, it is the price of inventory risk. "Two venues disagree on the same contract" is an edge. If you cannot say where the money comes from, it comes from you.
- Fees small relative to the gap. A one percent round trip against a three percent gross gap leaves very little once adverse selection takes its share.
- Queue position, not depth. You need to be near the front at your price, which usually means a wider spread and a thinner book.
- Choppy rather than trending. Both legs fill only if the price comes back. Markets resolving against a physical fact, like weather near settlement, trend by construction as information arrives.
- A rehedge plan before you need it. Decide the unwind ladder in advance. Deciding it while holding an unhedged leg in a falling market is how small losses become large ones.
Frequently asked
Is buying YES and NO below a dollar an arbitrage?
Only if both fill simultaneously. As resting limit orders it is a two-sided quote, and the gap is compensation for the inventory risk between the two fills.
Why does only one side fill?
YES and NO sum to a dollar, so the move that fills one leg pushes the other away. Both fill only if the price oscillates through both levels.
Does deep depth help?
Not for a maker at the touch. Your order queues behind the existing size. Deep books mean long queues and rare fills.
What is the biggest hidden cost?
Adverse selection. Your order fills precisely when the market is moving against it, and the size of that move is frequently larger than the spread you were trying to earn.
How do I check liquidity properly?
Look at the size resting at the exact price you want, not the market's headline volume. Tail buckets routinely show a few dollars of depth behind a large-looking mispricing.
Where to check the book
These markets settle on Polymarket's order book. SmartX is a terminal on top of it that puts the live probability next to what larger positions are actually holding, which is the cut that matters when the visible price and the visible depth are telling you different stories.