The moment a trader sees the same event priced at two different probabilities in two different places, one word jumps to mind: arbitrage. But a price gap on its own is not an opportunity. Standing between the visible gap and any actual profit are commission, settlement rules, liquidity, and the nature of the pricing mechanism itself. This guide separates genuine risk-free arbitrage from what is really just a price edge with positive expected value, and shows how the parimutuel pool used by PolySouq — with a 10% commission taken from the losing side only — rewrites the entire calculation. The conclusion we reach is blunt: true arbitrage is rare and practically out of reach for most individuals, and the realistic edge is disciplined price comparison driven by probability and expected value. Event trading carries real risk of real loss, and nothing here is a promise of profit.
A price in prediction markets is not a number issued by a central authority that knows the truth. It is the product of money and opinion pooled by one specific set of participants at one specific moment. Because every market gathers a different crowd, and because information reaches those crowds at different speeds, it is entirely normal for the same event — a team winning, Bitcoin clearing a price level, the outcome of a political decision — to be priced at an implied 58% in one place and 63% in another within the same minute.
That gap is the starting point of every conversation about arbitrage, but it is not proof that one of the two markets is wrong. Very often the gap is an honest expression of a difference in who is participating, or in how the event itself is defined. The main structural reasons probability diverges between two venues are:
That last point is precisely what kills most naive arbitrage attempts, and we will come back to it in detail. Before that, it helps to have the basic rules of the game clear in your head: read how prediction markets work and then reading prices and probabilities, because everything below is built on both.
The idea to take away from this section: a price difference between prediction markets is information, not necessarily an opportunity. Your job is to work out which kind of information it is.
Arbitrage, in the precise financial sense, is a set of simultaneous positions that guarantees a non-negative return no matter how the event resolves, after every cost. The key phrase is no matter how it resolves: if a single scenario exists in which you lose, what you are holding is not arbitrage — it is a risk position whose odds you believe are in your favour.
Most of what gets called "arbitrage" in event trading is really the second thing: a price edge, meaning you are buying a probability for less than you think it is worth. That is excellent and entirely legitimate, and it is the foundation of any sustainable approach — but it does not remove the possibility of losing on any individual position. Confusing the two is the single most dangerous mental error in this subject, because it pushes traders to size up on a false sense of safety.
The practical rule: before you call anything arbitrage, write out every possible scenario on paper, including the "administrative" ones such as a market being cancelled, or stakes being refunded on one side but not the other. If any scenario still ends in a loss, you are holding an ordinary position that must be judged with the logic of expected value, not the logic of a guarantee.
None of this diminishes the value of comparing prices. It simply puts price comparison where it belongs: a tool for choosing a better entry point, not a machine that prints profit.
On PolySouq you are not buying a "share" at a fixed price. You put your money into a shared pot alongside everyone else predicting the same outcome. At settlement, those who were right get their own stake back in full, then split what came in from the other side. This is a parimutuel pool, and it behaves fundamentally differently from an order book where the price is fixed at the instant you execute.
The formula is explicit and published: payout = stake × (1 + losing pool ÷ winning pool × (1 − commission)), where the commission is 10% and is taken from the losing side's pot only. In other words, the platform earns only when there are genuine profits to share. Nothing is deducted from your deposit, nothing from your own stake, and nothing from the winners' own money. A solvency identity is asserted on every market: total payouts + commission = total staked, so more can never leave a market than entered it. The mechanics are laid out in how PolySouq settles markets transparently.
Now consider what this means for any attempt at arbitrage:
The logical result: in a pooled model, the phrase "I locked the arb" is not even coherent, because you never bought a defined return — you bought a relative position inside a pool that has not finished forming. Anyone who ignores this builds an entire spreadsheet on a number that will change before kick-off.
The best way to understand a 10% commission is not to think of it as a slice of profit, but to translate it into a break-even probability. If the post-commission payout multiplier is m, the lowest true probability that leaves your position break-even is 1 ÷ m. The distance between that break-even point and the implied probability showing on screen is the real hurdle you must clear before talking about any "opportunity".
Take a simple binary market and run the numbers at different pool distributions:
Read it like this: if the implied probability of a team winning in football markets is 60%, you do not need the outcome to be more likely than 60% to profit over the long run — you need it to be more likely than 62.5%. Any price gap between two venues smaller than roughly two and a half points is not even covering the commission, let alone qualifying as arbitrage.
Notice another important behaviour: the gap peaks near the middle of the distribution (about 2.6 points at 50/50) and narrows as the market moves toward the extremes. Near-certain outcomes carry a lighter commission burden in points, but they hand back a very small absolute return, so a single miss wipes out several hits. There is no free corner in this table.
There are two cases where the commission is zero outright, and both are part of the design: if nobody took the other side — everyone was right and there are no losers — every participant gets their money back in full and the commission is zero. And if nobody backed the outcome that actually occurred, the market is void, all stakes are returned, and the commission is zero as well.
Take a realistic-looking case. A fixture in the football market, simplified to a binary question: "Does Team A win?" You enter early because the implied probability reads 40% while your own estimate — or the pricing on a parallel market you follow — says the true number is closer to 55%. You compute the payout multiplier, find 2.35×, and conclude you are sitting on an enormous margin.
What is invisible at that moment is that the pool is a living thing. Here is what actually happened through to close:
A
Check the solvency identity on the last row: winners split $4,200 × 0.9 = $3,780 on top of their fully returned stakes, the commission is $420, and the total is exactly $6,500 + $4,200 =
The central lesson: in a pooled market you are not buying a price, you are buying a relative position inside a pot that has not yet closed. Entering early usually gives you a better position, but it does not give you a guaranteed number — which is why any arbitrage equation that assumes the multiplier is fixed at entry is wrong from the first line.
Suppose, for the sake of argument, that you found a genuine gap wider than the commission threshold. The largest practical obstacle remains: executing both sides at the same time and in the same size. This is known as one-leg risk, and it is the number one reason "arbitrage" turns into a naked directional position.
Thin markets show large, tempting price differences precisely because they are thin. The paradox is that the gap widens in proportion to how little money is available to close it — and when you try to close it in meaningful size you find that:
Watching depth and volume is therefore not a technical footnote but a precondition for any discussion of price differences. Read liquidity and volume in prediction markets before planning anything multi-legged.
A blunt practical rule: if the position size you need to make the gap worthwhile is larger than the market can absorb without moving against you, the opportunity does not exist. A difference you cannot execute is not a difference — it is a number on a screen.
Even if price, size and timing all line up, one question remains that nobody thinks about until after they have lost: are the two venues talking about exactly the same event? In prediction markets, the event definition, the resolution source, and the cancellation rules are part of the product, not legal small print.
The questions that usually expose the mismatch:
PolySouq's rules here are clear and internally consistent, and that consistency is itself an advantage for the trader:
Now put that beside a second venue whose rules you have not read line by line. If one side of your "arbitrage" is refunded while the other settles normally, you no longer hold a hedge — you hold a single open position that was sized on the assumption it would never be open. Rule asymmetry does not reduce your edge; it converts a structure you believed was risk-free into an exposure you never agreed to.
The habit worth building: read the resolution rules before you read the price. A gap that exists only because two venues define the event differently is not an inefficiency to harvest — it is a translation error waiting to bill you.
If genuine arbitrage is out of reach, what is actually left? Something less glamorous and far more durable: using price differences as a signal about where your own estimate might be wrong, and as a way to pick a better entry when it is right. That is event trading done properly, and it is a skill rather than a loophole.
A workable sequence, in order:
One structural advantage of the parimutuel pool is worth holding onto here: because the commission comes only from the losing side, and because your own stake is returned in full when you are right, there is no drag on your capital simply for participating. That does not make you profitable — only a better probability estimate does that — but it means the arithmetic you have to beat is the arithmetic in the table above, and nothing hidden underneath it.
The honest bottom line: risk-free arbitrage across prediction markets is rare, short-lived, and largely inaccessible to individuals. Disciplined price comparison is neither of those things — it is available to anyone willing to do the work. But it is still trading, it still loses on individual positions, and no method, comparison, or system produces guaranteed profit. Only stake what you can genuinely afford to lose.
In theory yes, in practice almost never for individual traders. True arbitrage requires simultaneous positions covering every outcome, at prices summing to under 100% after all costs, with identical event definitions and identical settlement rules on both venues. Any one of those conditions failing turns the structure into an ordinary risk position. What most people call arbitrage is really a price edge with positive expected value, which can still lose on any individual trade.
At minimum, larger than the commission hurdle. On PolySouq the 10% commission on the losing pool translates into roughly 2.5 percentage points of break-even at mid-range distributions: at an implied 60%, your true break-even is 62.5%. So a two-point difference between two venues does not even cover the commission. Add execution slippage and the rule-mismatch risk on top of that before calling anything an opportunity.
No. PolySouq uses a parimutuel pool, so the multiplier shown at entry is a snapshot of how the pool is distributed at that instant, not a commitment. Every amount that enters afterwards changes the ratio, including your own, which enlarges the winning side and dilutes each participant's share. Any calculation that assumes the entry multiplier is fixed is wrong from the start.
From the losing side's pot only. Nothing is deducted from your deposit, nothing from your own stake, and nothing from winners' own money. The formula is payout = stake × (1 + losing pool ÷ winning pool × (1 − commission)). Winners always get their full stake back on top of their share, and a solvency identity is asserted on every market: total payouts plus commission equals total staked.
If everyone was right and there are no losers, there is no losing pool to distribute. Every participant gets their money back in full and the commission is zero. This is implemented behaviour, not a goodwill gesture — the platform earns only when there are actual profits to share.
The market is void. All stakes are returned in full and the commission is zero. The same applies to a market cancelled or voided by the operator: it is always a full refund, never a partial settlement.
Yes. On football markets a stake can be cancelled before kick-off for a full refund. That flexibility is one reason event definitions and cut-off times matter so much when comparing two venues — participation may close at different moments, and a position you assume is hedged may not be.
Creating an account is free. Funding is real money only — USDC on the Polygon network. You receive a personal Polygon USDC deposit address and send USDC to it; both native USDC and bridged USDC.e are accepted, with a
Disclaimer: Prediction markets are a legal and legitimate way to trade information about the outcomes of future events. However, trading carries risk and you may lose the full amount you trade — so only trade what you can afford to lose. This content is educational and is not financial or investment advice.