Prediction markets are designed to aggregate beliefs about future events. What they’re not designed to do is stay in sync with each other.
Polymarket, Kalshi, Limitless, Myriad, and Opinion each run independent order books. Each has its own liquidity pool, its own mix of market makers and retail traders, and its own lag in processing new information. The result: the same outcome – “Will the Fed cut rates in September?” – can trade at 62¢ on one platform and 54¢ on another at the exact same moment.
That gap is money. Buying YES on the cheaper side and NO on the other costs 62¢ + 38¢ = say $1.00 in a bad scenario, but when the gap is real and you execute both legs, you pay less than a dollar for something that pays exactly one dollar. Guaranteed profit, regardless of whether the Fed cuts or holds.
This is prediction market arbitrage. And in 2026, with five major platforms running simultaneously, the opportunities are more frequent than most traders realize – but also shorter-lived, harder to execute, and far more fee-sensitive than they appear on a surface-level scanner.
This guide covers the full picture: how arbitrage works mechanically, which platforms create the most opportunity, how to find gaps in real time, what eats your spread before you capture it, and why execution infrastructure is the single biggest variable separating traders who profit from those who break even.
How prediction market arbitrage works
Every binary prediction market resolves at two values: $1.00 if the YES outcome occurs, $0.00 if NO occurs. Because exactly one outcome must happen, a YES share and a NO share for the same event on any platform always sum to $1 at resolution. Always.
The arbitrage insight follows directly from this: if you can buy YES on one platform and NO on another for a combined cost below $1, you’ve locked in the difference as risk-free profit. The outcome of the event doesn’t matter. One of your positions pays $1, the other pays $0, and you net $1 on a sub-$1 investment.
As 0xinsider’s arbitrage guide explains: “The combined value of YES plus NO is always $1 at resolution. That is true on Polymarket, on Kalshi, on Limitless, and on every other platform listing the same event.”
A concrete example from 0xinsider
At time of writing, the 0xinsider scanner showed this live opportunity:
- Ousmane Dembélé – who will win the Ballon d’Or in 2026?
- Buy YES at 18.9¢ on Polymarket
- Buy NO at 44.9¢ on Myriad
- Combined pair cost: 63.8¢
- Pays $1 at resolution
- Locked profit: +36.2¢ per pair
That’s a 56.7% return on committed capital – at resolution, however long away that is. The catch: you need both legs to fill at those prices, you need capital tied up until Ballon d’Or results, and you need to execute before the gap tightens. We’ll come back to all three.
The two types of prediction market arbitrage trades
Understanding the two trade structures clarifies when each applies.
Type 1: The pair-under-$1 trade
This is the cleaner setup. Buy YES on Platform A at price P1, buy NO on Platform B at price P2. If P1 + P2 < $1, you have guaranteed profit at resolution.
The math works regardless of outcome:
- If YES wins: your YES pays $1, your NO pays $0. Net: $1 – (P1 + P2) = profit.
- If NO wins: your NO pays $1, your YES pays $0. Net: $1 – (P1 + P2) = same profit.
The profit is the same either way. You don’t need a market view to trade this – only fast execution and accurate fee accounting.
Type 2: The single-leg trade
When the same outcome trades at different prices across platforms – say YES at 7.0¢ on Polymarket and YES at 7.5¢ on Limitless – you can buy on the cheap side and sell on the expensive side simultaneously.
From 0xinsider’s arbitrage guide: “Buy 1,000 YES shares on Polymarket for $70. Sell 1,000 YES shares on Limitless for $75. Locked-in spread: $5.”
Both positions settle to the same value at resolution. The $5 spread is your profit. The downside: this requires holding shares on the expensive platform before the trade, or executing a two-leg order that must fill simultaneously. If one leg moves while you’re working the other, you’re holding a directional position you never wanted.
The five platforms to know
Not all platforms create arb opportunities equally. Scale, liquidity, and fee structure vary significantly.
Polymarket
Polymarket is the dominant global prediction market. It runs a central limit order book (CLOB) on Polygon, with USDC as the settlement currency. With the largest liquidity pools and highest trade volume of any prediction market, Polymarket is almost always one leg of a profitable arbitrage pair.
For arbitrageurs: Polymarket’s CLOB infrastructure is EU-adjacent. The matching engine responds to orders within milliseconds, but that latency advantage is dramatically amplified by physical proximity to the server infrastructure. Traders running bots from Dublin can execute and receive fills 80-200ms faster than those connecting from US residential internet – a gap that determines whether you catch an opportunity or watch it close. The TradoxVPS Dublin location was purpose-built to address this: 1ms round-trip to the Polymarket CLOB.
Fee structure: Polymarket charges category-specific taker fees, typically in the 1-2% range. Maker orders (limit orders posted to the book) often have lower or zero fees depending on the market. At small spreads, this fee asymmetry is the difference between profit and loss.
Best for: Any arb trade involving a liquid market. Polymarket’s depth means large orders are more executable with less slippage than thinner platforms.
Kalshi
Kalshi is a CFTC-regulated prediction exchange – the only federally regulated prediction market in the US. Its user base skews toward US traders, and its market mix reflects that: heavy on policy, economics, and finance markets.
For arbitrageurs: Because Kalshi’s user base and Polymarket’s user base are largely separate, they price the same events with different sentiment and liquidity. That divergence is exactly where arb gaps come from. The Fed rate cut example above – YES at 12¢ on Polymarket, NO at 80¢ on Kalshi – is representative of the kind of divergence that appears regularly between the two platforms.
Fee structure: Kalshi charges per-contract fees, typically $0.01-$0.02 per contract. For small-size trades this is manageable; at scale the math still works. Because Kalshi is CFTC-regulated, it’s US-only – access requires a US-based account.
Best for: Policy and macro markets where Kalshi’s regulated positioning creates pricing that diverges meaningfully from Polymarket’s crypto-native user base. The TradoxVPS Chicago location serves traders who prioritize Kalshi and other CME-adjacent infrastructure.
Limitless, Myriad, and Opinion
These three platforms appear on the 0xinsider scanner and frequently show wider gaps than Polymarket-Kalshi pairs – precisely because their thinner liquidity creates more persistent pricing inefficiencies.
Limitless is EVM-based with series-dependent fees. Thinner book depth means large fills are difficult, but visible gaps tend to be wider and slower to close.
Myriad and Opinion are newer platforms. Several of the largest spreads on the 0xinsider scanner at any given time involve Myriad – the Dembélé Ballon d’Or example above had a 36.2¢ gap between Polymarket and Myriad. The tradeoff: verify book depth before assuming you can fill your full intended size.
For a deeper comparison of these platforms, see TradoxVPS’s prediction market platforms ranking for 2026.
Finding opportunities in real time: the 0xinsider scanner
0xinsider.com/arbitrage is the most accessible real-time arb scanner for prediction markets. It compares Polymarket, Kalshi, Limitless, Myriad, and Opinion simultaneously, surfacing all pairs where the combined cost is below $1. Each entry shows the two platforms, the YES and NO prices, the pair cost, and the locked profit per pair.
At the time of writing, the scanner showed 598 live opportunities, with a best earn of 36.2¢ per pair.
How to use the scanner:
The scanner is a candidate generator, not a trade signal. It shows the headline prices at the top of each order book. Before executing, verify:
- Book depth. The price at the top of the book is only valid for the first few shares. A 36.2¢ spread on a market where only 20 contracts sit at that price becomes a 5¢ spread once you reach further into the book. Check actual depth before sizing.
- Fee impact. The scanner doesn’t account for your specific fee tier on each platform. Run the fee math: gross spread minus taker fees on both legs. A 4¢ spread with 2% taker fees on both legs can disappear entirely on small-size trades.
- Execution window. How long has this gap been open? A 36¢ gap that appeared 30 minutes ago and is still visible suggests either thin liquidity or a market where participants haven’t updated prices. A 5¢ gap that just appeared will likely close in seconds.
As 0xinsider notes: “Three defensive moves: place both orders at the same time, size to whatever sits at the top of each book rather than reaching for depth, and use limit orders at or just inside the best ask so you do not pay through the spread.”
What actually eats your spread
This is where most new arb traders get surprised. The gross spread visible on a scanner is not what you make – it’s the starting point before the market takes its cut.
Fees
Polymarket’s taker fees run 1-2% depending on market category. Kalshi charges $0.01-$0.02 per contract. On a two-leg trade, you pay taker fees twice – once on each platform.
From 0xinsider: “Two taker legs can swallow a 0.4¢ spread before you blink.” On Polymarket, category-specific fees can reach 2% per leg. A 0.3¢ gross spread on a $1 outcome evaporates completely against a 1% fee on each side.
The fix: Use maker orders where possible. Post a limit order inside the best price and let someone else fill you. Maker fees are typically lower or zero, which fundamentally changes the economics on small spreads.
Slippage
The quoted price is only good for the first few shares (or contracts) in the book. If you’re trying to fill 500 contracts and only 20 sit at the advertised price, the next 480 fill at successively worse prices. Your realized average fill can be significantly worse than the top-of-book number.
From 0xinsider’s guide: “The number you see at the top of the book is the price for the first few shares, not the next thousand. Try to lift more than what’s quoted and your average fill drifts north of where you started. Always read book depth, not just the headline price.”
For large-size arb, book depth is the binding constraint – not just the spread size.
Execution risk (one-legged exposure)
This is the most dangerous cost, and the hardest to quantify in advance. Between the moment leg one fills and leg two fills, you’re holding a directional position you never intended to take.
From 0xinsider: “Between leg one filling and leg two filling, you’re naked on a directional position. If the cheap leg fills first and the expensive side runs before you can hit it, you’re stuck holding one side of a trade you never wanted.”
The practical implication: place both orders simultaneously where the platform allows. And never size a two-leg trade larger than you’re willing to hold one-legged through adverse movement.
Why latency is the only moat
Once a gap appears in the scanner, the window is measured in seconds – sometimes milliseconds. Market makers update their quotes continuously. The moment a gap becomes visible, bots on both platforms start filling it. By the time a retail trader with a 120ms internet connection opens the scanner and clicks, a meaningful portion of the book depth at the advertised price has already been consumed.
This is why professional arb traders don’t use the scanner as their execution interface. They run automated bots that:
- Monitor both order books via WebSocket feeds in real time
- Identify gaps the instant they appear
- Submit synchronized two-leg orders before human eyes register the opportunity
And they run those bots from servers co-located near the exchange infrastructure.
A trader connecting to Polymarket from a US residential internet connection faces 100-200ms of round-trip latency to the CLOB. A bot running from a Dublin data center co-located near Polymarket’s infrastructure sees 1ms. That 100-199ms difference is the gap between reliably catching opportunities and consistently arriving to a filled book.
For Kalshi, the dynamics are similar. The exchange operates US-side infrastructure, and the Chicago TradoxVPS location provides CME-adjacent latency for traders who work both platforms simultaneously.
The latency advantage compounds on Polymarket specifically because its CLOB is a continuous-time matching engine. Unlike slower platforms where price updates are batched, Polymarket’s order book updates in real time – meaning a 1ms advantage over a 100ms competitor means you see roughly 100 price updates for every one they see before submitting an order. At scale, that’s a structural edge.
For a detailed breakdown of how latency impacts execution on Polymarket, see how to test latency of your Polymarket VPS and the local PC vs. VPS comparison for Polymarket bots.
Risk factors beyond execution
Even well-executed arb trades carry residual risks worth understanding before deploying capital.
Resolution disputes
Polymarket and Kalshi resolve markets through different mechanisms. Polymarket uses UMA’s optimistic oracle, a decentralized dispute resolution system. Kalshi is CFTC-regulated and resolves through its compliance process. On most events – election outcomes, sports results, Fed decisions – both platforms reach the same resolution.
On edge cases, they don’t. A market where the underlying event is ambiguous, delayed, or subject to interpretation can resolve differently on each platform. An arbitrageur holding YES on Polymarket and NO on Kalshi for the same event suddenly has two long positions instead of a hedged pair – with no guaranteed profit.
This risk is low-frequency but high-impact. Prefer markets with unambiguous resolution criteria.
Capital lock-up
Profit from an arb trade is only realized at resolution. A trade placed today on a market that resolves in 90 days ties up capital for the full period. A 10¢ spread that pays off in 90 days is roughly 37% annualized return – compelling in isolation, but the opportunity cost of that capital (and the inability to redeploy it on fresh opportunities during that window) has to be factored into the calculus.
Professionals prioritize capital velocity: frequency of turnover multiplied by average spread matters more than any individual trade’s yield.
Liquidity withdrawal
Book depth at the moment of scanner visibility can look different from book depth at execution time. Large market makers maintain the ability to pull quotes quickly. An arb opportunity showing 200 contracts at a 15¢ spread can thin to 20 contracts by the time your order hits. Size positions to what you can realistically execute, not what the scanner headline implies.
For a strategic overview of how to approach Polymarket as a market participant more broadly, see how to win on Polymarket in 2026.
Setting up your arbitrage infrastructure
Running a prediction market arb operation at any meaningful scale requires three things: always-on execution, fast connectivity to exchange infrastructure, and enough compute to monitor multiple order books simultaneously.
A home PC covers none of these reliably. Sleep, internet outages, and residential latency all introduce failure modes that automated arb simply cannot absorb.
The standard setup for a Polymarket arb trader:
1. A Dublin VPS for Polymarket execution. The CLOB is EU-adjacent; Dublin co-location gives you sub-1ms round-trip to the matching engine. Any additional latency you introduce by connecting from further away is a structural disadvantage in time-sensitive executions. Plan tiers starting from $50-80/month (single-core, 4GB RAM) are sufficient for a single-bot Polymarket arb operation.
2. A bot framework connecting to Polymarket’s API. Polymarket exposes a REST API and WebSocket feed for real-time order book data. A Python or Node.js bot can subscribe to multiple markets, calculate YES+NO pair costs in real time, and submit both legs of a trade simultaneously. Pre-configured VPS environments with Python, Node.js, Docker, and cron support – like those on TradoxVPS’s Dublin platform – eliminate the setup friction entirely.
3. Capital and fee accounting. Before going live, model your expected fee load across both platforms at your target size. Most arb strategies require minimum spreads of 1.5-2¢ or more to remain profitable after taker fees on two legs. The 0xinsider scanner is useful for identifying which categories and platforms tend to produce larger, more persistent gaps.
If you’re also trading Kalshi, a Chicago VPS running alongside your Dublin instance gives you the latency advantage on both sides simultaneously. The Polymarket vs. Kalshi comparison on TradoxVPS is a useful reference for understanding the fee and execution profile differences between the two platforms.
For traders wondering whether they need this level of infrastructure, the practical test is simple: are you executing automatically or manually? Manual arb on large, persistent spreads (10¢+, thin markets, long resolution windows) doesn’t require a VPS – you have enough time to execute from a browser. Automated arb on narrow spreads (1-3¢) in liquid markets requires co-location to compete.
See the full guide on running a Polymarket bot 24/7 on a VPS for step-by-step setup instructions.
TradoxVPS for prediction market arbitrage
TradoxVPS is purpose-built for exactly this use case. Two locations – Dublin and Chicago – cover the two most important exchange infrastructures for prediction market arbitrage in 2026.
The Dublin VPS delivers 1ms latency to the Polymarket CLOB. The hardware – Ryzen 9950X, DDR5 RAM, NVMe SSD – handles multi-bot deployments and simultaneous order book monitoring across multiple markets without resource contention. The 3Gbps base network (10Gbps burst) handles the data volume of continuous WebSocket feeds across five platforms without degradation.
Pricing starts at $50-80/month for a single-bot setup – a cost that a single profitable arb trade per week can more than justify.
The best VPS location for Polymarket trading in 2026 guide covers the full infrastructure decision in detail, including latency benchmarks and platform-specific recommendations.
Frequently Asked Questions
Prediction market arbitrage is the practice of exploiting price differences for the same event outcome across different platforms like Polymarket and Kalshi. Because each platform runs its own independent order book, the same YES or NO outcome can trade at different prices simultaneously. Traders buy the cheaper side on one platform and hedge on the other, locking in a guaranteed profit at event resolution.
In theory, a fully filled pair-under-$1 trade pays $1 at resolution regardless of outcome – making it mathematically riskless on resolved profit. In practice, risks include execution risk (one leg fills before the other and prices move), slippage on large orders, resolution disputes between platforms, and capital lock-up for weeks or months while waiting for resolution. Fees can also eliminate the spread entirely if it’s too narrow. See TradoxVPS’s Polymarket VPS guide for execution infrastructure considerations.
Profits depend heavily on spread size, capital deployed, and execution efficiency. A 0.3¢ spread on a $1 outcome is only 0.3% per trade – viable only at high volume or large size. Spreads of 5¢ or more (as shown on scanners like 0xinsider) offer more meaningful returns but are also the quickest to close. Professionals running automated systems with co-located VPS infrastructure can execute dozens of trades per day across multiple markets. Most retail traders find capital lock-up and execution complexity limit returns.
A VPS is not strictly required for manual arbitrage on large, persistent spreads. But for automated or semi-automated arb where opportunities close in seconds, a VPS co-located near the exchange servers is a material advantage. TradoxVPS’s Dublin location provides 1ms latency to Polymarket’s CLOB, while the Chicago location covers CME-adjacent infrastructure. A residential connection adding 80-200ms per round trip can mean missing an opportunity entirely.
The most accessible scanner is 0xinsider.com/arbitrage, which shows live cross-platform gaps across Polymarket, Kalshi, Limitless, Myriad, and Opinion in real time. More sophisticated traders build custom bots using Polymarket’s CLOB API (REST + WebSocket) and Kalshi’s official API to monitor order books, identify gaps, and submit synchronized two-leg orders automatically. Running these bots 24/7 requires always-on infrastructure – a dedicated Polymarket bot VPS is the standard setup.
