Arbitrage is the most appealing idea in trading: profit without directional risk. Buy one thing, sell another, pocket the difference regardless of what happens.
On Polymarket, genuine arbitrage does exist. But most of it is contested by well-capitalised operators with infrastructure you don't have, and the versions accessible to retail traders offer thinner returns than the word "arbitrage" suggests.
This guide covers the four types honestly — including which ones aren't worth your time.
What arbitrage means here
In a prediction market, the two outcomes of a binary market should sum to $1.00. Yes at $0.60 implies No at $0.40. If they sum to more or less than a dollar, there's theoretically free money.
Real arbitrage on Polymarket takes four forms:
- Complementary — the two sides of one market mispricing against each other
- Cross-platform — the same event priced differently on Polymarket vs Kalshi
- Temporal — related markets across different timeframes disagreeing
- Correlated — logically linked markets that imply contradictory probabilities
Each has a different difficulty and a different realistic return.
1. Complementary arbitrage
The idea: if Yes trades at $0.48 and No at $0.49, they sum to $0.97. Buy both for $0.97, and one of them pays $1.00 at resolution. Three cents of guaranteed profit.
The reality: these gaps close in seconds. Bots monitor every market continuously for exactly this. By the time a human notices, it's gone.
When you do see a persistent gap, ask why before assuming it's free money. Usually it's one of: - The order book is thin — that price exists for $5 of size, not $500 - Fees eat it — a 3-cent gap is not 3 cents of profit after fees and gas - The market is about to resolve and one side is functionally decided
Verdict for retail: not realistic. You're competing against automated systems watching the whole platform. The exceptions are thin, small, and usually illusory.
2. Cross-platform arbitrage
The idea: the same event trades on both Polymarket and Kalshi at different prices. Buy the cheaper side on one, the opposite on the other, and lock a spread.
The reality: this genuinely exists — pricing does diverge between venues, sometimes meaningfully. But the practical requirements are steep:
- Capital on both platforms simultaneously. Capital sitting on Kalshi isn't earning on Polymarket.
- Jurisdiction access to both. Kalshi is US-regulated; Polymarket International serves a broader audience. Many traders can't legitimately use both.
- Speed. Gaps close fast once anyone notices.
- Matching resolution criteria. This is the killer. Two markets asking a similar question may resolve on different sources, at different times, or with different edge-case handling. If they disagree, your "arbitrage" becomes two opposing bets that can both lose.
That last point deserves emphasis: read both resolution criteria in full before assuming two markets are equivalent. Superficially identical questions frequently aren't.
Verdict for retail: possible but capital-intensive and slow to scale. Realistic only if you already operate on both platforms.
3. Temporal arbitrage
The idea: related markets covering different timeframes imply inconsistent probabilities. If a 15-minute market implies one thing about the next hour and the hourly market implies another, one is wrong.
The reality: this is more accessible than the first two because it's less mechanically obvious — it requires a view about how probabilities should relate across timeframes rather than simple price comparison. Fewer bots chase it.
But it isn't risk-free arbitrage. It's a statistical trade: you're betting that a relationship reverts to normal, and sometimes it doesn't. Calling it arbitrage oversells it.
Verdict for retail: the most accessible of the four, if you understand the timeframe relationships. Treat it as a strategy with an edge, not free money.
4. Correlated market arbitrage
The idea: logically linked markets imply contradictory probabilities. If "BTC above $100k by December" trades at 30% and "BTC above $90k by December" trades at 25%, that's incoherent — the lower threshold must be at least as likely.
The reality: these do appear, particularly in less-trafficked markets where fewer participants are checking cross-market consistency. They require real thought rather than a price feed, which is precisely why they survive longer.
The constraint is that the mispricings are usually in thin markets. You've found a genuine inconsistency and can trade $40 against it.
Verdict for retail: genuinely findable, but capacity-limited. Good for a sharp trader with modest size, not scalable.
What arbitrage returns actually look like
Time to be blunt about the maths.
A 2% arbitrage sounds excellent. But:
- Capital is tied up until resolution. A 2% edge over three months is roughly 8% annualised, before costs.
- Fees and gas subtract from every leg. On a 2% edge across two legs, they can consume half of it.
- Opportunities are sporadic. You can't deploy consistently, so your average capital utilisation is low.
- Thin books cap size. The 2% might only be available for a few hundred dollars.
Professional arbitrage works through volume and automation — hundreds of small edges captured continuously. That's an infrastructure business, not a retail strategy.
The uncomfortable summary: for most retail traders, arbitrage delivers less than a well-executed directional strategy with proper risk management. It just feels safer because the word implies certainty.
The risks nobody mentions
Resolution mismatch. The big one for cross-platform trades. Different sources, different timing, different edge cases. Read both resolution criteria fully.
Execution risk. You fill one leg and the other moves before you get it. Now you have a naked directional position you never wanted. On fast markets this happens more than you'd think.
Liquidity risk. You can enter but can't exit at size without moving the price against yourself.
Capital lock-up. Money committed to a slow-resolving arbitrage isn't available for better opportunities.
Settlement design risk. Worth knowing: a July 2026 Stanford study found Polymarket's 5-minute Bitcoin contracts were vulnerable to settlement manipulation, with the pattern largely absent in 15-minute contracts. Any strategy relying on precise settlement behaviour should account for how a contract settles. Full details here.
A more honest alternative
If you're drawn to arbitrage because it sounds low-risk, the underlying instinct is good — you want defined risk and systematic execution rather than gambling on hunches.
You can get that from systematic directional trading with proper controls:
- Defined risk per trade — you know your maximum loss before entering, because it's what you paid
- Rules-based entries — no hunches, no revenge trades
- Odds discipline — refusing trades outside a set price range is the single highest-impact rule most traders skip
- Loss limits — a hard daily cap that stops you
That's not arbitrage. But it's achievable at retail scale, it scales with your capital, and it doesn't require competing with professional infrastructure.
Frequently asked questions
Is arbitrage on Polymarket legal? Yes. Arbitrage is normal market activity that improves pricing efficiency. Your eligibility to use the platform depends on your jurisdiction.
How much capital do I need? For cross-platform arbitrage, meaningfully more than most retail traders have — capital must sit on both venues simultaneously. Correlated-market arbitrage can work with less but is capacity-limited.
Can I automate arbitrage detection? Yes, and you'd need to. Complementary gaps close in seconds. But you're competing with existing bots, so you're competing on infrastructure quality.
Why do mispricings persist in some markets? Thin liquidity and low attention. Markets nobody's watching stay wrong longer — which is also why you can't trade size against them.
Is "risk-free arbitrage" ever real here? Complementary arbitrage is genuinely risk-free if you fill both legs at the stated prices and fees don't eat the gap. Those conditions are rarer than the theory suggests.
The bottom line
Real arbitrage exists on Polymarket, but the accessible versions are small, sporadic, and capacity-limited, while the profitable versions require infrastructure and capital most retail traders don't have.
If you're hunting arbitrage because you want defined risk and systematic execution, you can get both from disciplined directional trading — without competing against firms whose entire business is being faster than you.
Chase mispricings if you enjoy the hunt. Just don't expect the word "arbitrage" to mean what it implies.
Metazen Pulse runs 12 systematic strategies on Polymarket's 5M and 15M Bitcoin markets, with per-strategy odds windows and daily loss limits you control. Watch every non-premium signal live and free on our Telegram channel.
Related: Polymarket vs Kalshi for automated trading · The Polymarket API guide
Trading prediction markets involves real risk of loss. Past performance does not guarantee future results. Nothing here is financial advice.
See the strategies before you trust anything
Every non-premium signal is published live and free on our Telegram channel — follow the calls in real time and check them against actual market outcomes.