Is prediction market arbitrage legal?

Yes, in most jurisdictions. Kalshi is CFTC-regulated in the US. Polymarket operates globally but restricts US access at the protocol level. Trading on both from a US address can raise compliance questions, so most serious prediction market arbitrage traders use venues their jurisdiction explicitly permits. Unlike traditional sportsbooks, prediction markets rarely limit winning traders, which makes them structurally friendlier for arbitrage.

How much can you make from prediction market arbitrage?

Cross-venue edges between Kalshi and Polymarket typically run 2% to 5% on sports and 3% to 8% on politics or macro markets. With a $20,000 bankroll split across venues and 10 to 20 trades per day, a well-tuned prediction market arbitrage bot can generate $30 to $150 per day in risk-free profit. Returns scale with bankroll and execution speed, not prediction skill.

What is the biggest risk in prediction market arbitrage?

The biggest risk is naked leg exposure: one side of the arb fills but the other gets sniped or rejected, leaving you with an unhedged directional position. Secondary risks include settlement criteria mismatches (two venues resolving the same event differently), bridge or custody risk on crypto-settled venues, and sudden liquidity drain during high-volatility events.

Do I need a bot for prediction market arbitrage?

In practice, yes. Cross-venue edges on sports markets typically close within 15 to 30 seconds. By the time a human can open two browser tabs, find the matching contracts, and place both orders, the edge is gone. Automated execution with sub-second latency on both legs is the only reliable way to capture prediction market arbitrage opportunities at scale.

Prediction Market Arbitrage: Profit From Gaps

Q: What is prediction market arbitrage?

Prediction market arbitrage is the practice of buying and selling binary contracts on the same event across different venues (Kalshi, Polymarket, PredictIt, sharp sportsbooks) when their prices disagree. You pay less than $1.00 in combined cost for a pair of contracts that guarantee $1.00 in payout, locking in a risk-free profit regardless of outcome.

Prediction market arbitrage is the cleanest form of risk-free trading that exists in 2026. You find the same event priced differently on two venues, buy both sides, and collect the gap as guaranteed profit. No forecasting. No luck. Just math and speed.

The idea is simple. The execution is where most people get stuck. This guide covers what prediction market arbitrage actually is, where the edges come from, the venues where it works, three worked examples with real numbers, how automated execution handles the hard parts, and the risks that catch inexperienced traders off guard. If you want the full category in one place, our prediction market arbitrage and value betting hub links every guide, tool, and product page we publish on the topic.

What Is Prediction Market Arbitrage?

Prediction market arbitrage is the practice of buying a YES contract on one venue and a NO contract on another venue for the exact same event, when the combined cost is less than $1.00. Because prediction market contracts always resolve to either $1.00 (if the event happens) or $0.00 (if it doesn't), owning both sides at a total cost under $1.00 guarantees a profit regardless of outcome.

Core formula: If YES on Venue A costs Y and NO on Venue B costs N, and Y + N < 1.00, then buying one unit of each guarantees 1.00 - (Y + N) in profit per contract pair.

This is structurally different from traditional sports arbitrage. In sportsbook arbing, you're converting between decimal odds like 2.10 and 1.95 and solving for stake sizes that guarantee a positive return. In prediction market arbitrage, the math is already normalized: every contract pays $1.00. You just check if the sum of two prices is below $1.00 and act.

That cleaner math is one reason prediction markets have quietly become the most efficient place for automated arbitrage. Another reason: prediction markets don't limit winning traders. Sportsbooks do, aggressively. If you want to understand why that matters and how it shapes the strategy, our primer on sports arbitrage betting covers the basics of how arb math works across venues.

Prediction Markets vs Sports Arbitrage

People new to prediction market arbitrage often assume it's just sports arbing on different platforms. It isn't. The mechanics overlap, but the edges, risks, and tooling are different enough that you need to think about it as its own thing.

Factor	Sports Arbitrage	Prediction Market Arbitrage
Pricing	Decimal or American odds	Binary 0 to 1 contracts
Venues	Bookmakers (house takes the other side)	Order books (peer-to-peer)
Typical edge	1 to 3 percent	2 to 5 percent on sports, 3 to 8 on politics
Edge duration	Seconds to minutes	15 to 30 seconds on sports, minutes on politics
Account limits	Aggressive after 2 to 6 weeks of winning	None on Kalshi, Polymarket, PredictIt
Market coverage	Sports only	Sports, politics, macro, crypto, culture
Settlement	Immediate credit to account	On-chain (Polymarket) or regulated account (Kalshi)

The no-limits advantage is the big one. Sports arbers burn real operational time opening new accounts, rotating IDs, and managing gubbed books. Prediction market arbers don't. You set up your accounts once and run indefinitely. Our guide on value betting vs arbitrage explores why the two strategies fit different types of traders, and most of that thinking extends to prediction markets.

The Venues That Make Prediction Market Arbitrage Possible

Not every prediction market supports arbitrage. You need liquidity, a usable API, and a user base that disagrees enough with other venues to create pricing gaps. Here are the ones worth watching in 2026.

Kalshi

Kalshi is a CFTC-regulated event contract exchange based in the US. Contracts are priced in cents (0 to 100) and settle in USD held in regulated accounts. Kalshi lists sports, politics, economics, weather, and culture markets. It has a clean REST and WebSocket API, supports FOK and FAK order types, and charges a percentage fee on winning trades that varies by volume tier (typically 3 to 7 percent).

Kalshi is the cleanest venue for US-based traders. Funds are held in regulated accounts, the API is straightforward, and support for automated trading is explicit in the terms of service. If you want to go deeper on API integration and bot design specifically for Kalshi, our Kalshi trading bot guide covers the details.

Polymarket

Polymarket runs a Central Limit Order Book (CLOB) on Polygon L2 and settles trades in USDC. It's the largest prediction market by volume and lists thousands of markets across sports, politics, crypto, entertainment, and macro. Polymarket has no exchange fees, but traders pay Polygon gas costs per trade (fractions of a cent). The protocol restricts US access at the IP level, so traders operating globally use Polymarket while US-based traders typically stick to Kalshi.

The no-limits policy is one reason Polymarket is the best venue for a high-volume arbitrage strategy. Winning traders are welcomed, not restricted. The API is public, well-documented, and designed for algorithmic access. Our breakdown of the Polymarket arbitrage bot architecture explains the CLOB mechanics and fee modeling in detail.

PredictIt

PredictIt is a smaller, academic-focused prediction market run under a CFTC no-action letter. It primarily lists political events (elections, policy votes, approval ratings) and caps positions at $850 per market, which makes it unsuitable for scale but useful as an additional data point for cross-venue pricing. The API is basic and there's no WebSocket feed, so it's rarely the primary leg of an arb. Still, for political markets where Kalshi and Polymarket diverge, PredictIt can sometimes confirm which side is mispriced.

SX Bet

SX Bet is the largest decentralized peer-to-peer sports betting exchange, with over $780 million in lifetime volume. It's a P2P order book where users post limit orders and bet against each other (no house position), settles in USDC on SX Network or Arbitrum, and charges 0% fees on single bets. Unlike Kalshi and Polymarket, SX Bet focuses primarily on sports (NFL, NBA, MLB, NHL, soccer) rather than political or macro markets, and it routinely offers some of the best US sports pregame prices in the world. Because it has no house margin, no account limits, and a public REST API that explicitly welcomes bots, SX Bet has become a go-to venue for serious arbitrage traders. For the full breakdown, see our SX Bet arbitrage guide.

Manifold Markets

Manifold runs play-money prediction markets, so it's not a direct venue for prediction market arbitrage. But it's useful as a sentiment signal: Manifold users tend to be well-calibrated forecasters, and their aggregated odds on political or cultural markets often lead Kalshi and Polymarket by hours on slow-moving events. Professional arbers use Manifold as a signal layer, not an execution layer.

Cloudbet and Sharp Sportsbooks

Sharp sportsbooks like Pinnacle (PS3838), BetInAsia, and Cloudbet technically aren't prediction markets, but they price the same sports events as Kalshi and Polymarket using decimal odds. When a sharp book's line disagrees with a prediction market's binary contract price, that's a cross-book arbitrage opportunity. The decimal-to-probability conversion takes extra math, but the edges are real and often larger than pure prediction market arbs. Our post on the best sportsbooks for arbitrage betting covers which sharp books are worth connecting to.

Where the Edges Come From

If prediction markets were perfectly efficient, all venues would price every event identically and arbitrage would be impossible. They're not, and it isn't. Edges exist for five structural reasons that aren't going away.

1. Liquidity Asymmetry

Polymarket might have $100,000 of depth on a major NBA game's moneyline. Kalshi might have $8,000. When $20,000 of new flow hits the thin side, that book repriced before the deep side moves. For 15 to 60 seconds, the two prices disagree by 2 to 5 percent. An arbitrage bot captures the gap, and the two prices then converge.

2. Regulatory Fragmentation

Kalshi is US-only (CFTC-regulated). Polymarket blocks US IPs. This means Kalshi's user base is American, usually more institutional, and reacts to US sports news on a different timeline than Polymarket's global crypto-native user base. On a European soccer match starting at 3am Eastern time, Polymarket's crowd is awake and actively pricing while Kalshi is comparatively dormant. That timezone gap alone creates persistent small edges.

3. Binary vs Decimal Pricing

Kalshi and Polymarket price events as 0 to 1 probabilities. Sharp sportsbooks use decimal odds with an overround. Converting between them introduces rounding and vig-stripping errors that can leave 1 to 2 percent of persistent gap. Combined with the other factors, this pushes cross-book edges higher than pure cross-exchange edges.

4. Fee Asymmetry

Kalshi's percentage fee on winnings creates a permanent structural distortion: the "fair price" on Kalshi for a 50/50 event is not exactly 50 cents, because the winning side will give up 7 percent of its profit. Polymarket has no percentage fees but has gas costs. These fee structures mean the true no-arb equilibrium between Kalshi and Polymarket is not "identical prices" but "prices offset by the fee differential." An arb bot that models this correctly sees edges that naive traders miss.

5. Fiat vs Crypto Settlement

Kalshi settles in USD held in regulated bank accounts. Polymarket settles in USDC on Polygon. The two settlement paths have different operational costs, different bridge risks, and different withdrawal timelines. This structural difference means capital can't flow instantly between the two venues when prices disagree, which is why the gaps don't close immediately.

Three Worked Examples of Prediction Market Arbitrage

Theory is fine. Numbers are better. Here are three specific arb types with real math.

Example 1: Cross-Exchange Kalshi vs Polymarket

This is the bread-and-butter trade for most prediction market arbitrage bots. You watch the same event on both venues and wait for a pricing gap.

Setup: Lakers vs Celtics, NBA regular season. Polymarket: Lakers YES trades at 0.43. Kalshi: Lakers NO trades at 55 cents (0.55). Total cost: 0.98. Guaranteed payout per contract pair: $1.00. Gross edge: 2.04 percent.

Now subtract fees. On the Kalshi side, if NO wins, you collect $1.00 minus 55 cents cost = 45 cents profit. Kalshi takes 7 percent of 45 cents = 3.15 cents. On the Polymarket side, gas runs about 0.2 cents round-trip.

Expected net profit across both outcomes: roughly 1.4 cents per contract pair after fees. At $1,000 per side, that's about $14 in guaranteed profit per trade. Small but consistent, and the bot finds 8 to 15 of these per day across active sports markets. Use our arbitrage betting calculator to model edges with your exact fee tier.

Example 2: Intra-Market Bracket Arbitrage

Polymarket often lists the same question in multiple formats. "Who wins the 2026 NBA Finals?" might exist as a single market with 16 contracts (one per team), and also as individual pairwise markets ("Celtics win finals YES/NO"). If the implied probability on one team's contract in the big market disagrees with the matching pairwise market, that's an intra-market arb.

Setup: Polymarket bracket: Celtics to win Finals at 0.28. Polymarket pairwise: Celtics Win Finals NO at 0.69. Total cost: 0.97. Gross edge: 3.09 percent.

Because both legs trade on Polymarket, there's no bridge risk, no fee asymmetry, and no execution latency between venues. You just need to catch the gap before another bot does. Intra-market bracket arbs are cleaner than cross-venue arbs but rarer, because Polymarket's own market makers arbitrage these aggressively. Still, when new brackets launch or when news hits unevenly, the gaps open up.

Example 3: Sharp-vs-Soft Cross-Book Arbitrage

This is the most lucrative type of prediction markets arbitrage for traders with access to sharp sportsbooks. The idea: Pinnacle (PS3838) is the sharpest book in the world and repriced within milliseconds of any new information. Kalshi and Polymarket aren't. When Pinnacle moves, prediction markets lag, sometimes by 30 to 90 seconds. During that lag, the prediction market still shows the old, now-wrong price. A cross-book bot buys the correct side on the prediction market before it catches up.

Setup: Pinnacle just moved Arsenal from 2.10 decimal (47.6 percent implied) to 1.92 (52.1 percent). Kalshi's Arsenal YES still trades at 0.47 for 45 seconds after the Pinnacle move. Expected fair price post-Pinnacle-move: 0.51 after vig stripping. Buying Kalshi Arsenal YES at 0.47 and hedging with Pinnacle at 1.92 locks in an edge around 3.5 percent after Kalshi fees.

Cross-book arbitrage requires a more sophisticated pipeline than simple cross-exchange arbs. You need to strip the vig from the Pinnacle line, convert to no-vig probability, check depth on both sides, and execute both legs before the Kalshi price catches up. Claw Arbs' cross-book engine runs a 14-step validation pipeline specifically for this case. For a deeper look at how sharp sportsbooks fit into an arb portfolio, check our guide to the best sportsbooks for arbitrage betting.

How Automated Execution Actually Works

Manual prediction market arbitrage doesn't work at scale. The edges close too fast. A human opening two browser tabs takes 30 seconds from alert to placed trade. By then the gap is closed and you're holding a naked leg. Here's what an automated system needs to do to capture these trades reliably.

WebSocket Feeds on Both Venues

REST polling every 5 seconds is too slow. Kalshi and Polymarket both offer WebSocket feeds that push order book deltas in real time. A prediction market arbitrage bot connects to both simultaneously and maintains an in-memory price cache updated with every event.

At Claw Arbs, the price cache is a singleton that receives updates from three WebSocket feeds (Kalshi, Polymarket, Polymarket Sports) and notifies three arb engines whenever a price changes. The notification pipeline runs in under 25 milliseconds from price update to edge check.

Edge Detection With Fee Awareness

Every price change triggers an edge calculation for every pair that includes that contract. The calculation must account for:

Kalshi fee tier: the percentage fee on winnings, which varies by volume
Polygon gas: small but non-zero cost per Polymarket trade
Slippage buffer: how many cents of adverse movement the bot can tolerate before the edge dies
Order book depth: how much size is actually available at the displayed best price

Without accurate fee modeling, a 2 percent gross edge might look profitable but actually lose money after fees. A bot that doesn't model fees correctly is worse than useless, it's actively harmful.

14-Step Execution Pipeline

Once an edge passes the initial check, it still has to clear a series of validations before the bot actually places orders. Claw Arbs' cross-book engine runs a 14-step pipeline for every candidate trade:

Cross-book arbitrage enabled for this venue pair?
Kill switch off?
Scanning not paused?
Not in view-only mode?
Halt-if-naked check: no existing naked leg on this match?
Sharp venue price fresh (less than 2 seconds old)?
Edge above minimum threshold?
Edge below maximum sanity threshold (prevents fat-finger arbs)?
Sufficient order book depth on both sides?
Authentication tokens still valid?
Match not rate-limited (prevents retrying a broken pair)?
Pair cooldown expired?
Execution mode set to live or paper?
Execute

Each step takes microseconds, so the total validation adds almost no latency. But skipping any of them leads to bad trades, naked legs, or worse.

Dual-Leg Execution With Naked Tracking

When the bot fires, it sends both orders concurrently and waits for both to acknowledge. The dangerous moment is between "leg 1 filled" and "leg 2 filled." If leg 2 fails, you have an unhedged directional position. The bot must detect this immediately, mark the match as naked, and halt new trades on that pair until the exposure is resolved.

At Claw Arbs, both arb engines share a single naked match set. If the cross-venue engine has a leg 1 fill without a hedge, the cross-book engine also halts new trades on that match. This prevents stacking multiple naked positions on the same event, which is how most amateur arb bots blow up their bankroll during volatile games.

The Risks in Prediction Market Arbitrage

Prediction market arbitrage is risk-free in the sense that the outcome doesn't matter once both legs fill. But the path from "edge detected" to "both legs filled" has real risk, and so do the conditions after settlement. Here's what actually goes wrong.

Naked Legs

This is the number one failure mode. Your bot places two orders. Leg 1 fills instantly. Leg 2 gets rejected because someone else took that price 150 milliseconds before you, or because the exchange rate-limited your request, or because the order book depth changed. Now you're holding an unhedged position on a directional outcome.

A good bot handles naked legs three ways: immediate detection (within milliseconds), retry logic (attempt a new hedge at a slightly worse price), and halt mechanism (stop trading the affected match). Bots that don't handle naked legs properly stack multiple unhedged positions during volatile events and lose money faster than they made it.

Liquidity Risk

Displayed best prices on thin prediction markets can have only $100 to $300 behind them. If your bot places a $2,000 order, only the first $300 fills at the expected price. The remaining $1,700 walks the book into progressively worse prices, destroying the edge. A bot that doesn't check depth before sizing orders will consistently underperform its theoretical edge.

Settlement Timing Mismatches

Kalshi and Polymarket can define the same event with different resolution criteria. "Will the US send troops to Taiwan?" might resolve differently based on how each venue defines "troops" or "send." An arbitrage bot that pairs these as identical markets and one resolves YES while the other resolves NO will lose on both sides. Always verify resolution criteria match before enabling a pair for auto-trading.

Regulatory Changes

Prediction markets are a young industry. The CFTC's treatment of event contracts has evolved substantially from 2022 to 2026, and more changes are coming. Kalshi's list of approved contract categories can shift. Polymarket's US blocking enforcement can tighten. A prediction market arbitrage strategy that depends on a specific venue being available needs a fallback when regulations change. For the current state of legality, see our breakdown on whether arbitrage betting is legal.

Bridge and Custody Risk

Polymarket settles in USDC on Polygon. Your capital sits in a smart contract on an L2 chain. Polygon has been reliable, but any bridge carries non-zero risk. Diversifying capital across Polymarket (crypto-native) and Kalshi (CFTC-regulated, USD held in regulated accounts) is a natural hedge against platform risk, at the cost of some operational complexity.

Why Claw Arbs Is Built for Prediction Market Arbitrage

Most arbitrage tools stop at detection. They show you a dashboard with flashing green numbers and expect you to place the trades yourself. That doesn't work when edges close in 15 to 30 seconds.

Claw Arbs executes both legs. Here's the specific architecture:

Four venues connected simultaneously: Kalshi via WebSocket and REST, Polymarket via WebSocket CLOB feed, PS3838 via 0.2-second REST delta polling, and BetInAsia via Playwright headless browser with a 20ms MutationObserver flush.
Three arb engines in parallel: a WebSocket-driven cross-venue engine for Kalshi vs Polymarket, a cross-book engine for sharp-vs-soft edges, and an intra-market engine for Polymarket bracket arbs. All three share a single naked leg tracker.
Sports matcher automatically maps Polymarket token IDs to Kalshi tickers using fuzzy team-name matching, game dates, and market types. You don't have to manually pair contracts.
Fee-aware edge calculation models your exact Kalshi fee tier, Polygon gas, and a configurable slippage buffer. Edges below your minimum threshold are ignored.
14-step cross-book pipeline validates every candidate trade against kill switch, freshness, depth, cooldown, and match rate limits before firing.
Paper trading mode runs the full pipeline against live data without risking capital, so you can validate your configuration before going live.

The system runs as a single async Python process on your own machine. It's not a cloud service that sends you alerts. Your API keys stay with you. Your trades fire through your own accounts. Nothing routes through a third party.

If you're comparing this against other software, our best arbitrage betting software roundup compares Claw Arbs against OddsJam, BetBurger, and RebelBetting. The short version: most competitors only do detection. Claw Arbs is one of the few that actually places both legs of the trade.

Getting Started With Prediction Market Arbitrage

If you want to run this strategy, here's the realistic path from zero to live execution.

Fund both venues: USD on Kalshi (1 to 3 day wire transfer), USDC on Polymarket (bridge from Ethereum or buy directly on a supported exchange). Split your bankroll roughly 50/50 to start. A $10,000 total bankroll is the minimum to generate enough trades per day to matter.
Start with detection only: watch live edges for a few days before executing anything. Our free live scanner streams the exact same signals the engine detects, so you can see them in real time without installing anything. This builds intuition for what 2 percent edges look like, how long they last, and which markets produce them most often.
Run paper trading mode: simulate execution against real order book data. This surfaces matching errors, fee miscalculations, and depth issues before they cost you money. Run paper for at least a week.
Go live with small sizes: start at $50 to $100 per leg. Verify that your actual fills match your paper fills within a reasonable margin. If they don't, fix the fee modeling before scaling.
Focus on sports first: NBA, MLB, and NHL have the deepest liquidity and most frequent cross-venue gaps. Politics and crypto markets have wider edges but thinner liquidity and higher settlement risk.
Consider live in-play arbs once your base strategy is working: live markets have bigger edges but move faster. Our guide to live in-play arbitrage betting covers the tech stack requirements.
Track everything: log every trade, every naked leg, every fee. You can't optimize what you don't measure. Closing line value (CLV) is the only metric that proves your bot has real edge over a meaningful sample.

Frequently Asked Questions

What is prediction market arbitrage in simple terms?

It's buying YES on one venue and NO on another venue for the same event when the combined price is below $1.00. Both contracts pay out $1.00 total no matter who wins, so the difference between what you paid and what you collect is guaranteed profit.

Is it legal to arbitrage Kalshi and Polymarket?

Kalshi is CFTC-regulated in the US and trading on it is legal for US residents. Polymarket restricts US IPs at the protocol level. Traders outside the US can use both. US-based traders typically stick to Kalshi only. For specifics, see our post on whether arbitrage betting is legal.

How much capital do I need to start?

$10,000 is a reasonable minimum. Below that, the fixed costs of gas, fees, and trade frequency make the economics marginal. At $20,000 split across venues you can generate $30 to $150 per day in risk-free profit with a well-tuned bot. Scaling above $100,000 runs into liquidity constraints on thin markets.

What's the biggest mistake beginners make?

Going live before validating the fee model and the event matching. A bot that pairs two contracts with slightly different resolution criteria isn't doing arbitrage, it's placing two unrelated directional bets. And a bot that doesn't subtract Kalshi fees correctly will execute trades that look profitable on the dashboard but lose money on settlement. Paper trade for a week before going live.

Can I run a prediction market arbitrage bot on a VPS?

Yes, and you should. Running on a residential connection means your bot goes down every time your ISP hiccups or your power flickers. A $20 per month VPS in a US-East datacenter puts your bot close to Kalshi's servers, reduces latency, and runs 24/7 without supervision. Claw Arbs is specifically designed to run on your own hardware or VPS, not as a hosted cloud service.

The Bottom Line

Prediction market arbitrage is real, it's profitable, and in 2026 it's one of the cleanest alpha sources available to retail traders. The edges come from structural differences between venues (regulatory, fee, liquidity, settlement) that aren't going away. The math is simpler than traditional sports arbing. The no-limits policies of Kalshi and Polymarket mean you can run the same strategy indefinitely without operational overhead.

But it only works with automated execution. Edges close in 15 to 30 seconds, and a human can't place two orders across two venues in that window. You need a bot that connects to both venues via WebSocket, models fees accurately, tracks order book depth, executes both legs concurrently, and handles naked legs cleanly when they happen.

That's what Claw Arbs is built for. Four venues, three arb engines, 14-step cross-book validation pipeline, paper trading mode, and a single naked leg tracker shared across all engines. If you want to start capturing these trades instead of just watching them on a dashboard, check our Kalshi vs Polymarket arbitrage guide for the cross-venue playbook, or download the software at clawarbs.com.