Sports Arbitrage Bot: How Automated Arb Software Finds and Executes Trades
A sports arbitrage bot has two jobs: find prices that lock in a profit, and place both bets before the window closes. Here's how automated arb software works end to end, and what separates a real execution bot from an alert service.
Search for a sports arbitrage bot and you'll find two very different products wearing the same name. One is a scanner that lists opportunities and pings you to go place the bets yourself. The other is an automated system that detects the edge and fires both legs of the trade on its own. They sound similar. In practice they're not close, and the gap between them decides whether you capture the profit or just watch it close on your screen.
This guide covers what a sports arbitrage bot does, how automated arb software works from the odds feed all the way to the filled order, and what to look for when you pick one. We'll use Claw Arbs as the running example because it executes rather than only alerts, but the mechanics apply to any serious arb bot. If you're brand new to the strategy itself, start with our primer on sports arbitrage betting and come back.
What a sports arbitrage bot actually does
Arbitrage betting means covering every outcome of an event at prices that guarantee a profit no matter the result. It's a math problem, not a prediction. You're not betting on who wins. You're exploiting the fact that two venues disagree on the price by enough to leave a gap.
An arbitrage exists when the combined cost of backing every outcome is less than the payout. Cover both sides of a two-way market for a total of 97 cents on the dollar, and you collect $1 whoever wins. The 3-cent gap is locked profit, independent of the result.
The catch is that these gaps are small, they appear without warning, and they close fast. A person refreshing odds pages can't reliably find them, and even when they do, the price has usually moved by the time they've logged into the second venue. That's the job an arb bot exists to solve. It watches every price on every connected venue continuously, runs the arb math the instant a price ticks, and acts before the edge disappears.
A full arbitrage betting bot does four things a person can't do fast enough by hand: monitor thousands of prices at once, recompute the edge on every single update, size both stakes correctly, and place both bets in the same breath. Miss any one of those and the strategy leaks money.
Scanner, alert bot, execution bot: three different tools
The word "bot" gets attached to three products that do very different amounts of the work. Knowing which one you're looking at matters more than the price tag.
Odds scanners show a table of current arbs. You read it, decide, and place every bet manually. These are useful for learning what arbitrage looks like, but by the time you've clicked through two sportsbooks the edge is often gone.
Alert bots push a notification the moment an edge appears, which beats watching a table. You still place both bets yourself, though. You've shaved a few seconds off the discovery step and changed nothing about the execution step, which is where most of the delay actually lives.
Execution bots detect the edge and place both legs automatically, with no human in the loop between the price ticking and the money going down. This is the only category that reliably captures short-lived windows, and it's the category most tools marketed as an "arb bot" quietly are not.
| Capability | Alert-only scanner | Execution bot |
|---|---|---|
| Finds the arb | Yes | Yes |
| Places the bets | No, you click through by hand | Yes, both legs automatically |
| Time from edge to bet | 30 to 120 seconds (human) | ~20 milliseconds |
| Catches sub-60s live windows | Rarely | Yes |
| Naked-leg protection | None, you're on your own | Halts the market on an unhedged leg |
| Prediction market venues | Usually none | Kalshi, Polymarket, Predict.fun, SX Bet |
Most paid tools in this space, including the big names, live in the first two columns. They're odds-comparison products with good data and a clean interface, and they alert you well. They just don't place the bet. For a fuller breakdown of who does what, see our arbitrage betting software comparison.
How a sports arbitrage bot works, step by step
Under the hood, every automated arb bot runs the same five-stage loop, thousands of times a second. Here's each stage and where the hard engineering actually lives.
1. Odds feeds from every venue
The bot needs live prices, and how it gets them decides how fast it can act. The best connections are WebSocket feeds, where the venue pushes each price change to the bot the moment it happens. Claw Arbs uses WebSocket feeds on Kalshi and Polymarket that deliver updates in around 25 milliseconds. For venues without a push feed, it watches the odds board directly and detects a change in about 20 milliseconds. The slow option, and the one many commercial tools rely on, is polling a third-party odds API that itself scrapes the books, which stacks 500 to 2000 milliseconds of lag on top before the bot even sees the number.
2. Matching the same game across venues
An arb only exists between two prices for the same outcome, and that sounds trivial until you realize every venue names things differently. Kalshi encodes a game in a compact ticker. A sportsbook writes out "LA Lakers vs Boston Celtics." A prediction market lists "Will the Lakers beat the Celtics?" The bot has to recognize all three as the same market and the same side. This is fuzzy team-name matching, and getting it wrong is dangerous: match two different games and the bot will happily place two bets that aren't hedged at all. A good arb bot spends real effort here, and a careless one blows up here.
3. Calculating the edge after fees
Once two prices are matched, the bot computes the true edge, and the word that matters is true. Gross edge is easy: add the two implied prices and check whether they sum to less than the payout. Net edge is what you actually keep, and it has to subtract every cost. On Kalshi that's the trading fee, roughly 7% of price times one-minus-price per contract, which peaks near a 50-cent price. On Polymarket it's the gas cost of the Polygon transaction. On any venue there's slippage, the risk the price moves between decision and fill. A tool that shows gross edge and calls it profit is misleading you. The edge has to clear all of it before the bot pulls the trigger. Our guide on how to find arbitrage bets walks through this math by hand.
4. Sizing the two stakes
To be fully hedged you need both legs to pay the same amount whichever way the event resolves. Because the two sides trade at different prices, that almost never means equal dollar stakes. The bot solves for the stake on each side so the payouts match, then checks the size against the available liquidity, or depth, in each order book. There's no point computing a $2,000 arb if only $300 of it can actually fill on the thin side. Depth-aware stake sizing is what separates a bot that works on paper from one that works with money.
5. Executing both legs
This is the step scanners and alert bots skip entirely, and it's the hardest one. The bot has to place both bets almost simultaneously, because the instant leg A fills, you're exposed on leg B until it fills too. That exposure is called a naked leg, and it's the single biggest risk in automated arbing. If leg B fails, you're holding a one-sided bet you never wanted. Claw Arbs fires both legs inside the same event loop and runs a safety pipeline that halts a market the moment a leg goes unhedged, so the bot stops trading that game rather than piling on more exposure. We go deep on that machinery in our guide to automated arbitrage execution.
A worked example: one arb from edge to profit
Numbers make this concrete. Say the Lakers play the Celtics, and two venues disagree on the price. Basketball has no draw, so exactly one side resolves to $1 and the other to zero. Back both sides and you've covered every outcome.
| Leg | Venue | Bet | Price | Cost (1,000 contracts) |
|---|---|---|---|---|
| A | Kalshi | Lakers win (YES) | $0.51 | $510 + ~$17.50 fee |
| B | Polymarket | Celtics win (YES) | $0.44 | $440 + ~$1 gas |
Back leg A: buy 1,000 "Lakers win" contracts on Kalshi at $0.51. That's $510, plus Kalshi's fee. At a 51-cent price the fee runs about 1.75 cents per contract (7% x 0.51 x 0.49), so roughly $17.50 on 1,000 contracts.
Back leg B: buy 1,000 "Celtics win" shares on Polymarket at $0.44. That's $440, plus a small Polygon gas cost, call it a dollar.
Whichever team wins, exactly one side pays $1 per contract, so your payout is $1,000. Your total outlay was $510 + $440 + $17.50 + $1 = $968.50. Locked profit: $31.50, about 3.25% on the money committed, regardless of the result. No prediction required.
Notice what happened to the edge. The gross gap was 5 cents on the dollar (95 cents to cover a dollar of payout), but Kalshi's fee ate most of the difference, leaving 3.25% net. This is exactly why a sports arbitrage bot has to compute net edge, not gross. An alert tool flashing "5% arb" would have overstated your real return by more than half. The bot that quotes you the honest number after fees is the one you want.
What actually matters when choosing an arb bot
Once you know a tool actually executes, these are the dimensions that separate a good one from a demo. The best arbitrage bot for you is the one that scores well on the factors your own strategy depends on, not the one with the longest feature list.
- Execution, not just detection. This is the whole ballgame. A bot that places both legs captures edges a person never could. A bot that only alerts is a scanner with a nicer notification sound. If a tool can't answer "does it place the bet" with a clear yes, it's not an execution bot.
- Speed. How fast the detect-to-execute loop runs decides which windows you can catch. Direct venue connections in the 20 to 25 millisecond range catch live, in-play arbs. Tools relying on scraped third-party APIs at 500 to 2000 milliseconds miss them. For pre-match arbs with multi-minute windows, speed matters less, so match the tool to the markets you trade.
- Venue coverage. An arb needs two venues that disagree, so more venues means more chances. Look for prediction markets and sharp sportsbooks, not just soft books. Claw Arbs ships with 12 venues pre-installed (Kalshi, Polymarket, Polymarket US, Predict.fun, SX Bet, then Sportmarket, Stake.com, BC.Game, Cloudbet, Bet105, Pinnacle, and BetInAsia) plus calibration for any sportsbook you add yourself.
- Safety controls. Kill switch, naked-leg tracking, per-market cooldowns, exposure caps, and deduplication. Without these, one failed leg can leave you holding an unhedged position that never gets covered. This is the most under-discussed feature in the whole category, and the first thing that separates a serious arb bot from a weekend script.
- Paper trading. Before you fund a venue and fire real orders, run the entire pipeline on live data with orders simulated. Any tool without a paper mode is asking you to trust it with real money on day one, which is a bad ask no matter how good the marketing looks.
The prediction-market angle most arb bots miss
Almost every arb bot on the market covers traditional sportsbooks and stops there. That leaves the most automation-friendly venues in the entire market untouched: prediction markets.
Kalshi is a US-regulated exchange, overseen by the CFTC, where contracts settle in cents from 0 to 100. Polymarket is a crypto prediction market running an on-chain order book, priced 0 to 1 in USDC. Both publish real feeds a bot can connect to directly, which is exactly what you want for low-latency automation. Add Predict.fun and SX Bet and you have four venues that sportsbook-only arb tools ignore completely.
Two things make prediction markets special for an arb bot. First, they don't limit winners the way sportsbooks do. A soft sportsbook that notices you only bet arbs will quietly cut your stakes to pennies. An exchange doesn't care, because you're trading against other users, not against the house. Second, a price on Kalshi and a price on a sharp sportsbook for the same game are often out of sync, which creates cross-venue arbs a sportsbook-only tool can't even see. A bot that speaks both languages, prediction market and sportsbook, has a bigger opportunity set than one that speaks only one.
This is the gap Claw Arbs is built for. It runs an 11-book sharp odds aggregator that blends the sharpest references into one fair-odds line, then grades both prediction-market and sportsbook prices against it. There's a value-betting mode with limit-order sniping for venues that don't natively support limit orders, so you can pre-position at a target price and wait. If you want the detail on a single venue, the Polymarket bot page breaks down that side, and you can watch real edges appear as the bot finds them on our live arbitrage board.
What a sports arbitrage bot costs to run
Price matters, but not the way you'd expect. A free tool that executes correctly beats a $300-a-month tool that only alerts you. Claw Arbs is currently free during its testing period, with over 1,000 users running it live, so it's worth understanding what a genuinely free arbitrage betting bot includes before you pay a subscription for less. Whatever you land on, run it in paper trading mode first and confirm the edges are real before you fund anything. You still pay venue fees and need funded accounts either way. The software cost is the part that varies wildly, and the honest comparison is execution-for-free against alerts-for-a-fee.
FAQ
What is a sports arbitrage bot?
A sports arbitrage bot is software that watches odds across multiple betting venues, spots when two opposing prices guarantee a profit no matter the result, and places both sides of the trade. The best arbitrage bots go one step past detection: they execute both legs automatically, because the window between an edge appearing and closing is often shorter than a person can click through two sites. Claw Arbs runs detection and execution in the same loop, typically in around 20 milliseconds.
Do sports arbitrage bots actually place the bets, or just find them?
Most tools sold as arbitrage software only detect. They send you an alert and leave you to place both bets by hand across two tabs. A true arb bot places both legs itself. This matters because arbitrage windows on live markets frequently close in under a minute, and manual clicking usually takes longer than the edge lasts. Claw Arbs is built around execution: it fires both bets in the same event loop and runs a safety pipeline so one leg never fills without the other.
How fast does a sports arbitrage bot need to be?
It depends on the market. Pre-match arbitrage windows can stay open for several minutes, so speed matters less. Live, in-play windows often close in 5 to 60 seconds, where every hundred milliseconds counts. Bots that pull odds from third-party APIs typically sit at 500 to 2000 milliseconds of delay, while a bot connected directly to venue WebSocket feeds can detect and act in around 20 to 25 milliseconds. For live arbitrage, direct connections win.
Can a sports arbitrage bot trade prediction markets like Kalshi and Polymarket?
Yes, and this is where most arb bots fall short. The majority only cover traditional sportsbooks and ignore prediction markets entirely. Kalshi is a US-regulated exchange (CFTC) and Polymarket is a crypto prediction market, both with public feeds that make them ideal automation targets. Claw Arbs ships with Kalshi, Polymarket, Predict.fun, and SX Bet pre-installed alongside sharp sportsbooks, which opens up cross-venue arbs that sportsbook-only tools never see.
Related reading: Best Arbitrage Betting Software 2026, Automated Arbitrage Execution, Free Arbitrage Betting Bot. Want to see live edges first? Open the live arbitrage board.