Polymarket vs. Kalshi Arbitrage — What It Is, How to Spot It, and Who Actually Captures It

What “arbitrage” means here

A YES contract and a NO contract on the same question always resolve to exactly $1 total. If you can buy YES on Venue A for x and NO on Venue B for y, and x + y < 1, you have captured structural edge of 1 − (x + y). That's the pure theory.

The reality has eight friction layers, and the naive calculation above ignores all of them.

Why the spreads exist

Polymarket and Kalshi are fundamentally different venues. Polymarket runs on Polygon (USDC, on-chain, crypto-native users), Kalshi is a CFTC-regulated exchange (USD wire/ACH, KYC'd retail and institutions). The pools of capital rarely overlap:

• A Polymarket whale can't always get KYC'd on Kalshi — some are non-US, some are pseudonymous on purpose.
• A Kalshi retail trader often has never used a crypto wallet in their life.
• Capital can't move between the two in seconds: bridging USDC out and ACH-ing into Kalshi is a multi-hour-to-multi-day loop.

These frictions are why the spreads can exist. If both venues shared the same pool of capital and instant transfers, the arbitrage would close in seconds.

How to compute real edge

Start with the quoted mid-prices, then subtract:

1. Bid/ask spread. You are a taker, not a maker. On a thin market the quoted 52¢ is a mid — actual fill is 53−54¢.
2. Kalshi trading fee. Maker is free; taker is a flat per-contract fee that scales non-linearly near 50¢. Round-trip on a $1,000 notional at 50¢ is not trivial.
3. Polymarket gas. Small but non-zero, in MATIC. Negligible on $1K trades, meaningful on $50 trades.
4. Slippage. Prediction-market orderbooks are thin. Moving $20K through a market that has $3K on the best ask will rip through four levels and cost you 3–10 cents depending on liquidity.
5. Time-to-resolution carry. Capital locked in a market that resolves in 6 months is capital you can't deploy elsewhere. If you could earn 4% in T-bills, a 6-month lock-up costs 2% in opportunity cost. 2% is often more than the headline edge.
6. Resolution risk. The two venues sometimes have subtly different resolution criteria. “Will X happen before Dec 31” vs. “Will X happen in 2026” can seem identical and resolve differently.

After all six, the real edge on most headline “arbitrage” opportunities is 0 to −50 bps. Pros do this at scale; retail mostly doesn't make it through the math.

Who actually captures it

Three groups consistently run this strategy profitably:

1. Market makers on both venues. They earn the bid/ask spread regardless; arbitrage is a bonus. They run their own liquidity on both sides.
2. Quant funds with KYC'd accounts and dedicated USDC/USD rails. A handful of crypto-native quants have streamlined the bridge so capital rotation is hours, not days. With 8-figure size they can absorb the frictions above.
3. Operators of automated bots that front-run obvious mispricings. Once a 1¢ spread appears on a liquid market, it closes in under a minute. You can't manually catch those.

Where retail can still find real edge

The headline bid/ask arbitrage is generally gone. What's still available:

• Wide long-dated spreads. Markets that resolve 6–12 months out are thinner and less efficient. If you're willing to lock capital, the bps per month is sometimes competitive with fixed income.
• Narrative mispricings. Polymarket is often first to price news; Kalshi can lag by 15–60 minutes on political or breaking-story markets. Humans move faster than Kalshi's KYC'd liquidity pool for very specific event types.
• Resolution-criteria differences. Genuine differences, not ones people already noticed. These require reading the market rules carefully on both sides.

How the PolySharks arbitrage feed works

Our arbitrage dashboard does the matching + edge calculation automatically. We join Polymarket and Kalshi markets by entity-keyword overlap (names, dates, event types), pull orderbook mids on both sides, and surface the mid-price delta after estimated fees.

One thing we are honest about: the feed shows potential spreads, not confirmed fills. Actual capture requires you to still do the slippage/fee/carry math on your own size. Free tier shows 3 opportunities; Pro unlocks the full list.