Kalshi vs Polymarket: The Battle for Prediction Markets

We run against both feeds. Here's what a year of side-by-side data actually shows, updated for the Polymarket US era.

GhostLeg, 2026-05-30, 11 min

Most comparisons of Kalshi and Polymarket are written by journalists who have accounts on neither, or by crypto advocates who use one and have never touched the other. What you almost never get is a perspective from someone operating against both feeds simultaneously, on real sports markets, over a meaningful sample.

Kalshi vs Polymarket is the question sitting at the top of prediction-market search results right now, and the answer depends entirely on what you're trying to do. GhostLeg ingests both Kalshi and Polymarket data as live market signals. The numbers below cover the 90 days from early May through August 1, 2026, across MLB, NBA, WNBA, NHL, UFC, MLS, EPL, and a World Cup that turned out to be the biggest stress test prediction markets have ever had on sports. That's the lens this comparison comes through.

Updated August 27, 2026: Polymarket's regulated US exchange is now live (details below), Kalshi's fee mechanics are corrected against its current published schedule, and a new full-history accuracy ledger draws on our backfilled feeds, including a complete 2025 NFL season. The 90-day tables retain their original window, May 3 through August 1, 2026. An earlier refresh on August 1, 2026 re-pulled all data through our upgraded measurement pipeline (cleaner historical prices on both venues; divergence measured on like-for-like game-winner markets only).

Kalshi vs Polymarket: What They Actually Are

Both platforms let you take positions on real-world outcomes. But they were built from different starting points, and those origins shape everything about how they behave.

Kalshi is a US-regulated exchange with full CFTC oversight. It launched in 2021 and has steadily expanded into sports markets alongside political and economic events. Because Kalshi operates under federal regulation, it can be used legally by US residents in most states. Its structure looks more like a traditional exchange: contracts trade between $0.00 and $1.00, where price equals implied probability, and the platform charges a small trading fee at execution rather than baking a spread into the line.

Polymarket runs on crypto rails (Polygon blockchain), and after years as an offshore-only venue it is now US-accessible: Polymarket acquired the CFTC-licensed exchange QCEX in 2025, launched a regulated US exchange on December 2, 2025, and opened it to all US users (with identity verification) when the waitlist came off in May 2026. A handful of states are contesting that access (Minnesota's prediction-market ban took effect August 1, 2026 and is itself being challenged), so state-by-state friction remains. The global on-chain venue attracted enormous volume during the 2024 election cycle and has deep liquidity on political markets. Its sports coverage expanded significantly in 2025, and it approaches sports differently than Kalshi does: more player props and scenario-based markets alongside the standard moneyline-equivalent contracts. One structural feature matters enormously and gets almost no coverage: Polymarket settles on a public blockchain, so every wallet's positions are visible. That transparency is what makes wallet-level whale tracking possible on Polymarket and structurally impossible on Kalshi.

For a grounding in how prediction market prices relate to traditional sports betting lines, our post on prediction markets vs betting markets covers the conversion math in detail: moneylines to implied probability, vig structures, and why a $0.60 Polymarket contract and a -150 moneyline are saying the same thing.

The Fee Question Everyone Gets Wrong

The standard Kalshi vs Polymarket comparison leads with fees, and usually gets them subtly wrong by treating both platforms the same.

Kalshi's fee structure: Kalshi charges a trading fee at execution, computed per contract as roughly 0.07 times price times (1 minus price), rounded up to the cent and capped per contract; makers pay about a quarter of the taker rate. (Earlier versions of this post described it as a fee on winnings; that's wrong under Kalshi's published schedule, and it matters: the fee is paid win or lose.) The formula peaks at 50/50 contracts, where a taker pays 1.75 cents on a 50-cent contract, and falls toward zero at extreme prices. That parabolic shape is still structurally different from the traditional -110/-110 sportsbook model: at a coin-flip price the effective friction is around 1.75% of the dollar at stake, well under the ~4.5% vig baked into a standard two-way line.

Polymarket's fee structure: Polymarket charges a 2% fee on market-maker orders and is free for takers in most markets. Because it operates on a CLOB (central limit order book) model, the effective cost depends heavily on the bid-ask spread in any given market. Liquid markets trade with tight spreads. Thinner markets (niche props, smaller sports) have wider spreads that can add real friction.

How this compares to sportsbooks: A standard -110/-110 line at a sportsbook implies 4.5% vig. Both sides add up to 104.8% implied probability; that 4.8% is the house take. At Kalshi or Polymarket on a liquid market, you're typically operating in the 1-3% effective cost range. Over hundreds of bets, that's a material difference. Our deep-dive on the hidden tax in every bet walks through exactly how this math compounds. The short version: prediction market friction is structurally lower than sportsbook friction on most markets.

90 Days of Side-by-Side Data: What GhostLeg's Feeds Show

This is the part no third-party comparison can give you. The numbers come directly from our production feeds, aggregated over the 90 days ending August 1, 2026, matched to real games. The divergence column measures how far each platform's closing game-winner probability sits from the sportsbook consensus at the same moment, on the same games, in percentage points. (Measuring on game-winner markets only matters: earlier versions of this comparison mixed props and derivative markets into the divergence math, which exaggerated the gaps.)

Kalshi, May 3 to August 1, 2026

Polymarket, May 3 to August 1, 2026

Three things jump out of these tables.

First, on the biggest two-way markets, the three venues have converged. MLB is the flagship: Kalshi and Polymarket both sit 1.9pp from sportsbook consensus at close, on thousands of games. NBA reads the same at 1.6pp on both. Sharp money flows between all three pools now, and on liquid US team sports the prices are nearly interchangeable. That convergence is itself news: a year ago the conventional wisdom was that prediction market sports prices were noisy approximations of the books. On baseball and basketball, that's over.

Second, the World Cup was a monster. Nearly $6.9B in combined matched sports volume in 90 days, with Kalshi taking more than twice Polymarket's share. Tournament soccer is where event exchanges shine: one-off high-attention games, global participation, and three-way outcomes that map naturally onto event contracts.

Third, the divergence that remains is concentrated where participation thins out. Polymarket's EPL gap (17.4pp on a small sample) and both venues' MLS and NHL gaps are where market structure and participant mix still produce real price disagreements with the books. Those gaps are not noise; they're the raw material for divergence signals.

Track both feeds live in the GhostLeg intel dashboard and pull the historical data programmatically through the Data API. The prediction market endpoints expose both Kalshi and Polymarket snapshots per game, per sport, and per market type.

Who Was Right: Accuracy on Graded Markets

Divergence tells you the venues disagree. Grading tells you who was right. We grade every matched game-winner market after the final score: did the venue's closing price have the eventual winner above 50%?

On the big samples, everyone is basically tied: MLB within half a point across all three venues. That's what an efficient market looks like, and it's consistent with everything we know about closing lines being hard to beat.

The World Cup row is the eye-catcher. On graded tournament games, both prediction markets outperformed the sportsbook consensus by a wide margin, Kalshi by nearly 13 points. Part of that is structural: three-way soccer markets with draws are a different grading problem than two-way US sports, and event contracts handle the draw case more cleanly than a moneyline consensus does. But part of it looks like genuine crowd wisdom on a global event where the prediction market participant base was unusually deep and motivated. Soccer generally repeats the pattern in miniature: on MLS, both venues graded meaningfully better than the books over the window.

One honest caveat cuts the other way: several rows above are small samples from offseason or short windows, and 90 days is one summer, not a verdict. The MLB row, with more than two thousand graded games, is the number to trust most, and it says: parity.

The Longer Ledger: Full-History Accuracy

The 90-day window above is one summer. Since publishing it, we backfilled both feeds through our corrected measurement pipeline, and the full matched history now reaches back to April 2025 on Kalshi. That buys three things the summer window couldn't: full NBA and NHL seasons, samples three to four times larger, and a complete 2025 NFL season. Same grading question as above, all matched game-winner markets, as of August 27, 2026:

One methodological note before reading it: the windows are not symmetric. Kalshi's clean history extends back to spring 2025; Polymarket's corrected history begins in March 2026, when our current sportsbook-consensus measurement era starts. So the Polymarket column covers a shorter, more recent stretch, and the books column is graded on the Kalshi-matched games. Cross-venue comparisons are directional, not same-window.

What the longer ledger says:

The parity verdict survives, and sharpens. Over eight thousand graded MLB games, Kalshi sits at 55.4% and the books at 55.8%; the summer window's half-point Kalshi edge washes out to a half-point books edge over the full ledger. Same story on NBA (books ahead 1.5 points over a full season) and NHL (books ahead 1.9). On the biggest, most liquid US sports, nobody beats anybody: that's a market doing its job, on samples big enough to mean it.

The soccer pattern is not a summer fluke. Kalshi graded ahead of the books on MLS by 6.2 points over 1,348 games and on EPL by 4.7 points over 696, spanning a full year, and Polymarket's MLS reading agrees. The crowd-wisdom edge on lower-liquidity soccer that the World Cup made vivid shows up across the whole history.

And the NFL row is the one to bookmark. Kalshi carried a complete 2025 NFL season in our matched feeds: 550 graded games, roughly $4.5B in matched volume, average closing divergence of 3.7 points from the books, and 66.4% graded accuracy against the books' 65.7% on the same games. Polymarket's clean history starts after that season ended, so 2026 will be its first NFL season in this ledger, and the first NFL season of the US-access era for both venues. Kickoff is two weeks away; both feeds are live on the intel dashboard.

Where Each Platform Wins for Sports Bettors

Kalshi wins on:

• The regulatory head start. Kalshi has operated under full CFTC oversight since 2021, with the account infrastructure, state coverage, and legal track record that five years buys. Polymarket's regulated US exchange only opened to everyone in May 2026, and a handful of states are still contesting prediction-market access for both venues. Access is no longer Kalshi's exclusive, but maturity still is.
• Raw sports volume. Roughly $13.1B in matched sports volume across the window, more than triple Polymarket's, with MLB depth (25,000+ markets) that supports real position sizes.
• Big-event capacity. The World Cup numbers settled the question of whether a regulated US exchange can host institutional-scale sports volume. $4.79B in 90 days says yes.
• UFC. Kalshi carried $446M in matched UFC volume in the window; Polymarket currently has no matched UFC game coverage in our feeds. If fight markets are your focus, this one is no longer close.

Polymarket wins on:

• On-chain transparency. Every wallet's positions are public, which enables whale-wallet tracking: following specific proven-sharp accounts and grading their records. Nothing like this can exist on Kalshi's private accounts. For signal hunters, this is Polymarket's killer feature.
• Market variety. Player props and scenario markets that don't exist in the same form at Kalshi. If you're looking for lines the sportsbooks don't hang, Polymarket's menu is wider.
• International soccer participation. Polymarket's EPL volume ran roughly 7x Kalshi's in the window, and its globally distributed participant base is exactly why its soccer prices sometimes diverge hardest from US sportsbook consensus, in both the risky and the informative sense.
• Lower friction on liquid markets. The 2% taker structure and tight CLOB spreads on busy markets keep effective costs competitive.

The honest trade-off: Kalshi is the mature, convergent, high-volume platform. Polymarket is the transparent, divergent, globally-participated platform with a wider menu, now reachable from the US through its regulated exchange. For the first time they are genuinely competing for the same US user, which makes the differences that remain (volume depth, market variety, on-chain transparency) the whole ballgame.

Prediction Markets as Sharp-Money Signals

This is where the GhostLeg angle becomes most relevant. We don't just consume these feeds for the contracts themselves. We use them as leading indicators.

When Kalshi's or Polymarket's closing probability diverges meaningfully from our sportsbook consensus on the same game, it surfaces as a divergence signal in the intel dashboard. (Since July 2026 those divergence figures are computed against the no-vig sportsbook probability, so a gap means genuine disagreement, not the vig showing up as a phantom edge.) When large Polymarket wallets pile onto a side hours before a game, the whale tracker shows it next to the line. And when line movement at the books runs against public betting at the same time prediction markets lean the same way, you're looking at reverse line movement with independent confirmation.

Prediction markets are, in theory, the purest expression of market-clearing probability: no vig motivation to shade a line, no public-bias pressure. In practice, they carry their own distortions (liquidity gaps, participant-base differences, resolution-criteria ambiguity). The signal is real but it requires context, which is exactly what the cross-market comparison layer in our stack is built to provide. All of it is available programmatically: GhostLeg's Data API exposes the Kalshi and Polymarket endpoints with per-sport, per-market-type granularity, the same layer that powers our dashboards.

The Regulatory Wildcard

No Kalshi vs Polymarket comparison is complete without acknowledging the asymmetric regulatory risk.

Kalshi's CFTC-regulated status is both its biggest structural advantage and its biggest single-point-of-failure. Regulators have historically been unpredictable about what constitutes a "commodity" contract, and sports outcomes have been legally contested as a market category. Kalshi fought a significant legal battle to offer event contracts on congressional elections and won, which strengthened its legal footing. But the category isn't permanently settled.

Polymarket resolved its own access question the expensive way: it bought one. The 2025 acquisition of QCEX, a CFTC-licensed exchange and clearinghouse, gave it a regulated US on-ramp that launched in December 2025 and fully opened in May 2026. Its global on-chain venue still runs in parallel, and that participant base still skews international, which is part of why its soccer coverage is deep and its US-sport divergences have historically run wider. Whether the new US participant mix pulls its US-sport prices tighter to the books is one of the questions the 2026 NFL season will answer in our feeds.

The battleground has shifted from federal permission to the states. Both venues now operate under CFTC designations, and both face state-level pushback: Minnesota's ban took effect in August 2026 and is being challenged, and roughly a dozen states have issued cease-and-desist orders in some form. Both platforms grew substantially after prediction markets demonstrated forecasting accuracy on the 2024 election cycle, and the 2026 World Cup did for sports what that election did for politics: proved the category at scale. The landscape in 2026 looks more favorable than it did in 2023, but it is not static.

They Price Elections and the Weather Too

Everything above about sports carries into the venues' original territory: politics. And unlike liquid US team sports, where the two venues now close nearly identical, the 2026 midterm markets show real gaps. As of August 7, Kalshi priced Republicans to keep the House 25.7 points higher than Polymarket did, and the two venues named different favorites outright for Senate control.

GhostLeg now tracks both venues' election prices side by side on the politics board, recomputing the divergence hourly across 138 races. The full breakdown of the current gaps, and what genuinely explains them, is in Kalshi vs Polymarket election odds: same race, two prices.

Both venues have also moved into weather: daily contracts on high temperatures in US cities, settling against official National Weather Service climate reports. The read-the-contract discipline matters there too, because the two venues can settle the same city against different physical weather stations, which makes some apparent price gaps definitional rather than real. GhostLeg's weather board tracks every open contract on both venues against a forecast ensemble and grades every settlement against the official record.

FAQ

Is Kalshi better than Polymarket? For regulatory maturity, raw sports volume, and NFL/UFC depth, Kalshi. For market variety, on-chain transparency (including wallet tracking), and international soccer depth, Polymarket. Both are now US-accessible; on liquid US team sports the closing prices are nearly identical, so the choice comes down to features and market menu rather than price quality.

Which is more accurate, Kalshi or Polymarket? Effectively tied on big samples, and neither reliably beats the books: over our full graded ledger, MLB reads Kalshi 55.4% vs sportsbooks 55.8% on more than eight thousand games. Where the venues do separate from the books is lower-liquidity soccer (Kalshi ahead of books by 4 to 6 points on EPL and MLS over a year) and tournament events (World Cup: Kalshi 80.2% vs books' 67.6%).

Can US residents use Polymarket? Yes, since December 2025: Polymarket operates a CFTC-regulated US exchange (via its QCEX acquisition), open to all US users with identity verification since May 2026, with a few states contesting access. The global on-chain venue remains separate; its public wallet data is the signal layer GhostLeg tracks.

Do prediction markets have better odds than sportsbooks? Often, on liquid markets: effective friction of roughly 1-3% versus the ~4.5% vig in a standard -110/-110 line. On thin markets, wide spreads can erase that advantage. See our vig explainer for the math.

Which One Should You Use?

Neither platform is the obvious choice for every use case.

If you are a US bettor looking for a regulated, lower-vig alternative to standard sportsbooks on mainstream sports: for the first time you have two answers. Kalshi is the deeper and more battle-tested one, with the volume, the NFL and UFC coverage, and five years of regulated operation; Polymarket's US exchange is the newer door into the venue with the wider market menu. The MLB convergence data alone makes either worth monitoring for any bettor who regularly plays baseball markets.

If you want the richest signal surface rather than a place to trade: Polymarket's transparency is unmatched. Public wallets on its global on-chain venue mean you can follow proven-sharp money directly, which is exactly what our whale tracker does all day.

If you are operating at the infrastructure level (running models, backtesting signals, building licensed data products), you need both feeds. The divergence between them is as informative as either feed in isolation. That's the architecture the GhostLeg intel dashboard is built on, and the same data is available through our Data API for teams building their own systems.

The prediction market space is two or three years from the structure that will define the next decade. This summer's World Cup showed what the ceiling looks like: billions in matched sports volume, prices that held up against the sharpest books in the world, and two venues getting there by completely different routes.

Watch both. The market that moves first is often the one telling the real story.

For entertainment purposes only. Past performance does not indicate future results.