No — prediction market arbitrage between Polymarket and Kalshi is not drying up in 2026. What's changed is who captures it. As more traders and bots watch the same two order books, the easiest, widest mispricings get closed faster, which compresses the average spread size. But the structural reasons the gap exists in the first place haven't gone anywhere, and new markets open every day carrying fresh opportunities for divergence.
This is a different question from "how much can you make" — that depends on your capital and execution speed. This is about the supply of edge itself: is there still a real, persistent gap to capture, or has the market become too efficient for it to matter?
Why the "is it over" question keeps coming up
Both Polymarket and Kalshi have grown enormously since 2024, and with that growth has come more attention from traders looking for a market-neutral edge. More participants watching the same price gaps is a natural, expected phase in the life of any electronic market — it's the same evolution that happened in equities, then crypto, and it's now happening here. The instinct is to assume that once enough people are looking, the opportunity disappears entirely.
That instinct is usually wrong, for a simple reason: arbitrage between two independently operated platforms doesn't get arbitraged away the same way it does within a single order book. Closing the gap requires someone to actually place both legs, on two different platforms, fast enough. Competition makes that harder to do manually — it doesn't remove the underlying price divergence.
What's actually happening to spread sizes
Gross arbitrage edges on Polymarket/Kalshi pairs still typically fall in the 1.5% to 6% range per trade, with net edges (after platform fees) landing around 0.5% to 5%. Most day-to-day opportunities sit toward the lower end of that band — 1.5% to 3% gross — while high-activity events like elections and Fed decisions still produce the wider 3% to 8% spikes.
What's shifted isn't the ceiling, it's the distribution. Wide, obvious mispricings on heavily-traded markets get closed within seconds by whoever is watching. The edge that remains for a given trader depends heavily on how much of the market they're actually covering and how fast they're detecting it — not on whether arbitrage as a category still exists.
The structural reasons the gap can't fully close
Polymarket and Kalshi aren't the same market with a temporary pricing glitch between them. They're structurally different platforms: separate liquidity pools, different user bases (Polymarket skews more international and crypto-native, Kalshi more US-based), different fee structures, and occasionally different resolution criteria for the same underlying event. None of those differences resolve themselves — they persist as long as the two platforms operate independently, which is the entire reason the arbitrage exists at all.
New event contracts launch on both platforms constantly. Every new market is a new instance of the same structural gap, which means the supply of potential arbitrage doesn't deplete the way a single mispriced trade does once it's closed.
Where the edge concentrates now
The clearest pattern in 2026 is concentration around news-driven volatility. Breaking news, a large directional order, or a resolution update all move price on one platform before the other catches up — and that's precisely when spreads widen past the typical 1.5–3% baseline. Quiet, low-attention markets still diverge, but the spreads tend to be thinner and the fills smaller, constrained by order book depth on both sides.
The competitive squeeze: bots vs. manual traders
Execution speed is what actually decides who captures a given spread. Arbitrage windows on active markets still last roughly 30–200 seconds, but a growing share of that window is now being watched by automated agents rather than humans with two browser tabs open. That's the real shift: not "less arbitrage exists," but "manual traders are competing against detection systems that never blink." The opportunity didn't shrink — access to it did, for anyone still doing this by hand.
What actually threatens the opportunity — and what doesn't
Competition compresses individual spreads, but it isn't the binding constraint most traders should worry about. The more common failure modes are the same ones that mattered before competition intensified: event matching errors (treating two similar-but-different contracts as the same trade), thin order book depth that eats the edge on fills, and fee drag on thin spreads. A well-matched, correctly-sized trade at 1.5% net edge is still a good trade in 2026 — the risk isn't that arbitrage stopped working, it's executing badly on the opportunities that remain.
Realistic outlook for 2026 and beyond
Don't expect the wide, easy spreads that a smaller, less-watched market once offered — those get closed in seconds now. Do expect a steady, ongoing supply of smaller opportunities as long as Polymarket and Kalshi remain separately liquid platforms with different users and fee schedules, refreshed continuously by every new market that launches on both. The honest read is that this is a maturing, not a dying, opportunity — and it increasingly favors whoever is watching the most markets, the fastest. Join the Arbitrage Agent waitlist to get alerted the moment a net-positive spread is confirmed, across thousands of matched market pairs, without watching either platform yourself.