# Prediction Market Arbitrage After Fees: Kalshi vs Polymarket Gaps and Multi-Outcome Arbs

> Prediction market arbitrage after fees: break-even gaps for 8 Kalshi, Polymarket, Novig and ProphetX pairings, and when a multi-outcome field under $1 pays.

By Market Math Research · Updated 2026-11-17

Prediction market arbitrage means buying YES on one venue and NO on another for a combined price under $1, so one leg pays $1 whatever happens. Scanners flag these gaps all day between Kalshi and Polymarket. Most of them are not arbitrage. Either the fees on the two legs are larger than the gap, or the two contracts are not the same question. This page gives the break-even gap for eight real venue pairings, shows the arithmetic on a 2¢ gap, covers the single-venue version (a multi-outcome market whose outcomes sum to under or over $1), and lists what a bot has to model before any of it is real. Every fee figure below renders live from our fee registry, last verified November 16, 2026.

## Prediction Market Arbitrage Math: The Break-Even Gap After Fees

Buy YES on venue A at price a and NO on venue B at price b. One of the two contracts settles at $1. The raw gap is g = 1 − a − b. The trade is only an arb when:

**a + b + fee<sub>A</sub>(a) + fee<sub>B</sub>(b) < 1**, which is the same as **g > fee<sub>A</sub>(a) + fee<sub>B</sub>(b)**.

So the minimum raw gap is simply the sum of the two per-contract fees. Most venues price fees as a coefficient × P × (1 − P), which peaks at 50¢. That means gaps near 50¢ need the most room, and gaps near the tails need the least.

Here is what each leg costs per contract on the five venues that matter for US traders. Kalshi rounds its fee up to the cent **per order**, so the 100-contract row and the 1-contract row differ.

| Leg (taker, per contract) | 20¢ | 50¢ | 80¢ |
|---|---|---|---|
| Kalshi, 100-contract order | 1.12¢ | 1.75¢ | 1.12¢ |
| Kalshi, 1-contract order | 2.00¢ | 2.00¢ | 2.00¢ |
| Polymarket, politics | 0.64¢ | 1.00¢ | 0.64¢ |
| Polymarket, sports | 0.80¢ | 1.25¢ | 0.80¢ |
| Polymarket US, Tier 1 | 1.52¢ | 2.38¢ | 1.52¢ |
| Novig, live taker | 0.96¢ | 1.50¢ | 0.96¢ |
| Novig, pre-game | 0.00¢ | 0.00¢ | 0.00¢ |
| ProphetX, only if this leg wins | 1.60¢ | 1.00¢ | 0.40¢ |

Polymarket US is shown at Tier 1, the rate for accounts under $10K of 30-day taker volume. Higher-volume tiers pay less:

| Tier (30-day taker volume) | Θ | Fee on 100 contracts at 50¢ |
|---|---|---|
| Tier 1: Under $10K 30-day taker volume | 0.095 | $2.38 |
| Tier 2: $10K to $50K | 0.085 | $2.13 |
| Tier 3: $50K to $250K | 0.075 | $1.88 |
| Tier 4: $250K or more | 0.0695 | $1.74 |
| New account, days 1 to 7 | 0.0695 | $1.74 |

Now add the two legs. The table below is the **minimum raw gap** for each pairing, with YES on the first venue at the column price and NO on the second venue at the complement (an 80¢ NO for a 20¢ YES, and so on). For ProphetX, the fee only exists in the outcome where the ProphetX leg wins, so the table counts it in that worst case.

| YES venue + NO venue | YES at 20¢ | YES at 50¢ | YES at 80¢ |
|---|---|---|---|
| Kalshi + Polymarket politics | 1.76¢ | 2.75¢ | 1.76¢ |
| Kalshi + Polymarket sports | 1.92¢ | 3.00¢ | 1.92¢ |
| Kalshi (1-contract orders) + Polymarket sports | 2.80¢ | 3.25¢ | 2.80¢ |
| Kalshi + Polymarket US Tier 1 | 2.64¢ | 4.13¢ | 2.64¢ |
| Kalshi + Novig live | 2.08¢ | 3.25¢ | 2.08¢ |
| Kalshi + ProphetX | 1.52¢ | 2.75¢ | 2.72¢ |
| Kalshi + Novig pre-game | 1.12¢ | 1.75¢ | 1.12¢ |
| Novig pre-game + ProphetX | 0.40¢ | 1.00¢ | 1.60¢ |

The table uses the exact complement for the NO price. A real NO leg is cheaper than the complement by the gap itself, which moves its fee by a fraction of a cent. The [arbitrage calculator](https://marketmath.io/tools/arbitrage-calculator) uses the real prices.

### Worked example: a 2¢ Kalshi vs Polymarket gap at 50¢

A scanner shows YES at 50¢ on Kalshi and NO at 48¢ on Polymarket US. That is a 2¢ gap. Buy 100 of each.

- Cost of contracts: $50 + $48 = $98. One leg pays $100. Gross profit: $2.00.
- Kalshi taker fee: roundup(0.07 × 100 × 0.50 × 0.50) = $1.75, paid at the fill.
- Polymarket US Tier 1 fee: 0.095 × 100 × 0.48 × 0.52 = $2.37, also paid at the fill.
- Net in either outcome: $2.00 − $1.75 − $2.37 = **-$2.12**.

That is a guaranteed loss, not a guaranteed profit. Swap the NO leg to Polymarket's global sports book and the result is -$1.00. On a politics market it is -$0.75. No Kalshi plus Polymarket combination makes meaningful money on a 2¢ gap at 50¢; the politics result is the best case, and it is a loss or a rounding error depending on the category rate in force. The Kalshi leg alone uses 1.75¢ of the 2¢ per contract, leaving 0.25¢, and even the cheapest Polymarket categories, politics among them, charge 1.00¢ on a 48¢ NO. For the per-venue fee mechanics, see the [Kalshi fees guide](https://marketmath.io/blog/kalshi-fees-guide-2026) and [Polymarket fees explained](https://marketmath.io/blog/polymarket-fees-explained). Polymarket's global exchange restricts US users, which is why Polymarket US is the realistic NO leg for most readers.

## How Novig and ProphetX Change the Arbitrage Math

Two sports exchanges break the pattern. [Novig](https://marketmath.io/platforms/novig) charges a coefficient of 0 on pre-game straight trades, so a pre-game Novig leg adds nothing to the break-even gap. [ProphetX](https://marketmath.io/platforms/prophetx) charges 2% of net winnings per market at settlement, and $0 on a market you lose. For an arb, that changes the question from "what do both legs cost" to "what does each outcome cost."

ProphetX's charge on a winning contract is 0.02 × (1 − price). It shrinks as the price rises. So the ProphetX leg belongs on the **favorite** side of the pair, where the profit on a win is small.

**Worked example: same 2¢ gap, different NO venue.** YES at 20¢ on Kalshi, NO at 78¢ elsewhere, 100 contracts each. Contract cost is $98, one leg pays $100, and the Kalshi fee is roundup(0.07 × 100 × 0.20 × 0.80) = $1.12 in both outcomes.

| NO leg at 78¢ | NO-leg fee | Profit if YES wins | Profit if NO wins | Arb? |
|---|---|---|---|---|
| Polymarket US Tier 1 | $1.63 | -$0.75 | -$0.75 | No |
| Novig live taker | $1.03 | -$0.15 | -$0.15 | No |
| ProphetX | $0.44 only if NO wins | $0.88 | $0.44 | Yes |
| Novig pre-game | $0.00 | $0.88 | $0.88 | Yes |

The ProphetX row works because the fee is 0.02 × $22 of profit on the winning NO leg, not a charge on the $78 position. Flip it, with ProphetX holding the 20¢ YES and Kalshi the 78¢ NO, and the fee on a ProphetX win becomes $1.60. Add Kalshi's $1.21 and that outcome nets -$0.81.

The cleanest version skips Kalshi entirely. YES at 50¢ on ProphetX, NO at 48¢ pre-game on Novig, 100 contracts each. Novig charges $0.00. If YES wins, ProphetX takes $1.00 and you keep $1.00. If NO wins, you keep $2.00. Same 2¢ gap that lost $2.12 on Kalshi and Polymarket US.

One trap: the Novig advantage ends at kickoff. Live, Novig takers pay 0.06 × P × (1 − P), and the pre-game arb above becomes a loss if the second leg fills after the game starts. Both venues are sports-first, so these pairings only exist on games. The [Novig vs ProphetX comparison](https://marketmath.io/compare/novig-vs-prophetx) covers the rest of their fee schedules.

## Why Most Polymarket Arbitrage Gaps Are Two Different Questions

Fees kill thin gaps. Wide gaps usually die a different way: the two contracts settle on different rules. When you hold YES on one and NO on the other, any rule difference creates an outcome where **both legs lose**. Common mismatches:

- **"By" vs "on."** Venue A asks "Will X happen by June 30?" Venue B asks "Will X be true on June 30?" The "by" contract trades higher because it has more ways to resolve YES. Buy NO on A at 60¢ and YES on B at 30¢ for a "10¢ gap." If X happens in May and reverses before June 30, A resolves YES and B resolves NO. You lose $0.90 per pair.
- **Ties and three-way outcomes.** "Team A wins" on one venue and "Team B wins" on another look like complements. A draw resolves both NO.
- **Overtime rules.** One hockey or soccer contract settles on regulation time, another includes overtime or extra time. A game tied after regulation splits them.
- **Resolution source.** One economic contract settles on the first official release, another on a later revision or a different data series. One weather contract reads a named station, another a different one. One election contract settles on a media call, another on certification.
- **Timing and voids.** One venue settles a postponed game if it is played within a window and voids it otherwise; the other keeps the original rule set. A refund on one leg and a loss on the other is a one-sided bet you did not intend.
- **Who, not what.** "Party X wins" and "Candidate Y wins" diverge the moment the candidate changes.

Read both rulebooks line by line before counting a gap. Market Math's parent company runs [Prediction Genius](https://predictiongenius.io), a cross-venue data engine that matches the same question across venues, and that matching step is where most raw gaps fall out. Only compare exactly matched pairs: same event, same resolution source, same cutoff, same treatment of ties, overtime and postponements. The [sportsbook vs prediction market arbitrage guide](https://marketmath.io/blog/sportsbook-vs-pm-arbitrage) covers the same problem between books and exchanges.

## What a Polymarket Arbitrage Bot or Scanner Has to Model

A Polymarket arbitrage scanner that subtracts nothing but the quoted prices will flag hundreds of false positives. A bot that trades them loses money with perfect execution. Here is the minimum it has to model:

1. **Fees per fill, by venue and category.** Polymarket's taker rate depends on the market's category (and the category table changes on Nov 20, 2026), and makers pay zero and can earn a maker rebate. Kalshi charges takers per fill and nothing at settlement, and charges makers only on maker-fee series, which are mostly sports game and spread series. ProphetX charges on settlement, per market, only on wins.
2. **Kalshi's per-order round-up.** One 100-contract order at 20¢ costs $1.12. One hundred 1-contract orders at 20¢ cost 100 × 2¢ = $2.00. A bot that slices orders small to hide size pays 1.8x the fee. Polymarket US rounds each order to the cent with banker's rounding.
3. **Rate limits.** Both Kalshi and Polymarket cap API requests. A scanner polling hundreds of markets sees stale prices on most of them most of the time, and stale prices produce phantom gaps.
4. **Fill risk on the second leg.** The first fill moves you from hedged to naked until the second leg fills. Partial fills leave unmatched contracts. The bot needs depth at the price, not just a top-of-book quote, and a rule for what to do when the second leg misses. The [liquidity guide](https://marketmath.io/blog/prediction-market-liquidity) covers reading depth.
5. **Capital locked until settlement.** The Kalshi plus ProphetX trade above locks about $99 to make $0.44 in the worst case. Over 30 days, that is 5.4% annualized. Over six months it is a fraction of that. The same raw return means very different things at different lockups: a 6% arb that settles in one week is 312% annualized (6% × 52, simple), while the same 6% on a six-month election contract is 12%. A 3% arb that settles weekly beats an 8% arb that settles quarterly.
6. **Withdrawal rails and split balances.** Profit sits on two venues with different funding methods, minimums and processing times. Novig, for example, has a $10 withdrawal minimum. Moving capital back to the cheaper side takes days, not seconds, so each venue needs its own idle balance for the next gap. Keep most capital where you find the most gaps, hold minimum balances elsewhere, and rebalance after each settlement.

One more point for anyone searching for a Polymarket arbitrage bot that trades within a single market: YES and NO on one Polymarket or Kalshi binary share one order book, so the YES ask plus the NO ask is always at least $1. The in-venue version is the multi-outcome field below, where the YES prices across every outcome can sum below (or above) $1, and every leg pays the taker fee.

## Multi-Outcome Market Arbitrage: When the Outcomes Sum Under or Over $1

A multi-outcome market lists one YES contract per outcome: a championship, an award, a primary field, a range of CPI prints. Exactly one outcome settles at $1, so the YES prices across the field should sum to $1. They rarely do, because each outcome trades in its own order book and nothing forces the total back to $1. Two trades follow from that:

- **Sum under $1: buy YES on every outcome.** The full set costs the sum of the YES asks, S, and pays exactly $1. Raw profit per set is 1 − S.
- **Sum over $1: buy NO on every outcome.** With N outcomes, N − 1 of the NO contracts pay $1, so the set pays N − 1. On Kalshi and Polymarket a NO ask is $1 minus the YES bid, so the NO set costs N minus the sum of the YES bids. Raw profit per set is the YES bids' sum minus $1.

Either way, every leg pays the taker fee at the fill. That is where most multi-outcome arbs die.

### The break-even gap grows with the number of outcomes

On a P × (1 − P) venue, the fee on one full set is coefficient × Σ p(1 − p), which equals coefficient × (S − Σp²). For a field priced near $1, that is roughly coefficient × (1 − Σp²). More outcomes means a smaller Σp² and a bigger fee. A binary pair costs coefficient × 0.50 at 50¢; a 20-way field costs nearly the full coefficient on every dollar of sets. Here is the minimum raw gap per set for evenly priced fields, before Kalshi's per-order round-up (which adds up to 1¢ per leg per order):

| Field | Σ p(1 − p) | Kalshi taker | Polymarket politics | Polymarket sports | Polymarket US Tier 1 |
|---|---|---|---|---|---|
| 2 outcomes at 50¢ (binary, for comparison) | 0.50 | 3.50¢ | 2.00¢ | 2.50¢ | 4.75¢ |
| 4 outcomes at 25¢ | 0.75 | 5.25¢ | 3.00¢ | 3.75¢ | 7.13¢ |
| 5 outcomes at 20¢ | 0.80 | 5.60¢ | 3.20¢ | 4.00¢ | 7.60¢ |
| 10 outcomes at 10¢ | 0.90 | 6.30¢ | 3.60¢ | 4.50¢ | 8.55¢ |
| 20 outcomes at 5¢ | 0.95 | 6.65¢ | 3.80¢ | 4.75¢ | 9.03¢ |

So a ten-way Kalshi field has to sum below about 93.7¢ before a full YES set breaks even. The fee also weighs most on the longshots relative to what you pay: fee as a share of price is coefficient × (1 − P), so a 5¢ Kalshi contract pays 6.65% of its price in fees, against 3.5% at 50¢.

### Worked example: a five-outcome field at 96¢

The YES asks on a five-outcome market are 40¢, 25¢, 15¢, 10¢ and 6¢. They sum to 96¢, a 4¢ gap. Buy 100 of each: $96 of contracts, and one outcome pays $100, for $4.00 gross. Here is the same field priced identically on each venue:

| Venue, all five legs | Fees on 100 full sets | Net in every outcome |
|---|---|---|
| Kalshi taker | $4.93 | -$0.93 |
| Kalshi maker, maker-fee series (all five resting bids fill) | $1.24 | $2.76 |
| Polymarket politics | $2.81 | $1.19 |
| Polymarket sports | $3.51 | $0.49 |
| Polymarket US Tier 1 | $6.66 | -$2.66 |

A 4¢ gap on a five-way field is two full cents wider than the 2¢ binary gap that lost money earlier, and it still loses on Kalshi as a taker. The fee scales with Σ p(1 − p), not with the gap.

### The over-$1 side: NO on the whole field

Now the YES bids on a five-outcome market are 44¢, 27¢, 16¢, 11¢ and 7¢, which sum to $1.05. The NO asks are 56¢, 73¢, 84¢, 89¢ and 93¢, which sum to $3.95. Four of the five NO contracts pay, so 100 sets cost $395 and return $400: $5.00 gross. On Kalshi as a taker, the five fees total $5.21 and the net is -$0.21. On a Polymarket politics market the net is $2.04, which is 0.52% on the $395 tied up. The NO side needs about four times the capital per set of the YES side, so the return on capital is thin before you annualize it over a futures-length lockup.

The over-$1 side is the common one. Each outcome's price carries its own overround, and the excess concentrates at the bottom of the board. Favorites attract most of the volume and trade tight. Contracts at 3¢ to 8¢ sit in thin, stale books, and the favorite-longshot bias means recreational traders overpay for them: an outcome with a true 2% chance can trade at 5¢. Award fields with 8 to 12 nominees and primaries with 6 or more candidates are the usual places to find a field summing well above $1, and right after a candidate drops out or a new outcome is added, the longshots take the longest to reprice.

### Three traps specific to multi-outcome arbs

1. **The field has to be exhaustive.** A full YES set only pays if one listed outcome wins. An unlisted candidate, a tie, a "no winner" result or a cancelled event can leave every YES leg at $0. Read what the rules say happens when none of the listed outcomes occurs.
2. **Every leg has to fill.** Five legs means five books that can move while you work the order, and the longshot books are the thinnest. A partial set is a directional bet on the outcomes you missed.
3. **Normalizing is not an edge.** Dividing each price by the field total gives the market's fair probability: a 45¢ outcome in a field summing to $1.10 normalizes to 0.45 / 1.10 = 40.9%. That tells you the 45¢ YES is about 4.1¢ rich by the market's own view. It does not make a 45¢ YES worth buying when your own estimate is 43%; you still pay 45¢, so the expected value is 0.43 − 0.45 = −2¢ per contract before fees. Overround favors the NO side of the overpriced outcomes, not the YES side.

The [multi-outcome calculator](https://marketmath.io/tools/multi-outcome-calculator) sums the field, shows the overround, and strips it with six de-vig methods (power, Shin, multiplicative, additive, probit and worst case), which disagree most on the longshots. Run the fee on each leg through the [prediction market fee calculator](https://marketmath.io/tools/prediction-market-fee-calculator) before treating any gap as an arb.

## Kalshi Arbitrage Calculator Check: 6 Steps Before You Place One

**Prediction market arbitrage check**

1. Confirm the contracts match
2. Pull executable prices and depth
3. Add both fees per fill
4. Check the worst-case outcome
5. Annualize over the lockup
6. Fill the thinner leg first

1. **Match the question.** Same event, resolution source, deadline, tie and overtime rules, and void policy. If any differ, stop.
2. **Use executable prices.** Take the ask on both legs at your full size, not the last trade or the midpoint.
3. **Add both fees.** Enter both legs in the [arbitrage calculator](https://marketmath.io/tools/arbitrage-calculator); it nets the fees on Kalshi, Polymarket (global and US), Novig (pre-game, live or futures), ProphetX (charged only if that leg wins) and Underdog. For any other venue, use the Custom setting.
4. **Check the worst outcome.** With a ProphetX leg, profit differs by outcome. The arb is the smaller number.
5. **Annualize.** Divide the worst-case profit by capital deployed and scale by days to settlement. Compare it with the [turnover calculator](https://marketmath.io/tools/turnover-calculator) before tying up the money.
6. **Fill the thinner leg first.** Then fire the second leg immediately, and decide in advance what you do if it misses.

If the gap is under the number in the break-even table, there is nothing to check. For arbs where one leg is a sportsbook, read [sportsbook vs prediction market arbitrage](https://marketmath.io/blog/sportsbook-vs-pm-arbitrage), and for fee mechanics on every venue see the [fees page](https://marketmath.io/fees).

## Sources

- [Kalshi fee schedule (PDF)](https://kalshi.com/docs/kalshi-fee-schedule.pdf)
- [Kalshi series API: fee_type and fee_multiplier per series](https://api.elections.kalshi.com/trade-api/v2/series)
- [Polymarket trading fees documentation](https://docs.polymarket.com/trading/fees)
- [Polymarket US fees (effective Nov 14, 2026)](https://docs.polymarket.us/fees)
- [Fees on Novig (Novig Help Center)](https://support.novig.com/en/articles/16195057-fees-on-novig)
- [ProphetX trading fee schedule (CFTC filing, Aug 2026)](https://www.cftc.gov/filings/orgrules/rules08152617271.pdf)

## Frequently asked questions

**Is prediction market arbitrage between Kalshi and Polymarket real?**

Sometimes, but rarely at small gaps. Both venues charge takers a fee shaped like P × (1 − P), so a gap near 50¢ has to be wider than the two fees combined before it pays. A 2¢ gap at 50¢ loses money on every Kalshi and Polymarket pairing. Check the break-even table on this page and run real prices through the arbitrage calculator.

**How big does a Kalshi arbitrage gap need to be?**

At least the Kalshi fee plus the other venue's fee, per contract, at the prices you will actually fill. The requirement is largest at 50¢ and smallest near the tails. Kalshi rounds its fee up per order, so many small orders need a wider gap than one large order.

**Does a Polymarket arbitrage bot make money?**

Only if it models fees per fill by category, order rounding, rate limits, second-leg fill risk, capital locked until settlement and withdrawal times, and only if it trades exactly matched contracts. A bot that compares raw prices flags gaps that are either eaten by fees or are two different questions.

**Why are ProphetX and Novig useful for arbitrage?**

Novig charges no fee on pre-game straight trades, and ProphetX charges only on a winning market at settlement, as a share of net winnings. Pairing either one with Kalshi or Polymarket lowers the gap needed, especially with ProphetX on the favorite side. Current rates are on the /fees page.

**What is the biggest risk in a cross-venue prediction market arb?**

Settlement rules that differ between the two contracts. If one resolves on a different source, date or overtime rule, there is an outcome where both legs lose. Read both rulebooks and only trade exactly matched pairs.

**What is multi-outcome market arbitrage?**

When the YES prices across every outcome of one market sum to under $1, buying the full set locks in $1 for less than $1. When they sum to over $1, buying NO on every outcome does the same thing in reverse. Every leg pays the taker fee, and the combined fee grows with the number of outcomes, so the gap has to be wider than on a binary market.

**What is overround in a multi-outcome market?**

Overround is the amount the outcome prices sum to above $1. A field summing to $1.10 has a 10% overround. It concentrates on longshots priced at 3¢ to 8¢, where books are thin and the favorite-longshot bias pushes prices above true probability. The multi-outcome calculator strips it with six de-vig methods.

---

Source: Market Math (Market Math Research), fees verified November 16, 2026. https://marketmath.io/blog/prediction-market-arbitrage-after-fees
