Polymarket combos explained: how parlays fill, pay, and settle on chain

In the week of July 13, 2026, 21,010 wallets bought a Polymarket combo. The next week 5,753 did, and no week since has reached 7,500. The money fell far less: $4.31 million was staked that July week, and about $3.7 million in a typical week through September. Fewer buyers stayed, and they bet more each.

The wallets that left were mostly one-time buyers. Wallets with a single combo buy in the week fell from 9,846 to 1,984, while the median stake held near $10. Polymarket's changelog names no combo change that week, so the record shows who left and not why.

A combo is one position across several Polymarket markets. Each pick is a leg, and a share pays $1 only if every leg resolves your way. One miss pays $0. Sportsbooks call it a parlay. Through October 4, the combos feed had recorded 89,566 wallets buying 686,693 different combos in 1,229,063 taker buys, for $68.8 million staked.

You do not buy a combo off an order book. Polymarket prices it by request for quote: your request goes to market makers, who have 400 milliseconds to answer with a price. You get the best quote and 10 seconds to accept it, and a maker with Last Look gets one more second to confirm. The price is one maker's offer for the whole set, not a formula over the leg prices.

One winning share redeems for $1. So the stake is the collateral you put in, the payout is the number of shares you get, and the price is the ratio between them. A fill at 45.5¢ turns $14,602.21 into 32,081.08 shares, which pays 2.197 times the stake. A fill at 6.6¢ turns $2,675.08 into 40,525.40 shares, which pays 15.15 times the stake. Divide $1 by the price to get the multiplier.

There is no parlay table. On a sportsbook a 3-leg parlay pays a fixed number the book chose. On Polymarket the multiplier comes from the quote each request gets, and the feed prints the exact multiplier each fill got. 2 wallets buying the same combo an hour apart can get different multipliers, because each got its own quote.

Most of the money sits close to even. Of the 144,900 taker buys of $50 or more on record, 51.2% pay between 2x and 5x and 24.9% pay under 2x. The long shots are rare: 7.0% pay 10x to 25x, 2.1% pay 25x to 100x, and 99 buys in the whole record, 0.1%, pay 100x or more.

On September 4 at 4:39pm UTC, the wallet DegenKingBetter bought a 2-pick combo: the Los Angeles Dodgers to beat the Washington Nationals and the Milwaukee Brewers to beat the Cincinnati Reds, both that day. The stake was $14,602.21 at 45.5¢ a share, for 32,081.08 shares. The combos page shows it as $14,602.21 to win $32,081.08, a 2.197x chip beside those numbers, and Price 45.5¢ on the line under them.

The order filled against 15 maker orders in a single block, the largest for 26,794.19 shares, so makers were willing to take the other side of that pair in size. Most combo matches find one maker.

Both teams won. The combo settled Won, and the 32,081.08 shares redeemed for $32,081.08, a gain of $17,478.87 on the position before any fee. Every fill of that combo, including the later $683.72 buy that followed it, is on the combo's own page. The wallet carries an A grade on its profile, set by the profit it has booked across its record rather than by this one row.

2 days earlier, on September 2 at 6:24pm UTC, a wallet named mazalMONSTER staked $2,675.08 on 5 MLB games at once: the Cleveland Guardians over the Toronto Blue Jays, the Texas Rangers over the Athletics, the Pittsburgh Pirates over the San Francisco Giants, the Arizona Diamondbacks over the Philadelphia Phillies, and the Tampa Bay Rays over the New York Mets. The fill price was 6.6¢, so the position was 40,525.40 shares, a $40,525.40 payout if all 5 came in.

The Diamondbacks won. The other 4 lost. 1 leg short pays the same as 5 legs short, and the position redeemed for $0. That is the largest payout anyone chased in the week ending September 6, and it sits in the feed next to the rows that hit.

Both examples work the same way. A combo is a single bet on a conjunction. The price tells you what the book thinks of that conjunction, the legs tell you what has to happen, and settlement is binary.

Most lost combos are dead long before their last game. In a sample of 8,959 lost combos, 56% still had a leg that had not settled 3 hours after the first losing leg did. Whatever those games did next, the ticket was already worth $0. A combo can drop to nothing at any leg and only reaches $1 at the last one, so watch the first leg to settle, not the last.

If every leg resolves the way the combo picked, the YES side pays $1 a share. If any leg misses, the NO side pays. Each leg settles with its own market, so a combo across an afternoon game and an evening game settles in steps. Polymarket's docs say a partly settled position can be compressed into one that keeps only the open legs and pays out what the settled legs decided.

The feed shows 3 states. Open means no leg has lost and at least one is undecided. Won means every leg won. Lost goes up the moment one leg loses, even with games still to play, because nothing after it can change the result.

Polymarket has not published what a combo does when a leg is voided, for example a postponed game with a 50/50 payout. The feed does not guess: those combos print the reason next to their status.

You do not have to hold to settlement. A quote request can sell the YES side as well as buy it, so you can exit a combo that is halfway home at the best price a maker quotes. The feed shows a sale as shares sold for a dollar amount.

2-leg combos are 32.1% of every combo bought and 3-leg combos are 23.0%, so more than half of the ledger is a pair or a triple. 4 legs are 16.4%, 5 legs 9.1%, 6 legs 6.2%, 7 legs 3.4%, and 8 or more legs 9.8%. The longest combo bought so far has 38 legs, and it lost.

The dollars are more concentrated than the count. 2-leg combos hold $31,973,363 of the $54,376,768 staked on combos, which is 58.8%, and 3-leg combos hold another $12,244,186. Everything with 5 legs or more adds up to $4,567,722, which is 8.4% of the money on 28.5% of the combos.

Across the 1,006,027 settled buys on record, 23.7% won. Split by leg count the curve is steep. 2-leg buys won 34.3% of the time. 3 legs, 21.7%. 4 legs, 14.4%. 5 legs, 10.9%. 6 legs, 8.4%. 7 legs, 7.7%. 8 legs or more, 3.7%.

That is what multiplication predicts. If each leg is a 60% favorite and the legs are independent, 2 legs hit 36% of the time, 3 hit 21.6%, 4 hit 13.0%, and 5 hit 7.8%. The record tracks that arithmetic closely. Most combos are stacks of favorites, and a 5-leg stack of favorites is a long shot however safe each pick feels.

A win rate on its own says nothing about money. A 3.7% win rate is fine at 40x and ruinous at 10x. The next chart puts the two together.

For every settled combo, add up the dollars every buy staked and the dollars every winning buy redeemed for. On 2-leg combos, $31,267,972 was staked and $30,788,362 was paid, which is 98.5 cents back per dollar. 3 legs paid 91.4 cents. 4 legs, 83.6 cents. 5 legs, 73.5 cents. 6 legs, 71.9 cents. 7 legs, 72.1 cents. 8 legs or more, 56.5 cents. Across the 1,006,027 settled buys, $53,194,688 was staked and $49,416,748 was paid: 92.9 cents per dollar.

Read the method before the number. This counts each taker buy as held to settlement and paid at $1 a share if the combo won, and it leaves fees out. Sells are 13.1% of taker fills, so some buyers exited early on either side and their real result differs from the settlement figure. The comparison across leg counts is the point, and the method is the same for every bar.

The return fell with each leg up to 6, held near 72 cents at 6 and 7 legs, and dropped to 56.5 cents at 8 or more. Buyers of 8 or more legs, as a group, got back a little more than half of what they staked.

The feed opens at $50 and up, and that one default hides most of what happens: 82% of combo buys in the 30 days to October 11 were under $50. They held 12% of the money. That is the right cut for reading money and the wrong one for reading the crowd, so set Size to Any size when you want to see what most combo buyers do.

Each row is one taker fill. The first line is the result and the pick count: Open, Won or Lost, then 2 picks, 5 picks and so on. Under it the bet reads as one sentence, $250.00 to win $437.50, with the multiple beside it and the price on a quiet line below.

The picks column marks each pick with its state, a check for won, a cross for lost and a dashed circle for still open, and links the market it came from on Polymarket. The trader column carries the wallet's grade letter, its Polymarket username when it has one, and the address. Click the name for the profile, where the wallet's combos sit beside its single-market record.

Sort by Newest, Largest or Payout; Largest and Payout rank inside the last 24 hours, the last 7 days or all time. Filter by Result, Sport, Trader grade, Number of picks, Size and Odds, from 10¢ or less down to 1¢ or less, which is 10x down to 100x. Payout with odds at 1¢ or less lists the biggest long shots anyone has chased.

The header counts the combos open right now, recomputed every 5 minutes. Click a pick count to open that combo's own page: every fill of it, the makers and takers, and its life on chain, with the collateral split into it, merged out of it and redeemed from it.

Most combo trades are one quote meeting one request. In the 30 days to October 11, 73% of combo matches had a single maker on the other side, and 45 drew fills from 11 makers or more. The feed counts the taker once per match, so a request that 15 maker orders filled is one row, not 15.

Polymarket publishes no endpoint that lists every combo trade. Each Data API combo route takes one wallet address, so it answers for one user at a time. The feed reads Polygon instead: every 5 seconds the backend reads CombinatorialConditionPrepared events, which register new combos and their legs, and OrderFilled events, which record the fills.

Prices are execution prices on the YES side, the only side a quote request trades, and every taker fill on record is a YES. Legs map to markets by the condition id written into the combo, and the ones that do not match are described through Polymarket's Data API.

Without Pro the whole ledger is open, every filter and every page, a day behind, with one featured combo on a card. Pro shows each fill as it lands, for $25 a week.

A good grade did not carry over to combos. From September 20 to October 4, wallets graded S, A or B when they bought got back 72.3 cents per dollar staked, held to settlement. C, D and F wallets got back 87.9 cents. The gap is -15.6 cents, on an interval from -35.0 to +3.8 that includes zero: this sample found no measurable difference, and all three charted cohorts lost money. A fourth cohort outside grade-history coverage returned 106.05 cents per dollar across 8,237 buys; it is not one of those three.

So do not copy a combo for the letter beside it. A wallet that prices single games well is still buying a conjunction at one maker's quote, and over these two weeks whatever it knew about single games did not show up in its combo results.

Read a wallet's combos beside its other bets. The large-trade feed shows its single-market trades as they land, and the leaderboard ranks it by realized profit. A wallet that is long a team on the moneyline and long the same team inside 3 combos has told you its view 3 times.