How Split, Merge, and Redeem Work
Split turns a dollar of pUSD into a Yes and a No. Merge turns that complete pair back into a dollar. Redeem pays the winning token after resolution. All three sit outside the order book — not a how-to, and not a live book that sums to a dollar.
By Jordan NabigonPublished 8 min read

What you'll be able to do
- State that a position on the documented venue is a token balance, and that Yes and No are a fully backed pair.
- Distinguish split, merge, and redeem from a book trade.
- State the documented identities: $1 becomes 1 Yes + 1 No; 1 Yes + 1 No becomes $1; a winning token becomes $1 after resolution, with the rare 50/50 exception named.
- State the retrieved operational limits: merge needs equal amounts; redeem is after resolution and has no deadline on the manage page; redeem has no amount parameter.
- Refuse the readings this page does not support: split the gap; merge to lock; redeem now; a live book sums to $1.
Before you start
- what-is-a-prediction-market
- why-complementary-prices-add-up-to-a-dollar
What this guide won't do
- No live prices. The $100 scale examples, the $0.60 buy example, and the 1_000_000 = 1 pUSD unit are documentation examples retrieved 2026-08-16T00:48Z.
- Display, fee, and complementary-gap readings belong to companion pages and are not recensus'd here.
- Current UI availability, gas, and per-market adapter addresses are UNKNOWN. The Contracts page was not opened.
- UMA proposal/dispute timetable is out of scope.
- No trading advice. Vendor use-this-when wording is not adopted.
- Rules cited from one venue's public docs; they are not universal and are dated by retrieval.
Last reviewed 16 August 2026
A position on the documented venue is a balance of outcome tokens for a given market. Official docs retrieved 2026-08-16T00:48Z say every prediction is represented by those tokens. Three operations move money between collateral and the tokens outside the order book: split, merge, and redeem.
What this page does not claim. It does not tell you to split, merge, or redeem. It does not claim a live book sums to a dollar, or that a gap is a lock. The $100 scale examples and the $0.60 buy example below are documentation examples on the retrieved pages, not live fills. Why complementary prices should add up to about a dollar is a separate companion. How a market's rules and oracle path decide the winner is a separate companion. The negative-risk conversion — one No becoming Yes on every other outcome — is a separate companion. This is not a rewrite of those pieces.
A market is two tokens backed by one dollar
On the positions-and-tokens page (retrieved 2026-08-16T00:48Z), each market has exactly two outcome tokens. Yes redeems $1.00 if the event occurs. No redeems $1.00 if it does not. The tokens are ERC-1155 assets on Polygon, using the Gnosis Conditional Token Framework. Every Yes/No pair in existence is backed by exactly $1 of pUSD collateral locked in the CTF contract. The how-positions-work page says the same thing: the pair is fully collateralized.
The collateral itself is pUSD — Polymarket USD. The pUSD page (same retrieval) calls it a standard ERC-20 on Polygon mainnet, six decimals, backed by USDC, with that backing enforced onchain. No algorithmic peg. No fractional reserve. Wrapping USDC.e into pUSD, and unwrapping it back, is a different operation on that page. It is not split, merge, or redeem. This page does not teach wrap.
These three operations sit outside the book
The manage-positions page (retrieved 2026-08-16T00:48Z) says you manage positions outside the order book by splitting collateral into a complete set of outcome tokens, merging balanced tokens back into collateral, or redeeming winning tokens after resolution.
A book trade is a fourth path, on the positions-tokens page: buy or sell one side. Their documentation example is buy Yes at $0.60 — pay $0.60, receive 1 Yes token. That creates one side. Split creates both sides from a dollar of collateral. Those are not the same action.
Split: collateral becomes a complete set
Split converts pUSD into outcome tokens. Splitting $1 creates 1 Yes token and 1 No token. The docs' own scale example, on both the positions-tokens page and the manage-positions page, is $100 pUSD → 100 Yes + 100 No. Polymarket 101 says the same identity: $1 pUSD creates one Yes share and one No share.
The manage-positions page lists two preconditions: enough pUSD in the wallet for the amount, and an approval for the collateral adapter to spend that pUSD. The documented flow is: approve the adapter; call it with the market's condition ID and the amount; the adapter runs the underlying CTF operation and mints equal Yes and No balances atomically.
Amount is in pUSD base units. The same page says 1_000_000 splits 1 pUSD into 1 Yes and 1 No. That matches the pUSD page's six decimals. Those figures are documentation, not an invoice.
The positions-tokens page's "use this when" lines — create inventory for market making; obtain both sides — are vendor wording. This page reports them. It does not tell you to mint inventory.
Merge: a complete set becomes collateral again
Merge converts a complete set of tokens back into pUSD. It requires equal amounts of Yes and No. The docs' scale example is 100 Yes + 100 No → $100 pUSD. Every 1 Yes and 1 No returns 1 pUSD.
Preconditions on the manage-positions page: equal Yes and No balances, and an approval for the adapter to transfer those outcome tokens. The documented flow burns equal Yes and No balances and returns pUSD atomically. The amount cannot exceed either outcome-token balance. A documented max option merges the smaller of the two balances.
Vendor wording on the positions-tokens page: exit a position without trading; convert accumulated tokens back to collateral. Reported, not adopted. This page does not say merge is how you take a complementary-price gap. That reading belongs to the companion, and even there it is not a how-to.
Redeem: after resolution, the winning side becomes collateral
Redeem is not available on an open market. After a market resolves, winning tokens exchange for pUSD. In the usual documented case, each winning token returns $1 and each losing token returns $0. The scale example is 100 winning tokens → $100 pUSD. If the market resolves Yes, the manage-positions page writes 100 Yes → 100 pUSD and 100 No → 0 pUSD.
The same page says there is no redemption deadline. Winning tokens remain redeemable at any time after resolution. Redemption has no amount parameter: it redeems the wallet's balances for both outcomes. The adapter burns both outcome balances and returns the winning payout in pUSD atomically.
The resolution page (retrieved 2026-08-16T00:48Z) adds the settlement-side description: once a market resolves, trading stops on that market; winning tokens become redeemable for $1.00; losing tokens become $0.00. Redeem runs through the CTF collateral adapter, which burns the ERC-1155 outcome tokens, receives the released USDC.e collateral, wraps it into pUSD, and returns pUSD to the wallet.
One documented exception sits on that same resolution page. In a rare Unknown / 50-50 resolution — "neither outcome applicable" — the page says the market resolves 50/50 and each token redeems for $0.50. The usual $1 / $0 table is not the only documented payout. This page does not claim every redeem is a dollar.
Who proposes the outcome, what bond they post, and how long a dispute can run are a different guide. This page only needs: redeem is after resolution, and resolution is a process with an owner.
Two adapters, and a conversion this page does not own
Official docs use different collateral adapters for standard markets and negative-risk markets. The how-positions-work table: both still use the ConditionalTokens contract; standard markets trade on a CTF Exchange and keep multiple outcomes independent; negative-risk markets trade on a Negative Risk CTF Exchange and link outcomes through conversion. The manage-positions page selects CtfCollateralAdapter when negRisk is false and NegRiskCtfCollateralAdapter when it is true.
The conversion itself — 1 No becoming 1 Yes on every other outcome — is the companion negative-risk guide. This page does not retell it. Current contract addresses are UNKNOWN here. A Contracts page is linked from the docs; it was not opened this run.
What to do with the words on the screen
Name the operation. Split creates both sides from collateral. A buy creates one side from a book. Merge needs a complete set and returns collateral before resolution. Redeem needs a resolved market and, on the retrieved manage page, has no deadline.
If the sentence you are reading is about a Yes price plus a No price, you are on the complementary-price companion, not on this page. If it is about a named source and a deadline, you are on the resolution companion. If it is about "Other" and a conversion, you are on the negative-risk companion.
This page does not tell you to call any of the three operations.
Editorial review
Reviewed by Jordan Nabigon on 16 August 2026.
How AI was used here
Research packet, draft, reviewer/editor pass. Tim adversarial pass under Jordan 2026-08-15 auto-approve standing order. Not a claim Jordan fact-checked the numbers.
Sources
- Positions & Tokens primaryPolymarket · retrieved 16 August 2026
- How Positions Work primaryPolymarket · retrieved 16 August 2026
- Manage Positions primaryPolymarket · retrieved 16 August 2026
- Polymarket USD primaryPolymarket · retrieved 16 August 2026
- Resolution primaryPolymarket · Official description of market rules, resolution sources, end dates, and edge cases.
- Polymarket 101 primaryPolymarket · retrieved 16 August 2026
Read next
- What Is a Prediction Market?
A prediction market is a market in contracts that pay a fixed amount if a defined event happens. The price is what people will pay for that payoff right now — useful as a probability estimate, but not a fact about the world.
- What one academic paper documented about complementary-price extraction on Polymarket
Saguillo et al. (AFT 2025) classified complementary-price, NegRisk, and combinatorial extraction on markets that resolved 1 April 2024 to 1 April 2025 and reported $39,587,585.02. That is a pre-fee, dated measurement — not “bots made $40 million,” and not a claim the same gaps are still there in 2026.
- The Briefing
What moved in prediction markets, in short.
