LZLZL/Prediction markets/Mechanics
FREEDO IT A · MechanicsOn-chain

Split, merge, redeem

2026-08-21 · three actions that never touch the order book

Buying and selling is the surface. Underneath sit three actions that bypass the book entirely: turn $1 into a full set, turn a full set back into $1, and after settlement turn the winner back into money. They explain why price must equal probability, and why your winnings do not arrive on their own.

1The three

ActionDoesAvailablePrice
split$1 → one of every outcome tokenany timefixed $1
mergea full set → $1any timefixed $1
redeemthe winning token → $1only after settlementfixed $1

All three share one property: the rate is hard-coded and the market cannot move it. They are contract calls, not trades — no counterparty, no slippage, no trading fee (only gas).

2Why they set the price

Split and merge are bidirectional and always available, and that alone pins the price.

sides sum to < $1 → buy the set, merge, take the difference sides sum to > $1 → split a set, sell the pieces, take the difference both directions push the sum back to $1

So you never have to trust the claim that price represents probability — arbitrageurs hold it there. That lane is picked clean; I censused it and found zero live opportunity inventory. The value to you is not the trade, it is that you can read price as probability safely.

3Split: when it is actually useful

Paying $1 for a set where only one piece will be worth anything sounds pointless. Two real uses:

UseHow
You want to sell a side with no bidsplit a set, sell the side you do not want, keep the side you do — a "buy" executed as a sell
You want to quote both sidesstandard for makers: split, then post offers on both

The first occasionally matters for ordinary traders: when the bid side is thin and the ask side is deep, split-and-sell can beat buying directly. Do the arithmetic rather than assuming a direct buy is always best.

4Merge: cashing a matched position early

Hold 100 YES and 100 NO and those 100 sets are already worth exactly $100, whatever happens. Merging takes that $100 now, without waiting for settlement.

For makers this is the core loop: quote both sides, accumulate matched inventory, merge it to free capital for the next round. The matched part is locked profit; the unmatched part is the entire risk — that leftover is the naked leg, and it is where market making actually lives or dies.

⚠ Reconciliation note: merges are recorded

The activity feed carries a MERGE row — you do not need to infer it. (I originally believed otherwise; this corrects my own earlier note.)

So when profiling a market-making wallet, count trades, merges and redemptions separately. Looking only at trades blends locked-in matched profit with still-open naked risk.

5Redeem: the only one that will not happen for you

After settlement the winning ticket is worth $1 — but it is a ticket. Redeeming converts it into a balance.

⚠ Three things that make you think it already arrived

① The interface shows the position "worth" $1. That is a valuation, not a balance.

② The losing side zeroes and disappears, leaving no record — so reading the feed flatters you. Measured once at 8× off: 589 shares evaporated.

③ The "redeemable" field is not trustworthy. Once the book empties after close, the API may flag both sides redeemable. Judging outcomes from it is simply wrong — see the redeemable trap.

6The dependable test: ask the chain

"Did I win" has exactly one reliable answer, and it is settlement itself rather than any convenience field derived from price:

settlement complete AND my side's payout share > 0 on-chain: the payout denominator is non-zero, and my outcome's numerator is positive
Why it is better
It is the settlementnot inferred from price, volume or a helper flag
Unaffected by an empty bookno quotes, cold market, does not matter
Independently verifiablepublic on-chain; reproduce it on any settled market
No threshold to pickgreater than zero wins, equal to zero loses

Switching my redemption logic to read those two values eliminated the misjudgements. Re-checking history with it, agreement with market settlement was 100% (207:0 on the disputed windows, and 31:0 on a second batch).

7One line to keep

Trading decides what you hold; these three decide what it is worth and when it becomes money. Only redemption requires you — and it sits last in the sequence, after you have stopped paying attention.

EvidenceCheck it yourself

Fixed rates split / merge / redeem are contract calls with hard-coded ratios, visible on-chain.
Arbitrage pins price sum both sides on any board; my census of that lane found zero live inventory.
MERGE is recorded the activity feed exposes a MERGE type directly. This corrects an earlier note of mine.
Win/loss test payout denominator non-zero and own numerator positive; used to re-check all disputed windows against market settlement with full agreement (207:0, and 31:0 on a second batch).
Zeroed legs 825 unpaired shares, only 236 recorded, 589 vanished.
Checked 2026-07 to 2026-08. Contract and API behaviour can change; verify before use.

NextWhere to go

A · MECHANICS
Complementary sets and NegRisk
G · ENGINEERING
The redeemable trap and the phantom-empty position
A · MECHANICS
One trade, end to end
G · ENGINEERING
589 shares evaporated
This is an educational and research record. It is not investment advice, promises no returns, and offers no personalised trading recommendations. Rules and API behaviour are per the official documentation; this page states when it was checked and both can change without notice. Prediction markets are restricted or unavailable in some jurisdictions — confirm your own before taking part.