Strategies·November 17, 2026·8 min read

Calendar Spreads in Prediction Markets: Trading One Question Across Deadlines

How to read a deadline ladder as a term structure, price the conditional probability between two dates, and trade the spread after fees, rule differences and resolution delay.

calendar spreadsterm structurepolymarketrelative value
Whiteboard planning — illustrative image
Photo: Startup Stock Photos / StockSnap · CC0Source

A calendar spread in prediction markets is one position in two contracts on the same question with different deadlines. The near contract pays if the event happens by the first date. The far contract pays if it happens by the second. The price gap is not a discount or a mistake by itself. It is the market's implied probability that the event lands inside the window between the two dates, conditional on it not having happened earlier. Trading the spread means trading that conditional probability, and the trade only works when both contracts share the same definition of the event.

Deadlines are not interchangeable. A market asking whether the Federal Reserve cuts rates by March and one asking whether it cuts by June are separate instruments with separate end dates, separate resolution sources and often very different liquidity. Treating the later contract as a rolled version of the earlier one is the most common way to lose money in this structure.

The term structure of probability

For an event that can happen at most once, the probability by date must be non-decreasing. The March contract cannot rationally trade above the June contract, because every path where the event happens by March is also a path where it happens by June. A ladder of contracts on the same question is therefore a term structure, and its shape is a distribution over dates rather than a single forecast.

That framing matters because the level and the slope carry different information. A 2026 study of Polymarket deadline ladders grouped 1,551 contracts into 338 ladders and split prices into an occurrence component and a timing component. Its estimate was that only about one eighth of the information in prices concerns timing. A high price is a sign that the event is likely, not that it is imminent. For a spread trader the implication is blunt: the level is the well-priced part, the slope is the noisy part, and the slope is where a mispricing has to be found.

Reading the gap as a conditional probability

Take three contracts on a central bank cutting rates: by March at 38 cents, by June at 62 cents, by December at 81 cents. The ladder is monotonic, so no mechanical arbitrage exists. The June minus March gap of 24 cents is not the probability of a cut between April and June. It is the time value of the extra three months, priced as a conditional probability. The gap is that probability multiplied by the chance the event had not already happened.

The conditional is (0.62 - 0.38) / (1 - 0.38) = 0.387. The market is saying that if there is no cut by March, there is a 38.7% chance of one by June. If your own model says 30%, the June contract is rich relative to March. The expression of that view is to buy March Yes at 38 and buy June No at 38, since June No costs one minus the June Yes price. The pair costs 76 cents and pays $1 in two of three states: if the event happens by March, the March leg pays; if it happens after June, the June No pays. It pays nothing if the move lands in the April to June window. Maximum profit is 24 cents, maximum loss is 76 cents, and the break-even probability of the middle window is 76%.

That last number is the part traders skip. The market's implied probability for the middle window is 24%, so the pair is fairly priced at a 38.7% conditional. If your estimate of the conditional is 30%, the unconditional middle window is 18.6%, and the expected value per pair is 0.814 x 24 - 0.186 x 76 = +5.4 cents on 76 cents of capital, about 7.1%. The edge comes entirely from the conditional estimate, not from the ladder looking odd.

The mechanics of a ladder arbitrage

When the ladder is inverted, the trade is mechanical. Suppose June Yes trades at 37 cents and March No trades at 60 cents, so March Yes is effectively 40. Buying June Yes and March No costs 97 cents per set. The payoff is $1 if the event happens by March, $2 if it happens between March and June, and $1 if it happens after June. The worst case is a 3-cent gain on 97 cents, before fees.

Fees are charged per leg on the taker side, using shares x rate x price x (1 - price). At the politics rate of 0.04, 100 June shares at 37 cents cost 100 x 0.04 x 0.37 x 0.63 = $0.93. The 100 March No shares at 60 cents cost 100 x 0.04 x 0.60 x 0.40 = $0.96. Total fees are $1.89 against a gross profit of $3.00, leaving $1.11 net on $97 of capital, or 1.14%. If the spread resolves in three months, that is roughly 4.6% annualized before slippage. One tick of slippage on either leg cuts the edge by a third.

By-date and by-year markets are different contracts

A market asking whether something happens by December 31, 2026 is not the same as one asking whether it happens in 2026. The by-date contract pays for any occurrence up to the deadline. The by-year contract depends on how the rules assign an event that begins in December and concludes in January. The first may resolve on the date of the event. The second may resolve on the date of official confirmation, which can be later.

The same trap appears in ladders that mix scheduled and unscheduled events. A central bank decision has a fixed meeting calendar, so the deadline is unambiguous. A geopolitical event has no calendar, and the rule has to define what counts as the event, which source confirms it, and whether a partial or announced action qualifies. Two contracts in the same visual ladder can therefore have different event definitions, and the monotonicity constraint does not apply to them at all. Before computing any conditional, verify that the only difference between the two contracts is the date.

How to execute a calendar spread

  1. Pull the whole ladder for the event, not the single market you were shown. Group by event rather than searching by title, because titles vary between deadlines.
  2. Write each end date in UTC and name the resolution source for each rung. If the sources differ, stop and treat the rungs as unrelated markets.
  3. Compute the implied conditional for each adjacent pair and compare it with your own timing model.
  4. For a directional spread, quote the less liquid leg first as a maker. The far-dated contract is usually the thin one, and crossing its spread can consume the entire edge.
  5. Use GTD orders that expire before the first scheduled catalyst in the window. A resting quote that survives a meeting is not a calendar spread any more; it is an unhedged position.
  6. For the arbitrage version, verify monotonicity at executable prices. Polymarket displays the midpoint unless the spread is wider than 10 cents, so a ladder that looks inverted on screen can be flat at the touch.
  7. Size to the thinner leg and annualize the return over the expected lockup, including any resolution delay.

What can go wrong

The largest risk is definitional. Deadlines are a promise about dates, not about events. A rule that resolves early when the event happens, a source that publishes a revision, or a clarification that changes the trigger date can turn a monotonic ladder into two unrelated bets. Read the clarification history on every rung.

The second risk is that timing is genuinely hard to forecast. The evidence that only about an eighth of price information concerns timing cuts both ways. It says the slope is inefficient, and it says the slope is mostly noise. A spread trade needs a specific reason why your window estimate is better than the market's, such as a scheduled meeting, a known procedural step or a base rate you can defend.

The third is capital. The short leg ties up 60 or 70 cents to make 20 or 30, and a disputed resolution can lock both legs for days. A 1.1% net return is attractive at 30 days and mediocre at 120. Finally, check correlation with your other positions. Five ladders on the same central bank decision are one macro view, not five independent trades.

Checklist

  • Confirm the two contracts differ only in end date.
  • Write both deadlines in UTC and identify the source for each.
  • Check monotonicity at executable bids and asks, not midpoints.
  • Compute the implied conditional and compare it with your own estimate.
  • Add both taker fees and one tick of slippage per leg.
  • Quote the thin leg first and use GTD expiries around catalysts.
  • Annualize over the lockup, including dispute delay.
  • Cap total exposure per event, not per contract.

Sources

  1. 1.Markets & Events— Polymarket Documentation
  2. 2.Resolution— Polymarket Documentation
  3. 3.Prices & Orderbook— Polymarket Documentation
  4. 4.Fees— Polymarket Documentation
  5. 5.Whether Is Not When: Separating Occurrence from Timing in Polymarket Prices— arXiv
  6. 6.Calendar Spread— Investopedia

More from the journal

All articles