Strategies·November 17, 2026·7 min read

Threshold Ladders on Polymarket: Trading the Gaps Between Strike Prices

How monotonicity works across bitcoin, temperature and other threshold markets, how to compute ladder arbitrage after fees, and why different resolution rules can turn a sure payoff into a bet.

threshold marketsladder arbitragebitcoinpolymarket
Cryptocurrency finance — illustrative image
Photo: Icons8 team / StockSnap · CC0Source

A threshold ladder is a set of contracts that pay if the same underlying finishes above different levels: bitcoin above $100,000, $105,000 and $110,000 on the same date, or a daily high temperature above 70, 75 and 80 degrees at the same station. The prices must obey a simple ordering. The higher the bar, the lower the probability. When they do not, the gap is either an arbitrage or a warning that the contracts are not measuring the same thing.

That second possibility is the one that costs money. Most ladder arbitrage on prediction markets is not a pure mispricing. It is two contracts with different sources, different deadlines or different rounding rules, priced by people who read the rules more carefully than you did.

Monotonicity and implied probabilities

Let A(K) be the probability that the underlying finishes at or above threshold K. For any two thresholds K1 below K2, A(K1) must be at least A(K2). The probability of clearing a low bar cannot be lower than the probability of clearing a high bar.

The difference between two adjacent probabilities is the probability of landing in the bucket between them. If A(100,000) is 62 cents, A(105,000) is 48 cents and A(110,000) is 33 cents, the buckets are 14 cents for 100,000 to 105,000, 15 cents for 105,000 to 110,000, and 33 cents for above 110,000. The probability of finishing below 100,000 is 38 cents. The four numbers sum to 100.

That is a cumulative distribution built from market prices. It is useful even if you do not trade the ladder: it shows what the market implies about the distribution. A ladder with a 2-cent bucket and a 30-cent bucket says the market expects a jump.

For temperature markets the same arithmetic applies, but the threshold definition matters. High above 70 can mean strictly greater than 70 or at least 70. A one-degree difference is a different contract, and a market that rounds to the nearest degree can resolve differently from one that truncates.

Ladder arbitrage

The cleanest case is an inversion. Suppose the 100,000 Yes is offered at 52 cents, the 105,000 Yes is bid at 56 cents, and the 105,000 No is offered at 44 cents. The second Yes implies the first, so it cannot be more expensive than the first.

The trade is to buy the 100,000 Yes at 52 and buy the 105,000 No at 44, for a total of 96 cents. The payoff depends on where bitcoin finishes. Below 100,000, the first leg pays 0 and the second pays 1, so the pair pays 1. Between 100,000 and 105,000, the first pays 1 and the second pays 1, so the pair pays 2. Above 105,000, the first pays 1 and the second pays 0, so the pair pays 1. The minimum payoff is 1, and the cost is 0.96, so the gross profit is 4 cents per set.

Fees change the number. On 100 sets, the Yes leg costs 100 times 0.04 times 0.52 times 0.48 = $1.00 in a politics-rate market. The No leg costs 100 times 0.04 times 0.44 times 0.56 = $0.99. Total fees are $1.99, so the net profit is $2.01 on $97.99, or 2.05%. At the sports rate of 0.05, the fees are $2.48 and the net is $1.52, or 1.54%. That is still positive, but it is small enough that one tick of slippage on either leg can erase it.

The second case is an incomplete set of buckets. If three mutually exclusive and exhaustive outcomes are priced at 30, 30 and 30 cents, buying all three costs 90 cents and pays 1 dollar whichever wins, a gross profit of 10 cents. But a real event can have an other outcome, and the sum can legitimately be below 1. Polymarket's negative-risk documentation warns that new outcomes and placeholders can appear, and that the definition of Other can change. A ladder that ignores the placeholder is not a complete set.

The third case is a butterfly or convexity violation. For vanilla options, the implied density must be nonnegative, which imposes a shape on prices across strikes. Prediction-market threshold contracts are digital options, not vanilla calls, so the Breeden-Litzenberger formula does not transfer directly. The practical version is simpler: bucket probabilities cannot be negative, and the buckets must sum to 1. If a middle bucket is priced below zero, you have found the inversion above.

Fees and execution

Every leg pays its own taker fee. The formula is C times rate times p times (1 - p), and it peaks at 50 cents. A four-leg ladder at 50 cents in a politics market pays about 4 cents per share in total, which is 8% of the 50-cent purchase price. A 3-cent gross gap is negative after fees. At 30 cents the per-leg fee is smaller in dollars but larger relative to the stake, so the hurdle does not disappear at the extremes.

Maker orders avoid the fee but introduce leg risk. If you rest on three legs and only two fill, you are naked on the third. Polymarket's fill-or-kill order type applies to a single order, not to a basket, so there is no atomic way to execute a multi-leg ladder through the standard order types. You can leg in with limits, but you need a maximum naked exposure and a time limit.

Size to the minimum depth across the legs, not the sum. A ladder with $50,000 on the first leg and $2,000 on the third is a $2,000 trade. The top of the book is not the size that clears.

Non-comparable resolution rules

This is the risk that makes ladder trading dangerous. Two contracts can look like the same underlying and resolve on different data.

For bitcoin, one market may use a specific exchange's BTC/USD price at a specific time, while another uses a different exchange or a different index. A $105,000 touch on one venue may not happen on the other. For temperature, one market may use the official climate report from a specific station, while another uses a commercial weather service. The daily high can differ by a degree or two, which is the difference between a winning and a losing bucket.

The deadline matters too. On Friday can mean the observation at 11:59pm Eastern, the settlement price at a futures exchange, or the close of a trading session. A contract that resolves early when the threshold is crossed is not the same as one that waits until the end of the period.

If the rules are not identical, the ladder is not a single distribution. It is a set of separate bets that happen to share a headline. The inversion can be real in the prices and fake in the payoffs, because the two legs can resolve on different states of the world.

How to execute

  1. List every strike, deadline, source and settlement rule. If you cannot state them in one sentence each, you do not understand the ladder.
  2. Build the payoff matrix across the states. The minimum payoff is the number that matters; the expected payoff is a view.
  3. Compute the fees at the size you can actually fill on every leg.
  4. Check the depth on each leg separately.
  5. Prefer the version with the fewest legs and the same resolution source.
  6. If you must leg in, set a maximum naked exposure and a time stop.
  7. Verify the resolution language against the official rules before you enter, not after.

Risks and limits

The first risk is rule mismatch. The second is source mismatch. The third is time mismatch. The fourth is illiquidity: a ladder can be arbitraged on paper and impossible to execute because one leg has no depth. The fifth is early resolution, which can leave a hedge on the wrong side. The sixth is capital lockup: a 2% return over 60 days is about 12% annualized, but only if the capital is not needed elsewhere.

A ladder that looks like a 5% arbitrage may be a 5% bet on two rulebooks agreeing about a threshold. That is a different trade, and it deserves a different position size.

A short checklist

  1. Are the thresholds strictly ordered and measured on the same underlying?
  2. Do all legs share the same source, deadline and rounding rule?
  3. What is the minimum payoff across all states, not the expected payoff?
  4. What are the fees at my size on every leg?
  5. Is the depth on the weakest leg enough for the whole trade?
  6. Can I hold a naked leg if one order fails?
  7. Is the annualized return worth the capital lockup and rule risk?

Threshold ladders are a clean way to read a distribution from prices. They are a dangerous way to assume that every gap is free money. The arithmetic is simple; the rules are not.

Sources

  1. 1.Negative Risk Markets— Polymarket Documentation
  2. 2.Resolution— Polymarket Documentation
  3. 3.Fee Schedule— Kalshi
  4. 4.Prices of State-Contingent Claims Implicit in Option Prices— Journal of Business
  5. 5.Unravelling the Probabilistic Forest: Arbitrage in Prediction Markets— arXiv

More from the journal

All articles