The Quick Answer
A Kalshi order book gets thinner as the event gets closer, not deeper. Across 9,554 live books we captured, a cheap contract 24 to 48 hours from settlement was missing a quote on one side 24% of the time. Inside 12 hours of close, that rises to 84%.
That is backwards from almost every other market. In equities, futures and options, liquidity concentrates into expiry. Here it evaporates.
The practical version: the moment you most want to get out is the moment there is most likely nobody on the other side.
Where The Number Comes From
We captured the top of book — best bid, best ask and the size behind each — on Kalshi's daily high-temperature markets, every few minutes, for two weeks.
| Order Books Captured | 9,554 |
| Distinct Markets | 1,984 |
| Days | 14 (2026-07-30 to 2026-08-12) |
| Series | Kalshi daily high temperature, 8+ US cities |
Grouped by how long each book had left before its market closed:
| Time To Close | books | one-sided | median spread | median size at ask |
|---|---|---|---|---|
| 24–48h | 5,423 | 16.1% | 1¢ | 16 |
| 12–24h | 3,728 | 32.5% | 1¢ | 16 |
| 6–12h | 403 | 78.9% | 2¢ | 10 |
Every column moves the wrong way at once. The share of books with no quote on one side rises fivefold, the spread doubles, and the size behind the best offer drops by more than a third.
The Obvious Objection, And Why It Does Not Explain This
The first thing to suspect is price, not time. Near close a weather contract is usually nearly settled — it drifts to a penny or to 99 — and at a penny there is no room to bid underneath. One-sided books at the extremes could be pure arithmetic.
So we controlled for it: hold the price level fixed and vary only the time.
| Price Level | 24–48h | 12–24h | 6–12h |
|---|---|---|---|
| 1–4¢ | 24% (n=3,525) | 44% (n=2,649) | 84% (n=364) |
| 5–9¢ | 3% (n=963) | 6% (n=535) | 48% (n=27) |
| 10–24¢ | 0% (n=935) | 0% (n=535) | — |
The effect survives, and it is large. A 1–4¢ contract is three and a half times more likely to be one-sided in the last 12 hours than it was a day earlier, at the same price. In the 5–9¢ band the rate goes from 3% to 48%.
⚠️ The 5–9¢ figure in the last column rests on 27 books. Treat it as a direction. The 1–4¢ row is where the sample lives and the claim should be read off that.
⭐ And note the third row: a 10–24¢ contract was two-sided in every one of the 1,470 books we saw more than 12 hours out. Mid-priced contracts do not have this problem until they stop being mid-priced.
Why It Happens
This is the part the general guides get right in theory and never measure. A market maker quoting both sides is exposed to anyone who knows more than they do. As settlement approaches, the information that decides the outcome — the afternoon temperature, the last few innings — is arriving fastest, and the maker's edge against an informed trader is thinnest.
So they do the rational thing and step back. The book does not thin because interest disappears. It thins because the people whose job is to quote both sides stop wanting to.
That also explains the shape of the price control. At 10–24¢ the outcome is still genuinely uncertain and quoting is safe. At 1–4¢ the contract is a tail that only pays if something surprising happens, and a surprise is exactly what a maker cannot afford to be on the wrong side of.
What This Means If You Trade It
- Do not assume you can exit late. If your plan involves selling back into the market in the final hours, the measurement above says there is a good chance there is no bid.
- The cost of leaving is highest exactly when you most want to leave. Spreads double in the same window the book goes one-sided.
- Cheap tails are the worst case. They are the contracts most likely to be abandoned, and the ones where a one-cent spread is already a huge share of the price.
- If you need an exit, take it early. Two days out, five books in six were quoted on both sides.
More on this: Prediction Market Liquidity: Why It Decides If You Can Trade · Kalshi Liquidity Explained: Why Deep Books Win · Which Kalshi Markets Are Actually Liquid? We Measured 48 · When Do Kalshi Markets Close? When A Contract Actually Closes · Stranger Things Season 5: What The Market Says About The Release
What This Does Not Tell You
- One Market Family. Daily high-temperature series. Sports, politics and crypto markets have different maker activity and this pattern may not hold — we have not measured them.
- Nothing Under Six Hours. Our capture thins close to settlement, so we cannot describe the final hours, which are probably the most extreme.
- Fourteen Days Of Summer. Late July into mid-August 2026, a calm stretch.
- Top Of Book Only. We measured the best bid and best ask, not the depth behind them.
- This Is An Association, Not A Proven Mechanism. The withdrawal explanation above is the standard adverse-selection account and it fits the shape, but our data shows that the book empties, not why.
