How much is actually resting on a Kalshi book?
Crypto books multiply twenty fold into the final minute; commodity books do the reverse. The market this lab trades most is the thinnest on the exchange, and books lock at the extremes and never in the middle. From 8.9 million snapshots.
Short answer: less than you think, and it moves in opposite directions depending on what you are trading.
Depth is the first question anyone serious asks about a venue, and it is the one the public API cannot answer after the fact. Kalshi's book history is short and ages out, so unless you were recording at the time the answer is simply gone. This lab has been recording the top of book nightly since mid August. What follows is 8,906,068 snapshots from 23 to 29 August 2026.
Crypto books thicken twenty fold into the close
The median number of contracts resting at the best price, on the 15-minute crypto markets:
| time to expiry | median depth |
|---|---|
| 7 to 15 minutes | 37 |
| 3 to 7 minutes | 37 |
| 1 to 3 minutes | 69 |
| final 60 seconds | 755 |
That is the opposite of most venues, where liquidity evaporates as an event resolves. Here it multiplies by twenty in the last minute.
The reason is that these are binary contracts with a deadline. In the final seconds the outcome is close to known, so resting size at 97¢ is nearly riskless and people will post a great deal of it. Early in the window, when the thing is genuinely uncertain, almost nobody will.
Commodities do the reverse
The same measurement on the 15-minute oil, gold and silver markets:
| time to expiry | median depth |
|---|---|
| 7 to 15 minutes | 318 |
| 3 to 7 minutes | 202 |
| 1 to 3 minutes | 199 |
| final 60 seconds | 178 |
Deeper than crypto at every point, and declining rather than rising. Same exchange, same contract structure, same 15-minute clock, opposite behaviour. Whatever drives the crypto pattern is not a property of the instrument.
The market this lab trades most is the thinnest on the exchange
Median depth by price band tells the same story from another angle. Crypto sits between 33 and 63 contracts across every band. Esports runs 59 to 284. Commodities run 182 to 377.
So the fast crypto ponds, which are the most traded and the most written about, have roughly a fifth the resting size of the sleepy commodity ones. Anyone sizing a strategy on those books is working with far less than the volume figures suggest.
Books lock at the extremes and never in the middle
A locked book is one where the bid meets the ask and there is no spread to capture. It happens constantly at the edges and essentially never in the middle:
| price band | crypto | commodity |
|---|---|---|
| 1 to 25¢ | 31% | 30% |
| 25 to 90¢ | 0% | 0% |
| 90¢ and above | 63% | 52% |
Zero in the middle is not rounding. Across millions of observations between 25¢ and 90¢, the book was effectively never locked.
This has a practical edge to it. If your code treats a locked read as invalid data and skips it, it is blind for roughly two thirds of the ticks above 90¢, which is exactly where a favourite lives. This lab made that mistake and a trail went blind precisely where it should have been arming.
Why any of this matters
Volume tells you a market is busy. Depth tells you whether you can actually do anything there. On these books the honest reading is that almost everything is small: a strategy that works at one contract may have no room at ten, and the place it has least room is the place everyone trades.
The full tables, broken out by family, by price band and by time to expiry with sample sizes on every cell, are in the members edition, along with the method and what could not be measured.
Recorded by this lab, 23 to 29 August 2026. Kalshi's own book history does not reach back this far, so unlike most of what is published here you cannot check this one against the API. The method is described in the members edition so it can at least be checked against itself.