The Overround / The Week in Margins
Same exchange, same week, same screen. The difference between the cheapest and dearest book on the board is 31.6 points, and almost nobody trading either one could tell you that.
Every price you see quoted anywhere has a markup inside it. Add up what it costs to back every outcome of a market and a fair book comes to 100%. Anything above that is the house's, paid by whoever trades, on whichever side. That figure is called the overround, it is trivially computable from public order books, and essentially nobody publishes it.
So we write it down every day. Here is what the last 7 days say.
The dearest market on the board this week is “Oscar Winner: Best Picture” at 132%. The cheapest is “2028 U.S. Presidential Election winner?” at 100.4%.
There is a pattern in that, and it is not flattering to anyone who thinks of these venues as neutral infrastructure. The tight books are the ones being watched: heavily traded political contracts where a fat spread would be arbitraged away by lunchtime. The wide ones are the quiet corners. Awards. Appointments. Anything where the crowd is thin enough that nobody is checking.
That is not a scandal, it is how market making works everywhere. It is only a scandal that the number is invisible at the moment you trade.
The average book across everything we priced went up 2.2 points this week, from 111.5% to 113.7%. Which would be a story, except it mostly is not one.
Different markets carry a usable book on different days, so a raw weekly average drifts for reasons that have nothing to do with anybody's pricing. Restrict it to the 12 markets that had a book on every one of the 7 days and the move is +0.8 points. That is the honest number, and it is a lot duller than the first one.
| Market | 2026-09-07 | 2026-09-13 | Change |
|---|---|---|---|
| Who will be the next Supreme Court justice? | 101.6% | 121.5% | +19.9 |
| Next French presidential election winner? | 121.4% | 126.7% | +5.3 |
| 2026 Midterms: House popular vote margin of victory? (Generic ballot) | 108.5% | 109.7% | +1.2 |
| How much will the US acquire Greenland for? | 110.8% | 112% | +1.2 |
| 2028 Democratic VP nominee | 114% | 105.7% | -8.3 |
| Who will be Trump's next Press Secretary? | 117.8% | 110.2% | -7.6 |
| 2028 U.S. Presidential Election winner? | 102.4% | 100.4% | -2 |
Across 15 occasions this week where we could price the same question at both venues with enough confidence to publish, Kalshi's book averaged 117.2% and Polymarket's 121.5%, a gap of -4.4 points. Kalshi was the dearer of the two in 11 of 15.
Every figure above comes from prices this site recorded itself, once a day, at a fixed hour, and stored. Nothing is backfilled and nothing is modelled. The raw snapshots are public at /api/snapshot, one file per day, free to use with attribution. This page regenerates from them, so it is not an opinion held in a database, it is the arithmetic run again each time you load it.
A book is only published for a market when every active contract in it is quoting, because a partial sum is not a book. The scale on the chart starts at 100% for the same reason the number matters: 100% is what a fair market costs, and everything to the right of that line is the toll.