Market making
A wide spread is not a profitable one. What decides whether quoting a market pays is how much of that spread survives the people taking it.
Quoted +1.00pp, keeps +0.95pp. The flow taking it was not better informed than the quote.
Quoted +1.00pp and still loses -0.82pp. Widening would not fix it: the price moves against the quote after every fill.
| Market | Quoted | Effective | Keeps | Volume |
|---|---|---|---|---|
2028 Democratic presidential nominee: Alexandria Ocasio-Cortez Kalshi188 windows | +1.00pp | +0.95pp | +0.95pp | 104k |
US ceasefire against Iran continues through August 31? Polymarket24 windows | +0.25pp | +0.27pp | +0.50pp | 283k |
Fed decision in Sep 2026?: Fed maintains rate Kalshi183 windows | +1.00pp | +1.00pp | +0.49pp | 884k |
Will there be no change in Fed interest rates after the September 2026 meeting? Polymarket270 windows | +1.00pp | +1.01pp | +0.46pp | 926k |
Israel x Iran ceasefire continues through August 31? Polymarket60 windows | +1.21pp | +0.69pp | +0.35pp | 163k |
Swipe the table sideways to see every column.
You are seeing the 5 markets where quoting pays best. Trader shows every market measured, with the adverse selection and flow imbalance behind each number.
15 further markets are measured but not shown, including the ones that lose money, which are the ones worth knowing about.
What each plan includes →What these numbers are
Effective spread
What the taker actually paid away from the mid, rather than what the quote advertised. Usually smaller than the quoted spread, because not every fill happens at the touch.
Adverse selection
How far the price moved in the taker's favour afterwards. This is the cost of quoting to somebody who knew more, and it is the part a wider spread does not fix.
What a maker keeps
The first minus the second. Negative means quoting that market loses money at any width, which the quoted spread alone will never tell you.
Why sport is excluded by default
Measured across a day of live data, 154 of the 200 markets with enough flow to qualify were sport fixtures, and their numbers dominated everything. One showed adverse selection of 27 points.
That figure is not wrong, it is measuring something else. Adverse selection means a quote was picked off by someone who knew more. On an in-play fixture the price moves because the match is being played, and over a 15 minute horizon a football market is measured across a quarter of the game. The measure assumes a market whose clock is news; those have a clock that is a whistle. They are still there to look at, under the link above, they are just not the same question.
Rows marked thin have fewer than 10 windows with a mid recorded after the flow. Their arithmetic is the same and their evidence is not, so they are shown and labelled rather than hidden or averaged in.
None of this can be computed after the fact. It needs what was quoted and what traded against it, recorded at the same moment, and neither venue publishes either one historically. The window is therefore as long as we have been collecting, and it grows. Method →