The Tick Tax

Last weekend I saw this post on X discussing tick size changes and liquidity fragmentation and it got me thinking, what is the impact of tick sizes on liquidity. So I spent the week in our tick data.

A venue's tick sets the quoted spread, but nobody notices in majors like BTC because both figures are fractions of a basis point. Beyond BTC in the risker assets you see more impact from tick sizing. The issue here is that these ticks get set at listing and venues rarely follow the price down. This results in a token that fell 90% carrying an increment ten times too coarse.

For instance, on one venue OP cannot be quoted inside 110 bps and sits pinned at one increment 96% of the day, yet that coarse book carries 177 times the displayed size and nearly seven times the volume of the fine-tick alternative. The best null case I found is a fine-tick APE book streaming 139 quotes a minute all day against a spread near 100% and not one dollar traded. Availability without convertibility is not liquidity.

As for precedent, it turned out I did not need the equity market history, because two natural experiments ran on crypto venues this summer. In July, Binance cut the minimum price step on five small perp markets to a tenth of what it was, and these were books where the gap between best bid and offer had been stuck at that minimum for most of the day, meaning the step itself was what held prices apart. Shrink the step and the gap shrank with it, three to five times, exactly as the historical equity research says it should. But the improvement was paid for from the other side of the book, since the amount actually sitting at those best prices collapsed between eight and 56 times while overall trading barely changed. If you trade small you got a better price, and if you rest size waiting to be filled, you paid for it.

In June, Bitstamp had already run the opposite test, shrinking the minimum step on its whole book by ten to 100 times in one week. The difference was that its steps were not the thing holding prices apart, since the gap between best bid and offer sat at the minimum less than 1% of the time before the change. To judge the effect I compared each coin against the same coin on venues that changed nothing over the same hours, and the answer is that the cut bought nothing. The gaps ended up slightly wider than on those other venues, the amount waiting at the best prices fell by a third while it held steady elsewhere, and the books got noisier, updating their quotes 62% more often. Put the two together and you get what the equity research already proved, that shrinking the step only helps when the step was the problem. One number tells you which case you are in before you touch anything, the share of the day a book already sits at the minimum gap, and June is what ignoring it costs.

Now back to Kalshi and the example the piqued my interest in the first place. Typically on Kalshi every football market steps in flat one-cent increments, the books sit pinned at exactly one tick 73% of all hours, rising to 89% in MLS, and the wings are where it hurts, since a one-cent step on a five-cent longshot is a 20% increment and one contract in nine trades at prices where the step is at least a tenth of the cheap side. Kalshi's own 15-minute BTC markets run tenth-of-a-cent steps in exactly those wings and the makers use them fully, with sub-cent spreads 99% of wing minutes, so the remedy exists in production one menu over.

As the poster on X noted, Kalshi halved the tick on 60 Brasileirao Serie A (Brazilian Football League) markets opened 1 to 3 August. What they didn’t note was this was reverted for new listings within three days. The comparable cent-tick cohort was pinned 56% of hours, a book only partly constrained, and the yield matched, the median spread stayed at one cent, now two ticks, with sub-cent quotes in just 13% of hours. That is a third point on the same curve, binding ticks pay for a cut, slack ones do not, half-binding ones yield little, and the wings, where the tick truly binds at 20% of price, are still exactly where the football ladder does not go.

Every depth number here is displayed size at the best bid and offer, and a ten times finer grid mechanically spreads resting orders across ten times as many levels. Touch depth can therefore fall even when the dollars within a fixed band around mid did not leave. The same discipline applies on the taker's side, where our falls are quoted spreads and the quoted-to-effective wedge decides whether takers actually paid less. Until the book walk and the effective spreads are built, the honest reading is that cutting a binding tick visibly moves the trade-off between the two liquidities, while cutting a slack one costs you one of them for nothing.

The week ahead is quiet on paper, which after the fortnight we have just had is welcome. Japan CPI on Friday the 21st is the only print of consequence before Jackson Hole opens on the 27th, and Congress stays out until mid September with the CLARITY vote waiting. The sport to watch is Kalshi's final half-cent fixtures settling this weekend, which puts the spotlight's experiment live on your screen before the ladder goes back to the cent.