The tick size is the moat: why a finer price grid makes passive market making worse

What this is about

Polymarket is a prediction market. You buy shares in a yes-or-no question — will the highest temperature in Tokyo be 24°C on 22 September? — and at settlement each share pays exactly $1.00 or exactly $0.00. Nothing in between. Questions cover weather, sports, elections and crypto prices, and most resolve within days.

I run an automated market maker on it. The bot quotes both sides of a market and earns the spread: buy one YES and one NO on the same market and the pair redeems for exactly $1.00 whatever happens, so the profit is locked the moment the second leg fills. Polymarket also pays a maker rebate for providing liquidity. The risk is the leftover — shares whose other half never filled, which go to settlement alone and pay 1.00 or 0.00.

Everything below comes from six weeks of live quoting with real money: 701 settled markets, $19,889 of deployed capital, one account, one venue.

This piece compares that venue with a newer one. Hyperliquid's HIP-4 outcome markets, live since May 2026, are the same kind of instrument on different rails — binary contracts settling at 1.00 or 0.00, quoted on an order book shared with Hyperliquid's perpetual futures. Two frontends, outcome.xyz and trade.xyz, have grown quickly on top of it.

The question is narrow: what happens to a passive market-making strategy when the minimum price increment shrinks from a cent to a thousandth of a cent? HIP-4 is a credible product, not a curiosity — several of its markets have cleared a million dollars of cumulative volume. The answer is not the one the spread numbers suggest, and it is mechanical rather than a matter of taste.

The measurement

246 outcome markets enumerated through Hyperliquid's public API, every YES-side book pulled. Three venues share that book: out (99 markets), txyz (68) and skew (71). The three frontends are three deployers on one order book, not three exchanges.

226 of 246 books were two-sided (92%). Filtering to what a maker could actually work — spread at or under 10c, price between 0.02 and 0.98 — leaves 112 markets, 46% of the venue:

More workable markets than the bot currently quotes, at an eighth of the spread. On the face of it, the newer venue wins.

The spread difference is mostly the price grid

Polymarket's 1-cent tick is a hard floor on the spread. A market there cannot quote tighter than 1c, and the observed median is four ticks wide. HIP-4 permits a thousandth of a cent.

I expected a second driver and did not find it. HIP-4's outcome set is heavily price-linked (perp:BTC, perp:HYPE), so a maker can hedge an outcome against the same venue's perp — which should compress spreads regardless of tick. But 76 of the 112 workable markets are sports (participant 46, competition 30), entirely unhedgeable, and they quote at a 0.38c median. The hedging story does not survive contact with the composition. The grid explains it.

Why a coarse tick is worth paying for

A minimum tick does three things for a resting quote, and only the first is obvious.

It floors the spread. One tick is the smallest possible gross margin on a round trip.

It protects queue position, and this is the one that matters. To step in front of a resting order, someone must improve by a full tick. On a 4c spread that costs 25% of the entire edge — expensive enough that most participants join the queue instead of jumping it. Time priority becomes a real asset, and waiting becomes a strategy.

On a 0.001c grid, improving on you costs essentially nothing. Queue position stops being defensible, passive quoting collapses into a speed contest, and the counterparties in that contest are firms already running perps infrastructure on the same venue who can extend to outcomes at near-zero marginal cost.

It suppresses quote churn, because there are fewer levels to flicker between.

Three things that would have changed this conclusion, and did not

A maker rebate on HIP-4. There is none. Makers who would otherwise earn a rebate pay zero instead. On Polymarket the rebate is worth +0.65 per $100 of capital deployed — measured, from receipts, not modelled.

A prohibitive fee. Also no. HIP-4 uses HyperCore's spot schedule (0.070% taker / 0.040% maker at tier 0) and charges only on closing or settling — opening and minting are free. On a 0.50 share that is 0.020c maker, 0.035c taker. Against a strategy capturing 1.91c a share, the fee is about 1% of the edge. I over-weighted this in my first pass and was wrong to. The fee is noise; the spread and the rebate are the story. deployerFeeScale is 1.0 on all 170 outcomes that carry it, so no frontend is marking up on top.

Volume having arrived. Partly. Several markets show seven-figure cumulative volume on the frontends — but Hyperliquid's own page reports the same market at $53,205 in 24 hours, against $1.9M “Vol” on the frontend. Those are different quantities: cumulative against daily. Outside the top handful, HIP-4's daily flow runs in the hundreds of dollars — $195, $162, $120, $88, $22, $13, $5, $0.

The honest turn

A coarse tick is not a strategy. Over 701 settled markets and $19,889 of capital on Polymarket — the venue with the protective grid — the same passive strategy runs at −0.79 per $100 including the rebate.

The tick delivered the margin. It captured 3.82c per completed set. And the programme still lost, because the loss was never in the spread: it is in the unmatched residue that goes to resolution alone. Tick size protects the quote. It does nothing about inventory.

There is also a cost to the coarse grid that shows up plainly in the same data. A wide tick concentrates everyone onto the same level: on a settled Polymarket ladder, seven of eleven rungs had zero offers against 4,169 to 9,383 shares bid at 0.999. A four-cent spread with eight thousand shares ahead of you is worth less than a half-cent spread you can actually reach.

So the correct formulation is not “coarser is better.” It is:

There is an interior optimum, and tick size decides whether your edge comes from patience or from speed.

What this does and does not establish

Does: that the HIP-4 / Polymarket spread gap is substantially a tick-grid artefact rather than a liquidity or competition difference; that HIP-4 pays no maker rebate; that its fee schedule is immaterial at these sizes; that the frontend volume figures are cumulative rather than daily.

Does not: that HIP-4 is a bad venue. It is demonstrably a better one for anyone who can hedge into HyperCore perps, and for takers a 0.070% fee against a 0.51c spread is cheap. Nor does it establish anything about HIP-4 profitability over time: every figure here is a single snapshot of 246 books. A week of recorded book history would settle far more than a snapshot can, and the same recorder points at both venues with a changed URL.

One methodological caveat. The Polymarket median spread of 4.00c comes from the recorder's published universe, which selects for quotability. The HIP-4 median comes from every market on the venue, filtered only for a workable two-sided book. The two samples are not selected the same way, and that flatters neither side consistently.