Perp DEX spread feasibility: tick-floor microstructure across three venues
L1 spread measurements on Aster, GRVT, and Bluefin, and the difference between a venue where retail market making is economically possible and one where it isn't.
A passive market maker on any venue captures spread and pays adverse selection. For that trade to be net-positive, the quoted spread has to exceed fees and markout combined. On tick-floored markets — where the minimum-tick dominates the typical spread — this equation doesn't close before the first fill. We measured median L1 spreads across three perp DEXs (Aster, GRVT, Bluefin) to understand which, if any, offer tradeable microstructure for retail-fee market makers.
TL;DR
- On BTC and ETH, both Aster and GRVT spreads are pinned at one tick for most of the sample (~0.013 bps BTC, ~0.043 bps ETH). Round-trip retail fees alone exceed these spreads by two orders of magnitude.
- Bluefin is structurally different: five of six measured tokens show spreads of 5–7 bps, 3×–130× wider than the same tokens on Aster or GRVT. In raw-spread terms, Bluefin is the first perp DEX we have measured where retail-fee maker economics can plausibly close.
- Adverse selection still has to be characterised live; structural feasibility is a necessary but not sufficient condition.
Methodology
Per-sample spread in bps of mid: (ask − bid) / ((ask + bid) / 2) × 10000. Non-crossed-book filter (ask > bid > 0). Data: a 15-minute Bluefin+Aster WebSocket capture from Railway Singapore on 2026-04-22, merged with GRVT reference rows from a 5-minute capture two days prior. 134k valid samples, zero reconnects on the 15-minute run.
No bot; no capital was deployed. These are market-data observations — not live trading data.
Per-venue median L1 spread (bps of mid)
| Venue | Token | n | p25 | p50 | p75 | p95 |
|---|---|---|---|---|---|---|
| Bluefin | BTC | 34,920 | 1.05 | 1.68 | 2.23 | 3.77 |
| Bluefin | ETH | 19,229 | 4.68 | 5.14 | 5.61 | 6.64 |
| Bluefin | SOL | 17,412 | 4.64 | 5.09 | 5.76 | 6.78 |
| Bluefin | SUI | 12,244 | 4.13 | 5.16 | 5.16 | 6.20 |
| Bluefin | HYPE | 23,317 | 6.15 | 6.89 | 7.64 | 10.85 |
| Aster | BTC | 12,093 | 0.013 | 0.013 | 0.052 | 1.093 |
| Aster | ETH | 10,026 | 0.042 | 0.043 | 0.472 | 1.373 |
| Aster | SOL | 7,545 | 1.131 | 1.163 | 1.164 | 2.326 |
| GRVT | BTC | 706 | 0.013 | 0.013 | 0.013 | 0.013 |
| GRVT | ETH | 783 | 0.043 | 0.043 | 0.043 | 0.043 |
| GRVT | SOL | 691 | 1.162 | 1.163 | 1.163 | 1.164 |
Three findings
BTC and ETH on both Aster and GRVT are tick-size-floored
BTC's p50 of 0.013 bps corresponds to a $0.10 tick at ~$75k. ETH's 0.043 bps is a $0.01 tick at ~$2,300. These are not quoted spreads in the economic sense — they are the minimum-increment floor of the book. No passive-maker strategy can extract spread capture from a venue quoting at minimum tick; the spread to capture is literally zero.
GRVT's spreads are near-constant
p25, p50, p75, and p95 are identical to three decimal places across every token in our sample. Either the GRVT book is maintained with extreme stability by resting liquidity, or the sample is dominated by a few quote providers whose resting sizes clamp the book. Either way, the spread is not moving.
Bluefin is the outlier
Bluefin's p50 spreads — 1.68 bps BTC, 5.14 bps ETH, 5.09 bps SOL, 5.16 bps SUI, 6.89 bps HYPE — are 3×–130× wider than the same tokens on Aster or GRVT. This is not a tick-size artifact; it reflects genuine dispersion in the maker population. On the four tokens where Aster and GRVT run at the tick floor, Bluefin offers quoted spread of economic substance.
The tick-floor arithmetic
Round-trip retail maker costs at today's tiers:
- Aster retail with token-based discount: ~0.95 bps round-trip
- Bluefin standard retail: 2.0 bps round-trip (maker 1.0 bps × 2)
- GRVT tier 1: ~−0.01 bps (small rebate)
For spread capture to be net-positive on fees alone — before adverse selection — the quoted spread has to exceed the round-trip cost. Applied to BTC and ETH on Aster and GRVT, the spread is two orders of magnitude short. Spread capture is structurally negative before the first fill.
On Bluefin, net-of-fees capture is 3–5 bps on ETH/SOL/SUI and slightly negative (−0.3 bps) on BTC at current fee tiers. The pre-adverse-selection numbers clear the viability threshold on the mid-cap tokens.
Adverse selection is the other half
A quoted spread that survives fees doesn't automatically survive the next question: what is the average markout on a fill? Our per-fill economics on a comparable perp DEX (retail fee tier, tight books) measure at −1 to −2 bps at the venue level — i.e., the mid moves 1–2 bps against the maker in the first few seconds post-fill. That number is a venue-structural property in our data, not a symbol-specific one.
If Bluefin's adverse selection lives in a similar range (plausible — it is also a perp DEX with similar flow composition), realised net capture lands near +1 to +3 bps per fill on the viable tokens, and structurally negative on BTC. That survives as a strategy hypothesis. It needs live-capital validation to confirm.
Intra-window stability
The Bluefin spreads are not bursty. Within the 15-minute capture, 3-minute bucketed medians vary by ±0.3 bps on ETH, SOL, SUI, and HYPE. BTC ranges 1.19–1.86 bps. WAL tightened notably in the last few minutes — consistent with an external maker onboarding. A passive strategy targeting the mid-window median is not chasing a transient.
What this does not measure
- Queue position. A quoted 5 bps spread does not mean a new maker at 5 bps fills quickly. Fill-rate characterization requires a live probe.
- Adverse selection on Bluefin specifically. Requires live capital. The best proxy we have — similar-venue AS — gives a directional estimate, not a venue-specific number.
- Behaviour during stress. Fifteen minutes is enough for variance estimation but too short to capture large moves, funding events, or CEX-side outages.
- Cross-pair correlation in realised P&L. Each token's economics are independent in this analysis; in a live deployment they share margin and correlate through the underlying.
Conclusion
Three findings, ranked by decision value:
- Aster and GRVT are tick-floored on majors. Retail market-making on BTC/ETH on either venue is blocked by microstructure, not strategy. The spread to capture is below the tick-increment resolution. No tuning closes this.
- Bluefin has genuine quoted spread on majors. It is the first perp DEX we have measured that shows non-tick-floored spreads of 3–7 bps on tokens where other DEXs run at tick. Retail-fee maker economics plausibly close on five of six tested tokens before adverse selection.
- The venue picks the game, not the strategy. On a tick-floored venue, adverse selection is almost the entire economic signal; there is no spread to offset it with. On Bluefin, the spread budget exists but is not infinite. The question shifts from “can this work at all” to “how much of the quoted spread actually fills.”
Live-testing on Bluefin is the natural next step. The correct approach is the narrowest possible capital exposure — a single token, a single strategy profile, a few hundred dollars — and measure realised markout before scaling.