GL-03 · Market structure · 2026-09-22 · 6 min
Volume over open interest
A one-line test for recycled volume, run live across the perp DEXes that report both numbers.
A venue that pays for volume gets volume. Points programs, fee rebates for size and airdrop campaigns reward every trade the same whether anyone wanted the position or not. You can spot the result with one division, using two numbers most perp venues already publish.
The ratio
Open interest counts the positions still open. Volume counts every trade in the last day. A trader who opens a position and holds it adds to both once. A trader who opens and closes the same size forty times a day adds forty times to volume and nothing to open interest. So a high tells you that most of the volume came from positions nobody kept.
Volume over open interest, 24 h
CoinGecko, converted at BTC …, cached 5 min
| Venue | Open interest | Volume 24h | Ratio | Perps |
|---|---|---|---|---|
| Reading venues… | ||||
Reading the bands
Glassy colours each venue by four bands. They are a reading aid, not a verdict, and they come from how the ratio behaves, not from any venue's own statement.
- Under 1x. The day's trading turned over less than the open book. Positions last longer than a day on average.
- 1 to 3x. An active market where most of the flow still leaves positions behind.
- 3 to 10x. Heavy intraday churn. Market makers recycling inventory can produce this on their own, and so can a volatile day.
- Over 10x. Most of the volume cancels itself out. Check the venue's incentives before you read that volume as demand.
Where the test misleads
A liquidation cascade can push a clean venue past 10x for a day. A venue in its first weeks has small open interest, which inflates the ratio before anyone farms anything. Venues also count open interest differently: some count one side of each contract, some count both, and CoinGecko normalises what it receives without saying how for each venue. Compare a venue with itself over time before you compare it with its neighbours.
The ratio says nothing about book quality either. A venue can run at 0.8x with a spread of 5 bps and 20 bps of depth that vanishes on the first move. For that, open the terminal: depth within 25 bps and the widest spread of the session tell you what the volume figure cannot.
Source
The table reads CoinGecko's public /derivatives/exchanges endpoint through a Glassy route that caches the answer for five minutes, filters it to decentralized perp venues and converts BTC figures to dollars at CoinGecko's own BTC price from the same call. Lighter's Robinhood Chain instance appears on its own row because it runs separate books from Lighter on Ethereum.