Market making on perpetuals, from the inside
Everything here comes from a book we run ourselves: 3.4 million fills across 578 accounts and $276 million of turnover. Where the money goes is not where most explanations say it goes.
Why most explanations are wrong about where the money goes
The usual account of market making says you earn the spread. On liquid perpetuals that is arithmetically impossible as a standalone business: the spread on a major pair is one tick, a small fraction of a basis point, while a full round trip costs several basis points in fees alone.
What decides the outcome is a set of costs that appear in no fee schedule: who is on the other side of your fill, how long you are stuck with what you bought, and how you get out when you are wrong. Those three are the subject of this cluster.
- The spread is a rounding error against the cost of a round trip.
- The real costs are adverse selection and inventory, in that order.
- Fee tiers matter, but they rarely decide the outcome by themselves.
The single most surprising number we measured
On one of our books, removing the worst 1% of cycles moved the cost per million of turnover from 177 down to 43. One cycle in a hundred carried three-quarters of the loss.
The portrait of that 1% is consistent: a larger-than-usual position, held for hours, exited by crossing the spread. It is not a pricing problem. It is an exit problem, and it is why every serious change we have made since has been about getting out rather than getting in.
- 1% of cycles held roughly three quarters of the loss.
- That 1% shares one shape: big, old, and closed as a taker.
- Six different entry filters were tested against it. All six failed.
Where to start reading
If you have never run a maker, start with what the job is. If you have run one and cannot work out where the money went, go straight to adverse selection, then to inventory.
- New to it: what a market-making bot really does.
- Losing money and not sure why: adverse selection.
- Positions piling up: inventory and hold time.
- Choosing a venue: maker and taker fee arithmetic.
Questions
Is market making on perpetuals still profitable?
It can be, on the right market with the right cost discipline. It is not a spread-capture business any more: the winner is whoever turns inventory over most cheaply and avoids the flow that runs them over. Venue choice matters more than cleverness.
How much capital do you need to start?
Less than most people assume, because the strategy is about turnover rather than position size. What matters is that your clip is large enough to clear the venue's minimum and small enough that a normal adverse move does not force you out.
Do I need to write my own bot?
No. What you do need is a bot that runs on your own exchange account through a key limited to trading, so your funds never leave your custody and you can revoke access at any time.
What is the biggest mistake new makers make?
Quoting deeper to get more fills. We tested it: quoting further from the touch raised the cost per million from 143 to 350 and dropped the maker share from 99.8% to 81.9%. Deeper quotes fill precisely when the market is running, which is the worst moment to be filled.