Six checks before you quote anywhere
Most of the outcome is decided before a single order is placed. Pick the wrong book and no parameter set rescues it; pick the right one and a mediocre strategy survives.
The first check is arithmetic, and most venues fail it
Add the maker and taker fee for one full cycle, express it in basis points of notional, and compare against the market's typical spread measured over a full day. If the spread does not clear that floor with room to spare, stop. There is no parameter set that fixes a market where the round trip costs more than the book pays.
We learned this by failing it for weeks. On one book the measured spread was 0.07 basis points against a round-trip cost of 2.50. Eighteen of the twenty-six markets on that same venue were better. We were quoting the worst one, and no amount of tuning was going to change the arithmetic.
- Round-trip cost is the floor. The spread has to clear it with margin, not just meet it.
- Measure the spread across a day. One snapshot cannot separate two markets.
- If the arithmetic fails, change market before you change anything else.
| What to check | Why it decides the outcome | What good looks like |
|---|---|---|
| Round-trip cost | It is the floor the spread must clear | Maker + taker for one full cycle, in basis points |
| Tick value in bp | Decides whether queue priority is affordable at all | Below ~0.1 bp: nearly free. Above ~1 bp: it costs your spread |
| Spread persistence | A spread that appears for ten minutes a day is not the spread | Measured across a full day, not a snapshot |
| Depth within a few ticks | Sets your maximum clip | Your clip is a small fraction of it |
| Behaviour under load | Decides whether you can exit during a fast move | Cancels still work, API responsive, mark price sane |
| Fee tier | Real money, and the smaller half of the picture | Compared only between venues that already cleared the floor |
Convert the tick to basis points before anything else
A tick is a fixed price increment, and what it is worth depends entirely on the price level. On a gold perp near $4,450 with a one-cent tick, one basis point is 46 ticks, so a single tick costs about 0.02 bp and buying queue priority is nearly free. On the silver book alongside it, the same one-tick improvement is worth 1.47 bp: sixty-eight times more.
Same exchange, same code, opposite verdict. This is the single number that decides whether the standard toolkit of a market maker applies on your chosen book, and it takes thirty seconds to compute.
- Tick value in bp = tick size ÷ price × 10,000.
- Below about 0.1 bp, queue priority is cheap. Above 1 bp, it costs you the spread.
- A preset that is right on one market is silently wrong on the one next to it.
Depth, persistence, and behaviour under load
Depth decides your maximum size, and the number that matters is not the top of book but what sits within a few ticks of it. Persistence decides whether the spread you measured is still there when you are filled. A book whose spread is wide for ten minutes a day is not a wide-spread book.
Behaviour under load is the check people skip and regret. Watch the venue during a violent move before you commit capital: whether orders can still be cancelled, whether the API stays responsive, whether the mark price does something strange. A venue that pauses while you hold inventory turns a controlled position into an uncontrolled one.
- Measure depth within a few ticks, not at the touch.
- A spread that appears for ten minutes a day is not the market's spread.
- Watch one violent move on the venue before you quote size on it.
Fee tiers come last, and they come last for a reason
Volume tiers and maker rebates are real money and the smaller half of the picture. A rebate improving your cost by a fraction of a basis point does not rescue a market where adverse selection costs several. Settle the arithmetic first, then let the fee schedule choose between books that already pass.
One number worth measuring on yourself rather than assuming: your actual maker share. Ours runs 99.8 to 99.9 per cent, which is what makes the rebate side of the schedule the relevant one. A book running at 82 per cent maker pays taker fees on nearly a fifth of its turnover, and that swamps any tier benefit.
- Rebates improve a viable book; they do not make an unviable one viable.
- Measure your real maker share: the fee schedule applies only as far as it allows.
- Compare tiers only among venues that already cleared the arithmetic.
Questions
What is the best exchange for market making crypto?
There is no single answer, because the question is really about a market rather than a venue. The right book is one where the day-long typical spread clears the round-trip cost with margin, where a tick is worth a small fraction of a basis point, and where the venue stays responsive during fast moves.
Should I market-make on a DEX or a centralised exchange?
Judge it per book rather than per venue type. DEX perps often have wider spreads and flatter fees, which favours smaller accounts, but also thinner depth and more variable execution. The arithmetic check applies identically to both.
How much depth do I need to start?
Enough that your clip is a small fraction of what sits within a few ticks of the touch. If filling one clip visibly moves the book, you are the liquidity event rather than the liquidity provider, and adverse selection will find you immediately.
Does a new venue with high rebates make sense?
Sometimes, and the rebate is rarely the reason. New venues can have genuinely wide spreads because nobody is competing yet, which is the real opportunity. Check whether the spread is wide all day or only during the hours nobody trades.