01.The basic trade
Post a bid below and an ask above where you think fair value sits. If a buyer and a seller both cross to you, you have bought low and sold high without taking a directional view, and the spread is your compensation for standing there. The subtlety is that you do not get to choose which side fills first.
- Quote both sides around your estimate of fair value.
- The spread is payment for providing immediacy to others.
- You are compensated for inventory risk, not for being right about the outcome.
02.Adverse selection is the whole problem
Your quotes get taken most eagerly by people who know something. When news breaks, the informed trader hits your stale bid before you can pull it, and you are left holding a position that is already wrong. This is not an edge case — it is the structural cost of the business, and every real market-making strategy is organised around limiting it.
- The trades you get filled on are disproportionately the ones you did not want.
- Speed of cancellation matters more than speed of quoting.
- Widen or withdraw around scheduled events rather than getting picked off.
03.Inventory, and the binary problem
Ordinary market making assumes you can hedge or flatten inventory. In a prediction market, an accumulated one-sided position goes to zero if the event goes the wrong way, and there is often nothing to hedge with. That makes inventory limits harder than usual: you need a rule for how lopsided you will let yourself get, enforced automatically rather than by judgement.
- Set a hard inventory cap per market and enforce it in code.
- Skew your quotes to attract flow that flattens you when you are lopsided.
- Remember the downside is total, not a drawdown you can ride out.
04.What it takes in practice
Continuous quoting is not something you do by hand. You need a websocket feed to see the book change, an automated loop to reprice, and reliable cancellation — a bot that quotes but cannot reliably cancel is worse than one that does nothing. Resolution risk also puts a natural end date on every market, unlike in continuous markets.
- Stream the book; do not poll it.
- Cancellation reliability matters more than fill rate.
- Every market ends, so inventory has a hard deadline you must plan around.