Adverse Selection & Toxic Flow

Every market maker has a nightmare counterparty: the one who's right.

Here's the setup. We quote both sides of a book — a bid and an offer, all day. Most of the time the trades that hit us are noise: someone rebalancing, someone with a view that turns out wrong, someone who just needed liquidity. We earn the spread and move on.

Then there's the other kind. You get lifted on your offer, and three seconds later the price has ripped through your fill. You weren't unlucky. You traded against someone who knew something you didn't — a signal, faster data, a read on flow. That's adverse selection: the systematic tendency to trade with the counterparty who's better informed than your quote.

When a stream of orders is disproportionately made of those trades, we call it toxic flow.

So how do we tell toxic from benign in a business where every order looks identical at the moment it arrives?

We mostly can't — in real time. What we do instead is measure after the fact and let the pattern speak:

• Markouts. We track where the mid-price is 1s, 10s, 60s after a fill. If our fills are consistently underwater seconds later, that flow is picking us off.
• Timing. Flow that clusters right before news, funding flips, or large prints tends to know more than we do.
• Source. Toxicity is not evenly distributed across venues, order types, or counterparties. It concentrates. Over time it becomes very learnable.

Here's the part that matters for everyone else, not just us.

A liquidity provider can't identify the informed trader at the point of the trade — only in aggregate. So to survive, we price the average toxicity of the flow into the quote. The more toxic a book is, the wider we have to quote and the less size we're willing to show, because every fill carries a higher expected loss to someone who's ahead of us.

Which means the informed traders don't just take money from market makers. They tax everyone who trades alongside them. The uninformed participant — the person just trying to get filled — pays a wider spread to cover the losses we take to the sharp flow they're mixed in with.

That's the real economics behind "why is the spread so wide here?" It's rarely greed. It's usually a book that's been getting picked off, and quotes that have widened to stay solvent.

It's also why serious liquidity provision is increasingly about segmentation: separating flow that's benign from flow that's toxic, so the benign flow can be quoted tighter and doesn't have to subsidize the sharp. RFQ, relationship-based liquidity, and curated venues all exist largely for this reason. Tighter prices for flow we understand; wider prices for flow we don't.

The takeaway

A spread isn't a fee the venue charges you. It's the market maker's insurance premium against being adversely selected — and the toxicity of the flow sets the premium.

Understand that, and most of the "unfair" things about spreads stop looking unfair. They start looking like risk being priced.

Contact: sales@graniteriver.io