I keep coming back to the same mistake, usually after watching a fill come in three or four percent worse than the screen quote suggested it should. The pair looked fine. Reported daily volume was in the millions, the spread on the ticker was a couple of basis points, everything said this was a liquid market. Then the order hit the book and the price walked away from me, and the fill blended together at a level I never would have accepted if I had seen it up front. The problem was never that the market was illiquid in some absolute sense. The problem was that I measured it with the one number that lies the most, which is reported volume.
Volume is a headline. It tells you that trades happened over some window, but it says nothing about whether they happened in a way that would have helped you at the moment you needed to trade. A pair can print big daily volume from a handful of wash trades, from market makers passing inventory back and forth, or from one whale that came and went hours before you showed up. None of that tells you whether the book can absorb a real order right now without moving. So I stopped trusting volume as a liquidity measure and started scoring venues with three things I can actually check in a few minutes.
Two percent depth is the number that actually matters
The first and most useful yardstick is how much size sits within a small distance of the mid price. I use two percent as the default band, meaning I add up all the resting bids from the mid down to two percent below it, and all the resting asks from the mid up to two percent above. That gives me two numbers, bid-side depth and ask-side depth, in dollar terms. This is the quantity that tells you how large an order the book can eat before your average fill price slips by roughly two percent.
The reason this beats volume is that it is a snapshot of what is available to you, not a record of what happened to other people. If I want to move a position of a certain size, I compare that size against the two percent depth on the side I need. If my order is a meaningful fraction of the depth, say more than a fifth of it, I already know I am going to move the price and I should either split the order over time or trade somewhere else. If my order is a small slice of the depth, I can be more relaxed. This one comparison replaces a lot of guessing.
Two details make this measurement honest. First, look at both sides separately, because books are often lopsided. A pair can have deep bids and thin asks, which matters enormously depending on whether you are buying or selling. Second, be a little suspicious of very large orders sitting right at the edge of your band, because those are the easiest to fake. A resting order that never gets hit and keeps reappearing at a round number is often there to make the book look deeper than it is. Depth that is real tends to be layered and messy, not one clean wall.
Time-weighted spread, not the spread you happened to glance at
The spread on the ticker is a single instant, and single instants are easy to cherry pick. Market makers can hold a tight quote most of the time and let it blow out during exactly the moments you would want to trade, like right after a large print or during a news spike. So the spread that matters is not the one you see, it is the one you would have paid on average across a stretch of time.
You approximate this by sampling. Watch the top of book over some window, a few minutes is enough to get a feel, and note the bid-ask spread repeatedly rather than once. What you care about is not just the average but how often it widens and by how much. A venue whose spread sits at two basis points nine times out of ten but jumps to thirty basis points during the tenth sample is a very different place to trade than one that holds ten basis points steadily. The average might even look similar. The steady one is far safer, because the moments a spread blows out are exactly the moments you are most likely to be trading with size, so you tend to pay the bad spread more often than a simple average would suggest.
A quick rule of thumb I use: if I cannot watch the top of book settle down and hold a tight quote for a sustained stretch, I treat the tight quotes I do see as decorative. A quote you cannot rely on being there when you send an order is not really a price, it is an advertisement.
How fast the book refills after a sweep
The third measure is the one almost nobody checks, and it is often the most revealing. When a real order comes through and clears out several levels of the book, what happens next? On a genuinely liquid venue, resting orders repopulate those levels quickly, within seconds, because there are participants who want to be there and step back in. On a thin venue dressed up to look deep, a sweep leaves a hole that stays open, and the spread yawns wide until someone decides to quote again.
This resilience, how quickly the book refills after it gets hit, is what separates a market that can support repeated trading from one that only looks good until you actually use it. It matters most if you are working an order in pieces, because each of your child orders is a small sweep, and if the book does not refill between them you are effectively trading against a book that gets thinner with every clip you send. You end up chasing your own footprint.
You do not need a lab to observe this. Watch the book during a moment of real activity, see a level get cleared, and count how long the gap stays open. Fast refill is a green light. A gap that lingers, followed by a jump to a new price, tells you the depth you measured a minute ago was fragile and will not be there when you lean on it.
Scoring a venue in a few minutes
Here is the workflow I actually run before committing real size to a pair on a new venue.
- Pull the order book and sum the resting size within two percent of the mid on each side. Compare that dollar figure against the position you intend to trade. If your order is more than roughly a fifth of the depth on your side, plan to split it or look elsewhere.
- Watch the top of book for a few minutes and note the spread repeatedly. Care about how often and how far it widens, not just the tightest number you catch.
- Wait for a level to get cleared during live activity and count how many seconds until it refills. Seconds is healthy. A gap that lingers is a warning.
- Sanity check the depth for fake walls. Layered, messy, frequently traded depth is real. One clean round-number wall that never gets hit usually is not.
- Only then size the position to what the book can absorb, not to what the reported volume implies it should be able to.
None of this requires exotic tooling. A live order book, a clock, and a little patience get you most of the way. If you want it faster, this is roughly the kind of thing Blockcircle surfaces across the venues and pairs it covers, so you can compare depth and behavior side by side instead of eyeballing one book at a time. Either way, the discipline is the same. Measure what the book can actually absorb, then size to that number and not to the story the volume is telling you. The venues that fail this test do not announce themselves. You find out from a fill that came in worse than it should have, and by then the lesson is already expensive.