The worst fills I have ever taken were not on volatile days. They were on the day after a delisting notice, trying to get out of a pair where the book had already gone hollow. You put in a sell and the price you see is a fiction, because the depth underneath it is thin enough that your own order walks the market down. By the time the notice is public, the people who were going to leave have mostly left, and you are selling into whatever is left, which is not much.
The thing I keep relearning is that exchanges rarely delist a healthy, liquid pair on a whim. Most delistings are the last step in a slow decline that leaves fingerprints. If you know where to look, you can usually see the book deteriorating well before anyone posts anything official. That does not give you certainty. It gives you a reason to trim early and exit into a market that can still absorb you.
What actually degrades first
The signals show up in liquidity long before they show up in price. Price can hold flat for weeks while the machinery underneath rots. A few things I pay attention to, roughly in the order they tend to appear.
- Maker depth thins out. The resting liquidity within a percent or two of mid gets shallower. Market makers are usually the first to notice a venue losing interest in a pair, and they pull quotes or widen them because the inventory risk stops paying. This is the earliest and most reliable tell, because it is a business decision by people who watch the pair full time.
- Spreads widen and stay wide. A spread that used to sit tight starts gapping and does not snap back. Occasional widening is normal. A structural shift where the typical spread doubles or triples over a couple of weeks is the book telling you fewer participants care.
- Volume drifts below listing-maintenance thresholds. A lot of exchanges publish, or at least imply, minimum activity requirements to keep a listing. When a pair's rolling volume slides toward those floors and stays there, it moves onto an internal list you cannot see. You can approximate that list by watching volume decay against the venue's stated rules.
- Monitoring or caution tags appear. Several exchanges attach a formal label to at-risk assets before they act, something like a monitoring tag or an investment-caution designation. This one is public and explicit, and it is often the last warning before a decision. If a pair you hold gets tagged, treat it as a countdown, not a coin flip.
- The team goes quiet. Communication silence from the project is softer and easier to misread, but it correlates. When a team stops shipping, stops posting, and stops answering, the venue relationship usually erodes in parallel. Silence alone means little. Silence stacked on top of thinning depth means a lot.
Why the exit window closes so fast
What makes delistings dangerous is the reflexivity. The moment a notice drops, everyone reads the same book you are reading. Liquidity providers cancel because they have no interest in warehousing an asset that is about to lose its main venue. Other holders rush the door at the same time. So the depth that was already thin collapses to almost nothing precisely when the most sell pressure arrives. You end up crossing a spread that has blown out to something absurd, and slippage does the rest.
This is why watching the precursors matters more than reacting to the announcement. The announcement is the point where your remaining options get much worse. Everything useful you can do happens before it, in the stretch where the book is quietly deteriorating but still deep enough to sell into without moving the price much. You are trying to exit into liquidity that still exists, not liquidity that used to exist.
Building a watch that flags at-risk pairs
You do not need anything exotic to catch most of this. You need to measure a few things per pair, on a rolling basis, and flag when several move together. Here is the shape of a watch I would build.
- Track depth within a fixed band. For each pair you hold, snapshot the resting bid and ask liquidity within, say, one percent of mid on a schedule. Store a rolling baseline. Flag when current depth drops meaningfully below its own recent normal. You care about the pair's trend against itself, not against some universal number.
- Track spread as a percentage of price. Log the typical spread over each window. Flag a sustained widening, not a single spike. The word sustained is doing the work here, because noise will trip a naive threshold constantly.
- Track volume against the venue's maintenance floor. If the exchange publishes minimums, encode them. If it does not, use a conservative floor of your own and flag pairs that spend more than a few days near it.
- Scrape the tags. Monitoring and caution designations are usually posted somewhere public. Watch that page. A new tag on something you hold should page you, not sit in a log.
- Score, do not alarm on any single input. One signal firing is noise. Three firing together on the same pair is a reason to start trimming. The value is in the correlation, so weight the combination rather than treating each input as an independent alarm.
The output you want is a short ranked list of pairs where the risk is climbing, updated often enough to be useful and quiet enough that you actually read it. If you already run a platform that tracks order-book depth and market scorecards, this is mostly wiring existing signals into a delisting-specific view. We built something along these lines into Blockcircle for exactly this reason, because the underlying data was already there and the only missing piece was framing it as an exit-timing question.
The rule of thumb I actually use
When two or more of these fire on a position at the same time, I stop treating it as a hold and start treating it as a wind-down. Not a panic sell. A deliberate scale-out over days, into whatever depth is still there, before the crowd shows up. If the notice never comes, I have given up a little on a position that was fading anyway. If it does come, I am already mostly out, watching the book collapse from the sidelines instead of from inside a frozen order that will not fill.
The asymmetry is what makes this worth doing. Being early costs you a bit of upside on a pair that was already losing interest. Being late can cost you most of the position. I would rather be early and slightly wrong a lot of the time than late and badly trapped once.