Why late buyers lose

What exit liquidity means

Exit liquidity means being the buyer whose purchase lets somebody else get out. Every sale needs a buyer, so in one sense everybody who ever buys anything is exit liquidity for whoever sold to them, and nobody loses sleep over it. The phrase only becomes meaningful in a specific situation: where the buying is concentrated, finite, and drawn from a group of people who were all told the same thing at slightly different times.

In that situation, the supply of later buyers is not endless. It runs out. Whoever is holding when it runs out absorbs the entire difference between the price they paid and the price the asset returns to. They are not unlucky. They are structurally last, and somebody had to be.

The rest of this page shows that falling out of ordinary arithmetic. There is no interpretation involved, and the conclusion does not depend on anyone behaving badly.

How a pool prices a purchase

Most small tokens trade against an automated pool rather than an order book. A common design holds two assets and keeps the product of the two reserve quantities constant across every trade. If the pool holds SOL and a token, then buying tokens means adding SOL and removing tokens, in whatever quantity keeps that product the same.

Two consequences follow immediately, and they are the only two you need.

First, every purchase pushes the price up as it executes, so the average price you pay is worse than the price quoted before you started. Second, the size of that effect depends on how much is in the pool. A large pool absorbs a purchase with a small price change. A small pool does not.

This is not a defect. It is how the pricing works, it is documented behaviour, and the transactions are all visible afterwards on a public Solana block explorer for anyone who wants to check the reserves before and after a trade. What matters here is that the same property that makes a shallow pool move impressively upward makes it move just as hard downward.

The worked example

Round numbers, chosen for readability. No real token, no real pool, no real person. A pool starts with 400 SOL and 40,000,000 tokens, so the quoted price at the start is 0.00001 SOL per token.

Four groups of buyers arrive in sequence. Each group spends exactly the same amount, 40 SOL, so nobody is being greedier than anybody else and nobody has an advantage in size. The only difference between them is when they arrived. Then, after the buying stops, each group sells its entire holding, in the same order they arrived.

At the moment the buying stopped, the pool held 560 SOL against 28,571,429 tokens, which quotes a price of roughly 0.0000196 SOL per token, almost double where it started. That number is what the chart showed. It is also a price at which nobody sold anything, because the act of selling moved it.

Reading the result

The total column is the part worth sitting with. One hundred and sixty SOL went in and one hundred and sixty SOL came out. Nothing was created. Every unit that the first group gained came from the third and fourth groups, and no argument about the quality of the asset changes that, because the pool ended exactly where it began.

The distribution is also worth noticing. The first group roughly doubled its money. The second group barely moved. The third and fourth lost, and the fourth lost most. All four spent the same amount. All four made the same decision. The only variable was position in the sequence, and position accounted for the entire spread of outcomes.

There is one more detail hiding in the table. The fourth group paid an average of 0.0000182 while the quoted price at the top was 0.0000196. The chart never showed the price they actually got, in either direction. Charts report the last trade; they do not report what a quantity would fetch, and in a shallow pool those two numbers are meaningfully different exactly when the chart is at its most persuasive.

Why the round trip is not symmetric

People often assume that if buying pushed the price up by some amount, selling the same quantity pushes it back down by the same amount, so a round trip is roughly neutral. For the pool as a whole that is true. For an individual participant it is not, and the difference is what the table shows.

When you buy, you pay progressively worse prices as your own order moves the ratio. When you sell, you receive progressively worse prices for the same reason. Both effects work against you. The only participants for whom the round trip is favourable are the ones whose buying happened before the price moved and whose selling happened before it came back, which is a description of position rather than a description of skill.

This asymmetry gets larger, not smaller, as the pool gets shallower. That is the uncomfortable part, because the assets involved in these episodes are selected for shallowness. The property that makes the chart move is the same property that makes exiting expensive, and it is not possible to have one without the other.

The costs that never come back

The illustration above is generous, because it ignores every cost. Add them back and the total column stops summing to zero and starts summing to a negative number.

• Pool fees. Automated pools charge a fee on every swap, taken out of the trade. It is paid on the way in and again on the way out, by everyone, regardless of outcome.
• Network fees. Every transaction pays to be included. Small individually, and paid on failures as well as successes.
• Failed attempts. During a burst of activity, a meaningful share of transactions do not land. Retrying costs more fees, and a retry that succeeds at a worse price costs twice.
• Slippage beyond expectation. A trade sent during rapid movement can execute at a materially different price than quoted, which is a cost that lands unevenly and lands hardest during exactly the moments that feel most urgent.

None of these is dramatic on its own. Together, across a whole group, they mean the arrangement is not a redistribution with a zero total. It is a redistribution with a negative total, where the shortfall is paid to pools and validators. Even in the version where the plan works perfectly, the participants collectively finish behind.

Order of arrival is the whole variable

Everything above collapses into one sentence: in this structure your outcome is decided by when you arrived, and by nothing else you can influence.

You cannot see your position. The chat does not show it, the chart does not show it, and the number of people talking does not indicate it, because the people talking are not the same set as the people who have already bought. Somebody who arrived first is not necessarily loud, and somebody being loud is not necessarily early.

What makes this genuinely hard is that the experience of arriving late is indistinguishable from the experience of arriving early. In both cases you see a message, look at a chart that is moving, feel the pressure of a moment passing, and buy. The information that would tell you which situation you are in is precisely the information the structure does not produce.

This is why "get in early" is not advice. It is a description of a variable you cannot observe, offered as though it were an action you could take.

Why fair rules do not change it

Arrangements of this kind sometimes come with rules that sound protective: everybody buys in the same window, nobody sells before a stated point, allocations are equal. It is worth understanding why none of this helps, because the rules are often sincere.

Consider what perfect compliance would mean in the illustration above. If all four groups genuinely bought simultaneously, they would all pay the same average price, and when they all sold simultaneously they would all receive the same price, minus fees. Everybody would finish slightly down and nothing would have happened. The arrangement produces gains for anybody only when the buying and the selling are separated in time, which requires the rules to be unevenly followed.

So the rules cannot be doing what they appear to do. At best they slow the sequence down. In practice they mainly ensure that whoever ignores them does best, which is a well understood property of any agreement where defecting is profitable and enforcement is impossible.

A regulator would describe the underlying activity rather than the etiquette around it, and the customer advisories published by the Commodity Futures Trading Commission are written at exactly that level: they describe the shape of coordinated buying followed by insider selling without any interest in whether the participants had house rules.

Why the pool never got deeper

A reasonable objection to the illustration is that a real episode brings attention, attention brings participants, and more participants ought to make the market less shallow. It is a fair thought and it is worth answering directly, because the answer is the last piece of the arithmetic.

Trading through a pool does not add to a pool. Every swap takes one asset out and puts the other in, leaving the total commitment roughly unchanged apart from fees. Depth grows only when somebody deposits both assets into the pool on purpose, which is a separate action, taken by a different kind of participant, for reasons that have nothing to do with a chart moving for an afternoon.

In practice, depth during an episode of this kind is more likely to fall than rise. Whoever supplied the original liquidity is watching an asset become volatile, which is the condition under which supplying liquidity is least attractive, and withdrawing is a single transaction away. So the buyer arriving at the busiest moment may well be trading against a shallower pool than the one that produced the impressive part of the chart.

That is the quiet reason the fourth row of the table does so much worse than intuition suggests. The chart implies a market that got bigger. The reserves say otherwise, and the reserves are what price the exit.

Why it looks like it works

If the arithmetic is this clear, why does the impression persist that these things pay?

Because you only ever hear one side of the table. The first row of the illustration has every reason to talk about it, and screenshots of the first row are what circulate. The third and fourth rows have nothing to post and every reason to stay quiet, which means the visible record of any episode is systematically drawn from its winners.

There is also a timing effect. During the buying phase, everybody who has bought is showing a gain, including the people who will finish worst. If you look at a group chat at that moment, the mood is unanimous and the evidence appears overwhelming. The people who will absorb the loss have not yet absorbed it, and they are, at that moment, contributing to the impression that it works.

Both effects are ordinary features of how information gets selected, and neither requires anybody to be lying. The record you can see is not a sample of what happened. It is a sample of what people chose to publish.

This is not a story about foolish people

It would be easy to end a page like this with a lesson about greed. That would be both unkind and wrong.

The structure described here works on ordinary judgement operating on incomplete information, under time pressure, with visible social confirmation and a chart that appears to corroborate everything. Every single input is designed, or at minimum selected, to point the same way. Reasonable people reach the predicted conclusion, and the fact that many reasonable people reach it is what makes the mechanism function at all.

If you finished in the third or fourth row of a real version of this table, the arithmetic on this page is a description of a mechanism, not a verdict on you. The most useful thing about seeing the numbers is that they are legible in advance next time, and the second most useful thing is knowing that the outcome was set by a variable nobody showed you.

If something already happened, what to do afterwards covers the practical first steps and is honest about how often funds come back. If you want the mechanism described from a participant point of view rather than through arithmetic, how coordinated buying works does that, and the tool-or-scheme checklist is the version you can run against a specific offer before anything happens at all.

Questions readers send about this page