The Volume Spike Playbook: Turnover Against Depth

A volume spike is only readable as a ratio. Divide session turnover by the pooled liquidity it traded through, then count how many distinct signers produced it. Those two numbers separate a spike caused by many participants arriving from one caused by the same capital cycling. The playbook below assigns a rule to each class, and each rule states what following it costs.

Why a volume number means nothing alone

Turnover is a sum of trade values. It is not a count of people, not a measure of interest, and not a statement about liquidity. Two pairs can report identical turnover while one has hundreds of participants trading through a deep pool and the other has a handful of wallets pushing size back and forth through a shallow one.

The confusion is understandable, because in traditional markets turnover carries more information: it is reported against a venue with known participants, and a large figure usually implies breadth. On a permissionless chain the same figure can be manufactured by anyone willing to pay fees, and the cost of doing so on Solana is small enough that the possibility is always live.

What turnover does tell you is real and worth keeping. It tells you the pair was used. It tells you fees were generated for whoever provides the liquidity. And, compared against the pair's own history, it tells you that something changed. That is a starting point for an investigation, not a conclusion.

The ratio that makes it readable

Divide turnover by the pooled liquidity the trades passed through. The result is dimensionless: how many times the equivalent of the whole pool changed hands in the period. This converts a number that depends on pair size into one that can be compared across pairs and, more importantly, across sessions of the same pair.

One caution about the denominator. On a concentrated liquidity design the total pooled figure overstates the depth your order will meet, because much of the capital sits in ranges away from the current price. Where the venue exposes it, use active liquidity near price instead. Where it does not, treat the resulting ratio as an underestimate of how concentrated the flow really was.

Four spike classes

The classes are not exclusive and a real session frequently contains several at once. That is not a defect of the taxonomy; it is a reason to record what share of the session each class appears to account for rather than forcing a single label onto it. A spike that is mostly arbitrage with a distribution tail is a different trade from either one alone.

How to classify a spike in five minutes

1. Read pooled depth first. Open the pool account, note the reserves, and identify the owning program so you know whether the total describes depth near price.
2. Compute the ratio. Session turnover divided by that depth. Write the number down; it is the anchor for everything that follows.
3. Count distinct signers. Read the recent transaction list for the pair and count separate signing accounts rather than transactions.
4. Look at spacing. Human and heterogeneous flow is irregular. Trades arriving at strikingly even intervals, or in repeating size patterns, are evidence of a program rather than a crowd.
5. Check both sides. Compare buy and sell value over the window. Near symmetry with high turnover points to cycling or arbitrage; sustained one-sidedness points to distribution.
6. Check whether depth moved. Compare the pool reserves now against the start of the window. Turnover with growing depth and turnover with flat depth are different events.
7. Assign a class and write it down. Include your confidence. A recorded low-confidence classification is far more useful in review than a confident label you cannot reconstruct.

A worked classification, end to end

The steps read as bureaucracy until you run them once. Here is the sequence applied to an invented session, using round numbers chosen to show how the evidence accumulates rather than to describe any real pair.

Step one and two: the pool holds the equivalent of 45 SOL in total and the session shows 540 SOL of turnover, so the ratio is twelve. That number alone rules nothing out. It is high, and high is ordinary for a small pair, so the only conclusion available is that the session deserves the remaining five steps.

Step three: the transaction list shows 610 trades produced by 23 distinct signers. Twenty-six trades per signer on average is not a crowd, and it is the first substantive piece of evidence. Step four: the spacing between trades clusters tightly around a repeating interval, with occasional gaps that look like retries rather than hesitation. Step five: buy value and sell value across the window differ by under two percent, which is not what a directional crowd produces.

Step six is the one that decides the sizing. Pool reserves at the start of the window and at the end are within a few percent of each other, so the pair absorbed twelve turnovers of activity without gaining any capacity to absorb a larger order. Whatever produced the flow did not deepen the market.

Step seven: classify as cycling, confidence moderate, with a note that a distribution tail cannot be excluded because the small sell-side excess is within measurement error. The rule that follows is the cycling rule, which means the position is sized against 45 SOL of depth as though the 540 SOL figure did not exist. Recording the confidence and the reason matters more than the label, because in review it is the reasoning that can be corrected.

The rule attached to each class

Produced activity, described plainly

Some Solana turnover is generated deliberately. Teams run activity through their own pairs so that the pair appears on venue and aggregator screens that rank or filter by turnover, and so that a chart shows continuous trading rather than gaps. The tooling for this is openly sold, and a multi-DEX Solana volume bot is a product that exists precisely because visibility on those screens is a real distribution channel.

Treating this as a scandal is a mistake, and so is treating it as invisible. It has a signature: many transactions, few underlying decisions, regular spacing, near-symmetric buy and sell value, and depth that does not grow. Reading that signature is the same skill as reading arbitrage or distribution, and it feeds the same single output, which is the size you are willing to take.

The reason it matters mechanically is unglamorous. Produced flow is funded by a budget, and budgets end. A pair whose turnover is largely produced can revert to its underlying depth without anything visible happening first, which means your exit assumptions should be based on that underlying depth throughout. Nothing about the practice changes what the pool can absorb.

What honest measurement looks like

Measurement of activity is a solvable problem, and the way it is solved tells you how seriously to take any figure. An honest measurement states its window, its source, whether it counts both sides of a swap, whether it deduplicates routed legs that belong to one user action, and which venues it covers. Any number missing those attributes is a headline, not a measurement.

That standard applies to figures published by tooling as much as to venue dashboards, which is why how volume campaigns are measured is a more useful question to ask a vendor than what results they claim. A measurement methodology can be checked against the chain; a result cannot be checked against anything.

For your own records, prefer raw quantities to derived ones. Store the reserves, the turnover, the signer count and the window rather than a ratio, because the ratio can be recomputed later and the inputs cannot be recovered from it. Public explorers such as Solscan expose the raw material, and the discipline of storing inputs is what allows a review to reach a different conclusion than you did at the time.

Where this playbook fails

It fails on timing. Classification takes minutes and some spikes are largely finished inside that window. The playbook has no answer to this beyond accepting the loss, and any claim that a faster heuristic would be equally reliable would be an invention.

It fails on ambiguity. Coordinated wallets and independent wallets are indistinguishable on chain. Every signature described here is evidence rather than proof, and a trader who converts evidence into certainty has made the classification worse than useless by attaching confidence to it.

It fails on the denominator. Where a venue does not expose active liquidity near price, the ratio is computed against a total that overstates real depth, and every conclusion drawn from it inherits that error. Note the limitation in the record rather than pretending the number is exact.

And it fails, as every rule set does, when the output is treated as a prediction. The playbook classifies what has already happened. It contains no statement about what price will do next, because turnover is a record of the past and no amount of processing turns a record into a forecast.

Questions the desk gets asked