A spike is a question, not an answer

Turnover on a Solana pair jumps ten times in twenty minutes. That single fact is compatible with at least six different things having happened, and the chart shows the same shape for all of them.

This desk classifies a surge before explaining it, and explains it before anyone acts on it. Every reading names the class, the fields that put it there, and the one observation that would move it somewhere else. Where the frequency of a pattern is unknown to this desk, the word printed is unknown.

Open the volume console Start with the head note

The fields a surge is read from

Six properties of the window, none of which is the turnover figure itself.

New addr: The share of trading addresses that have never appeared in this pair before, and whether that share holds up as the window runs on.
Tail: Whether the counterparty list is still growing at the end of the window or has settled into a set that keeps recycling.
Spacing: The distribution of gaps between trades, which is where a crowd and a fixed configuration look least alike.
Sizes: How widely trade sizes are spread, and whether they cluster on a small repeating set of values.
Inventory: What each participant is holding when the window closes compared with when it opened.
Reserves: Whether pooled liquidity moved during the window, which separates a plumbing event from everything else.

The six surge classes

Classification comes first because the same jump in turnover means completely different things depending on which of these produced it. The classes are deliberately coarse. A coarse label that can be defended beats a precise one that cannot, and two of these six routinely get mistaken for each other.

Attention

New participants arrive because something off chain pointed at the pair.

Plumbing

The liquidity structure under the pair changed, and trading followed the change.

Distribution

The holder set is changing hands rather than growing.

Routing

Turnover is generated by intermediaries doing their job, not by end demand.

Produced

Turnover was generated on purpose so the pair reads as busier.

Echo

The surge is a reaction to the surge: alerts, screeners and follow-on automation.

Featured watch notes

Four notes carry the rest of the site. Each states the question it answers, the evidence behind the reading, the finding that would overturn it, and how firmly the conclusion is worded.

Solana volume spike analysis

Why turnover is the least informative number a spike produces, the eight fields that separate the six classes, and why this desk publishes no figure for how common any class is.

Open the head note

Classifying a surge

The sequence itself. Which question is asked first, why the cheap structural checks run before the expensive behavioural ones, and how a class is recorded with its own falsifier attached.

Read the framework

Volume without price movement

Heavy turnover with a flat price and unchanged depth is a strong constraint, not a mystery. What it mechanically requires, and the mirror case where price moves on almost no volume at all.

Read the divergence note

Reacting without being baited

A twenty-minute triage for a live spike: what to read in the first two minutes, what to refuse outright, and how to write the note so a later review can judge whether the reading was sound.

Read the procedure

Two sections, one standard

Surges covers the shape of the event: what a spike is, how it is classified, and the divergence that separates turnover from demand. Causes covers the mechanisms that put the activity there and what to do while it is still running.

Surges

What a spike is, how it is classified, and the one divergence that separates a surge in turnover from a surge in demand. These notes describe the shape of the event itself.

Open the section

Causes

The mechanisms that generate sudden turnover, how to tell arriving demand from produced turnover, and a triage sequence for reading a live spike without being pulled into it.

Open the section

How a class earns its label

A class is not a guess dressed up in a code. Each of the six has a required evidence set, and a reading that cannot produce that evidence is written down as undecidable rather than nudged towards the most interesting neighbour. Undecidable is published here as often as anything else, because a spike observed for four minutes usually cannot support more.

Confidence is expressed in four named bands, and the band is chosen by what was measured rather than by how confident the writer feels. Firm is reserved for structural facts recorded on chain, such as a pool being created inside the window. Inference does not reach Firm, however many fields agree.

The method note sets out the definitions, the evidence requirements, the four bands and the specific sentences this desk refuses to write about any named token, wallet, team or venue.

Read the method note

What this desk will not claim

That a named token, wallet, team or venue manipulated anything. Intent is not recorded on chain.
That a surge implies a future price. No note here contains a target, an entry or an exit.
That any class is common or rare, unless a frequency has actually been measured and can be cited.
That a single field settles a class. Every field on this site has an ordinary explanation attached to it.
That a reading is final. Each class is published with the observation that would move it.
How produced activity could be shaped to defeat these checks. Detection is explained; concealment is not.

One class of surge is manufactured on purpose

Class S5 is different from the other five in one respect: somebody chose it. Attention, plumbing, distribution, routing and echo are all side effects of something else happening. Produced turnover is the output of a configuration that a person filled in, with a wallet count, a size band, an interval and a budget.

That matters to a reader for a practical reason rather than a moral one. A configuration has parameters, parameters repeat, and repetition is exactly what the spacing and size fields are built to notice. Knowing what the tooling can be set to is what turns a narrow size band from a curiosity into something interpretable.

It also sets the tone this desk works in. Produced flow is openly sold, priced and documented, and is a normal part of Solana market structure in the way that execution algorithms are a normal part of any other market. Treating it as a scandal makes the reading worse, not better, because it pushes an analyst towards accusation and away from measurement.

See how produced activity works

Parameters that leave a trace

How many wallets the activity is spread across, and how those wallets were funded.
The interval between trades, and whether it is fixed, randomised inside a band, or event driven.
The size band each trade is drawn from, and how narrow that band is relative to pool depth.
Whether buys and sells are balanced by design, and over what window the balance is enforced.
Which venues are included, and whether one pool ends up carrying nearly all of the turnover.
When the run starts and when it stops, which is usually the sharpest edge in the whole record.

How this desk works

Three commitments decide what appears on this site and, much more often, what does not.

Classify before explaining

The first output of a reading is a class code and a confidence band, not a story. Stories are easy to build around any chart and are almost impossible to dislodge once written, which is why the class comes first and the narrative is allowed only after the fields have been recorded.

If two classes fit equally well, both are named. A reading that lists two live candidates is more useful than one that picked the more quotable of them.

No invented base rates

The honest answer to how often a spike belongs to a given class is that this desk does not know. No sample here is representative, no venue coverage here is complete, and a percentage invented to fill that gap would be exactly the failure this site exists to describe.

Where a frequency matters to an argument, the argument is rewritten so it does not depend on one, or the limitation is stated in the paragraph itself.

Every class carries its exit

No class is published without the observation that would move it into a different one. That requirement does more work than any single check, because it forces the writer to decide in advance what evidence would have been needed, which is far harder than describing what was found.

Where no such observation exists, the correct output is undecidable, and undecidable is a finished reading rather than a failed one.