Top Holder Concentration: Classify Before You Read It

Top holder concentration is the combined balance of the largest token accounts divided by supply. On a young Solana token that list is typically led by a liquidity pool, followed by a program vault and sometimes a burn address. None of those is a holder. Until every address in the list is classified, the percentage is a number about accounts rather than about ownership, and reading it directly produces confident wrong conclusions in both directions.

What the raw percentage contains

The computation is simple: take the largest N balances for a mint, sum them, divide by supply. Two of those inputs are choices. N is usually ten or twenty and the choice is rarely displayed. The denominator may be total supply or circulating supply, which on a token with a large excluded allocation produces very different percentages from identical balances.

Neither choice is where the real problem lives. The real problem is the numerator, because a token account is a token account regardless of what controls it. The query cannot distinguish a person's wallet from a pool's reserve account, and the largest balance on a newly launched token is frequently the pool.

On a freshly launched token, the single largest holder is often the market itself. Counting the liquidity pool as a concentrated owner inverts the meaning of the figure: the account holding the tokens everyone can buy is recorded as the account hoarding them.

Classify first, then compute

The discipline is to treat the top holder list as raw material rather than a result. Pull the list, classify each entry, then compute. Classification takes a few minutes for twenty addresses and it changes the answer more than any other adjustment available to you.

The signals you use are the owner program of the account, the pattern of transfers in and out, the age of the account relative to the token, and whether the address appears in the same role for other tokens. An address that holds reserves for many different mints on the same program is a pool or a router account. An address funded once and never active again is a passive holding. An address with constant small outflows to many destinations is behaving like a custodian or a distributor.

The address categories that matter

How each category should be treated when recomputing concentration, and the consequence of getting it wrong
Category	Recognition signal	Include in concentration	Consequence if miscounted
Liquidity pool reserve	Owned by a DEX program, mirrored by the paired asset	No	Overstates concentration, often by tens of percent
Program vault or escrow	Program-derived owner, single large balance, no swaps	No, but note the program	Reads as a whale with no ability to sell
Vesting contract	Program owner with a schedule, periodic outflows	No, record schedule separately	Confuses locked supply with liquid supply
Burn address	Known incinerator or authority-less account	No	Overstates both concentration and available supply
Exchange omnibus	High transfer frequency, many counterparties, long history	Yes, flagged as aggregate	Reads as one whale when it is many users
Team or treasury wallet	Funded at launch, low activity, round amounts	Yes	The most consequential category to miss
Coordinated trading wallet	Peer wallets funded from a shared source in a short window	Yes, grouped	Understates concentration when counted individually
Independent holder	Acquired by swap, unrelated funding, varied activity	Yes	The baseline the other categories are separated from

Worked example: a top-ten list reclassified

Illustrative arithmetic on an invented holder list

All figures are chosen by the desk to demonstrate the method. They describe no real token and are not observed data.

A dashboard reports top-ten concentration of 61 percent of total supply. The ten balances, as percentages of supply, are: 24.0, 11.5, 6.2, 4.8, 4.0, 3.1, 2.6, 2.2, 1.4, 1.2.

Classification: the 24.0 percent account is the primary liquidity pool. The 11.5 percent account is a burn address holding tokens removed at launch. The 6.2 percent account is a second pool on another venue. The remaining seven are ordinary wallets, and four of them, holding 4.8, 4.0, 3.1 and 2.2 percent, were funded from the same address within eleven minutes of each other.

Recompute. Remove pools and burn: 61.0 - 24.0 - 11.5 - 6.2 = 19.3 percent held by classifiable wallets. That is the honest concentration figure, and it is under a third of the headline.

Now group the four commonly funded wallets: 4.8 + 4.0 + 3.1 + 2.2 = 14.1 percent, which is 73 percent of the classified concentration sitting behind one funding source. The dashboard showed ten separate entries; the funding graph shows something closer to four independent parties.

Both adjustments matter and they pull in opposite directions. The raw figure was far too high as a measure of ownership concentration, and far too low as a measure of control concentration.

The figure errs in both directions

This is the part most treatments of concentration miss. The error is not one-sided, and correcting only one side produces a different kind of wrong answer.

Upward error. Pools, vaults, burns and staking contracts appear as large holders while representing supply that is either available to everyone or unavailable to anyone. On a new token this can be most of the reported percentage.
Downward error. One operator splitting a position across thirty addresses appears as thirty small holders and never enters the top-ten list at all. The percentage looks healthy precisely because the position was distributed.

The downward error is the harder one, and it is why concentration cannot be read without some form of relatedness analysis. It is also why the figure is weakest exactly where you most want it to be strong: a deliberately structured position is designed to not show up in a top-ten list, and the tooling to spread supply and activity across many addresses is ordinary. Anyone operating automated volume management across a wallet set produces this shape as a side effect of normal operation, without any intent to disguise ownership.

Funding graphs and coordinated wallets

A funding graph traces where the SOL in each wallet came from. Every wallet on Solana needs SOL to pay fees and to open token accounts, and that SOL arrived from somewhere. Following it one or two hops back is the single most informative thing you can do with a holder list.

Take the classified wallets from the top twenty. Ignore pools, vaults and burns; they have no funding story worth tracing.
Find the first SOL inflow to each. Order the account's transaction history ascending and locate the first transfer that funded it.
Record the source address and the timestamp. These two fields are the entire dataset.
Group by source. Any source funding more than one wallet in your list is a relatedness candidate.
Check the timing spread. Wallets funded within minutes of each other, in similar amounts, are a much stronger group than wallets funded months apart.
Note where the trail ends. A funding transfer from a large, high-traffic address is usually an exchange withdrawal and breaks the trace. Record it as unresolved rather than as independent.

Two cautions keep this honest. A shared funding source does not prove common ownership; wallets funded by the same exchange hot wallet are unrelated. And an unresolved trail is not evidence of anything. The technique narrows the space of explanations, and narrowing is the most any on-chain method can do about identity.

Concentration on curve-launched tokens

Tokens launched on bonding-curve platforms have a specific concentration profile worth knowing. Supply is minted in full at creation, and while the token is on the curve, most of it sits in the curve's own account. Concentration measured then is almost entirely an artefact: one account holds nearly everything, and that account is the mechanism through which anyone can buy.

The figure only becomes meaningful after migration to a standard pool, and even then the first hours are dominated by the new pool account. Early buyers who accumulated on the curve appear as the first genuine concentration, and their share is a real signal about how the distribution phase went. Details of the curve mechanism itself are documented by the platform, and the practical point for metric readers is simply that a concentration percentage taken before migration is measuring the venue rather than the holders.

Concentration over time is the real measurement

A single concentration reading is a snapshot of a distribution that was produced by a history you did not observe. The same 19 percent can be the residue of a wide distribution that consolidated, or the starting position of an allocation that has not moved at all. Those are opposite situations and the snapshot cannot separate them.

Two readings can. Take the classified figure at two dates, with the same classification rules and the same denominator, and the direction tells you something the level never will. Three movements are worth naming, and each has a fairly specific meaning.

Falling concentration with rising funded holders. Supply moving from large accounts into a widening base. This is distribution in the neutral sense, and whether it is welcome depends entirely on who is on the receiving end.
Falling concentration with a flat holder base. Large balances splitting into smaller ones without new participants arriving. Often ordinary account management, and sometimes a position being spread across addresses. The funding graph is what separates the two.
Rising concentration with flat supply. Accumulation. The interesting question is whether the accumulating accounts share funding, because a single party accumulating and twenty parties accumulating look identical in the percentage.

The mechanics of doing this are unglamorous and matter more than the interpretation. Use the same N, the same denominator and the same exclusion rules on both dates, record the block height for each reading, and keep the classified list itself rather than only the resulting percentage. Without the list you cannot tell whether a change came from balances moving or from your classification of a borderline address changing between passes.

Two readings a fortnight apart, on a token you are actually watching, produce more usable information than a hundred single-reading comparisons across tokens you are not. This is the general shape of most on-chain analysis: instruments are noisy in absolute terms and much better behaved in differences, provided the instrument itself does not change between measurements.

What a classified figure can and cannot support

Once classified, concentration supports a narrow set of claims. Being precise about the boundary keeps the analysis defensible.

It supports: a statement about how much supply could be sold by a small number of decision-makers, and how that compares with the depth available to absorb it.
It supports: a comparison over time for the same token, since the same classification applied at two dates is a like-for-like measure.
It does not support: any claim about intent. A large holder who has not sold is not thereby a holder who will not sell.
It does not support: a claim of identity. Grouped wallets are wallets with correlated funding, which is a behavioural finding, not an ownership finding.
It does not support: cross-token league tables, unless the same classification and the same N were applied to both.

The most useful pairing is concentration against depth. Classified concentration tells you how much supply sits with a few decision-makers; the depth analysis in the liquidity ratio tells you what happens if a fraction of it is sold. Neither is interesting alone; together they describe a specific, quantifiable fragility.

A classification pass you can repeat

The pass below takes about fifteen minutes for a token and produces a record rather than an impression. Repeating it at intervals for the same token is where it earns its value, because the change over time carries more information than any single reading.

Pull the top twenty accounts. Record address, balance and percentage of total supply.
Classify each using the category table. Record the category and the signal you used, so a later reader can disagree with you specifically.
Recompute concentration excluding pools, vaults, vesting and burns. This is your classified concentration figure.
Trace funding for every remaining wallet. One hop is usually enough, two if the first hop is small and recent.
Group by common funder and recompute again. Report both the ungrouped and grouped figures; the gap between them is the finding.
Record the denominator you used. Total or circulating, stated explicitly, because it changes the percentage.
Note unresolved addresses. Any wallet you could not classify, with its percentage, so the uncertainty is visible in the record.

The output is three percentages and a list of unknowns: raw, classified, and classified-and-grouped. A token where those three numbers are close together has a genuinely dispersed holder base. A token where they diverge sharply has a structure the headline figure was never going to show you, and that structure belongs in the scorecard as evidence rather than as a conclusion.

Questions the desk gets asked

What is top holder concentration?

The share of supply held by the largest token accounts, usually the top ten or twenty, expressed as a percentage. The count of accounts included is a choice the provider makes and rarely displays, so two dashboards can report different concentration for the same token without either being wrong.

Does the liquidity pool count as a top holder?

It appears in the list because a pool holds its reserves in a token account, but it is not a holder in any useful sense. Pool reserves are available to the market by definition. Leaving the pool in the calculation inflates apparent concentration, sometimes dramatically on young tokens.

What top holder percentage is dangerous?

No single number works as a threshold, because the answer depends entirely on who the holders are. Twenty percent held by a locked vesting program is a different risk from twenty percent held by four wallets funded from one source an hour before launch.

How do I tell if wallets are related?

Trace the SOL that funded each one back a hop or two and look for common sources, and compare timing. Wallets funded from the same address within a short window and behaving similarly afterwards are a strong constraint. Funding via an exchange withdrawal breaks the trace.

Can concentration be hidden?

Yes, by splitting a balance across many addresses. That is why concentration should always be read alongside funding-graph evidence and wallet-count metrics rather than on its own, and why a low top-ten percentage is not by itself reassuring.

Do burn addresses count in concentration?

They should not, because burned supply cannot trade. Many providers include them, which inflates the percentage. Where a large burn exists, recomputing concentration against supply excluding the burn usually changes the figure substantially.

Is high concentration always bad?

No. A treasury under multisig control with a published policy, or supply in a vesting program with dated unlocks, is concentration with a visible constraint on it. The risk is unconstrained concentration, which is a property of the holder rather than of the percentage.