What a Top 10 Holders Figure Actually Measures

A top-10 holders figure is the share of a coin’s supply held by its ten largest addresses, and some of those addresses can be trading pools rather than people. That changes what the number is telling you.

Updated 16 October 2026: the measurement now runs to 16 October with a new section on what changed after 5 September, the data providers are named, the note about this site’s safety page reflects its 13 October correction, and a short version and three common questions were added.

On $CATE, over four weeks of hourly readings to 5 September 2026, the number moved 2.6 percentage points, and the thing it tracked most closely was not anybody buying or selling. It was the quantity of tokens parked inside the trading pools, which rises automatically when the price falls.

That matters because the figure is normally read as a safety signal. A rising concentration number is taken to mean insiders are accumulating. On the evidence below, most of the movement in this particular number meant something much duller, and knowing the difference changes what the chart is worth to you.

What the number is supposed to tell you

Concentration is the risk that survives every other check. A coin can have no ability to print new supply, no hidden sell tax and a burned trading pool, and still be one wallet’s plaything if that wallet holds a large enough slice. So price sites publish a top-10 share: add up what the ten largest addresses hold, divide by total supply, show a percentage.

The reading people take from it is intuitive. A low number means the supply is spread across many hands, so no small group can move the price at will. A high number means a handful of addresses could sell into your bid whenever they felt like it. Watching the number rise is therefore supposed to be an early warning that a few large holders are quietly building a position.

This site has published that figure since 9 August 2026, reading it hourly from GeckoTerminal, the same public source that the live stats page shows. What follows uses 1,344 of those hourly readings, running from 9 August to 5 September 2026. Nothing here is a forecast, and none of it says what the number will do next. It is only a record of what it did. A later section carries the same record forward to 16 October 2026.

The crowd grew 68% and the figure went the other way

The obvious explanation for a moving concentration figure is a moving crowd. If the number of holders grows, each new arrival owns a little of the supply, so the largest ten should command a smaller share. Dilution alone should push the line down.

Over this window the crowd grew a great deal. The holder count read 70,386 on 9 August and 118,480 on 5 September 2026, a rise of 68%, sourced from GeckoTerminal throughout. If the concentration figure were mostly a function of how many people hold the coin, that growth should have dragged it steadily downward.

It did not. Across the whole window, taking the daily median of each, the correlation between the level of the top-10 share and the level of the holder count is negative 0.11, which is close to no relationship at all. The two lines cross each other repeatedly and in both directions. The concentration figure fell hardest during the days when holders were arriving fastest, then rose again through a fortnight when the holder count was essentially flat.

Note also that the crowd did not only grow. The holder count peaked at 119,230 on 30 August 2026 and was slightly lower by 5 September, so this window is not a clean upward line, and it should not be described as one.

Three stretches, and none of them look like accumulation

Splitting the window into its three obvious regimes makes the mismatch easier to see than any correlation figure does. The first stretch is a stall, the second is a rush of new arrivals, the third is a long plateau.

If concentration tracked accumulation by large holders, the middle stretch is where you would expect trouble. It is the period when the most money was moving and the most new wallets appeared. Instead the concentration figure fell by nearly two points during exactly those three days, then recovered over the following fortnight while almost nothing was happening.

Read as a whale story, that sequence makes no sense. Large holders would have to have sold heavily into the busiest days of the month and then bought steadily back through the quiet ones, ending roughly where they started. Read as something mechanical, it makes immediate sense, and the next two sections show what the mechanism is.

There is a further reason to doubt the accumulation reading. The three stretches line up much more tidily with the price than with anything else. The stall covers a long slide, the rush covers the sharpest recovery in the window, and the plateau covers a drift back down. Whatever is driving this figure appears to be responding to price rather than to the arrival or departure of people, which is a strange property for a measure of who owns what.

A liquidity pool is a holder

Here is the part that reframes the whole number. A trading pool is an address, and it holds tokens. Nothing in the arithmetic of a top-10 list distinguishes an address that belongs to a person from an address that belongs to a machine, so the pools that let you buy and sell the coin are eligible to be counted among its largest holders like anybody else.

That would be a footnote if pool inventory were stable. It is not, and the reason is the design of the pool itself. These pools hold two things at once, the coin and the network’s own currency, and they quote a price from the ratio between them. When people buy, coins leave the pool and currency goes in, so the pool ends up holding fewer coins. When people sell, coins go into the pool and currency comes out, so the pool ends up holding more.

The consequence is worth stating slowly, because it inverts the usual reading. A falling price means selling pressure, which means coins accumulating inside the pool, which means the pool’s share of total supply goes up. If that pool sits among the ten largest addresses, then the published concentration figure rises during a decline without a single person having decided anything. The number goes up precisely when the intuitive reading says whales are moving in, and it is the machine filling up.

This site has covered the pools themselves before, in the sense that a burned pool is not necessarily a deep one. What follows is a different question: not how much is in there, but whether what is in there is quietly being counted as a holder.

The measurement

To test that, we need the pool’s token inventory over time, and the hourly log stores enough to reconstruct it. For a pool of this kind, the value of the coins held inside is half the pool’s total quoted liquidity, so dividing half the liquidity by the price of the coin gives the number of coins sitting in the pool.

That derivation is worth checking rather than trusting. Taking a single reading on 5 September 2026 and running the calculation returns 29,304,018 coins for the main pool. DexScreener, a separate market data source, reported the pool’s actual holding at that same moment as 29,359,222 coins. The two agree to within 0.19%, which is close enough to treat the reconstruction as sound. Both figures drift through the day, which is the whole point of what follows.

Applying it across the log gives a pooled-inventory series to set against the concentration series. One caveat had to be handled first: the liquidity field changed meaning on 11 August 2026, from the main pool alone to every pool at once, so the two days before that change are excluded rather than silently compared against the rest. That leaves 1,309 hourly readings across 26 days.

The result is unambiguous. Taking the daily median of those readings, pooled inventory and the top-10 share move together with a correlation of positive 0.78, meaning about 61% of the variation in the published concentration figure is accounted for by how full the pools happen to be. Pooled inventory against price is negative 0.96, close to a mechanical relationship, exactly as the design of the pool predicts. Against those, the holder count’s negative 0.11 is not a competing explanation.

How much of the top ten is not a person

Correlation says the pools are moving the figure. The next question is how big a bite they take out of it.

On 5 September 2026, total supply stood at 964,162,080 coins, and the main pool held 29,359,222 of them. That is 3.045% of the entire supply sitting in one address that is a piece of software. The published top-10 share that morning was 15.0994%. If that pool is one of the ten being counted, it alone is roughly a fifth of the headline concentration number, and it belongs to nobody.

There is a second clue pointing the same way, and it is the neatest number in this whole exercise. Fitting a straight line through the two series gives a slope of 0.55, meaning the top-10 share moves about half as far as total pooled inventory does. Separately, the main pool accounted for 55.7% of all liquidity across the token’s pools that day. Those two figures matching is what you would expect if the main pool is inside the top ten and the smaller pools are not, because then only the main pool’s share of the movement would show up in the figure.

That is an inference from two independent measurements agreeing, not a proof, and the next section is about why it has to stay an inference.

What we could not check

The honest limit of this work is that the ten addresses are not public. GeckoTerminal, which publishes the percentage, does have an endpoint listing the individual top holders, and requesting it returns an authorisation error, because that list sits behind a paid tier. It still did on 16 October 2026. The percentage is free and the composition of it is not.

The usual fallback is to ask a public node for the largest accounts directly. That was tried on five separate public endpoints on 5 September 2026. One rate-limited the request, three now require an account for this class of query, and one was unavailable. So the specific claim that the pool address sits inside that specific list of ten is supported by the correlation, by the size of the pool relative to the figure, and by the slope agreement, but it has not been confirmed by reading the list.

One further caution about the figures on this page. Only the top-10 share is logged hourly. The deeper bands, ranks 11 to 20 and 21 to 40, exist only as a present-day snapshot and were not recorded over time, so they can be quoted for a single dated moment and must not be drawn as a series. On 5 September 2026 they read 6.473% and 6.705%, with 71.72% of supply outside the top forty addresses altogether.

What happened after 5 September

The measurement above stopped on 5 September 2026. The hourly log kept running, and by 16 October 2026 it held 3,370 top-10 readings. Running the same comparison on the longer record keeps one half of the story and breaks the other, and it is worth being plain about which is which.

The half that holds is the short-term movement. Comparing each day’s change in the top-10 share with each day’s change in pooled inventory gives a correlation of positive 0.72 from 6 September to 16 October, close to the 0.79 the same test gives for the original window. So a sudden one-day jump in the figure is still, most often, the pools filling up.

The half that breaks is the slow drift. Between 5 September and 16 October the daily median top-10 share climbed from 15.05% to 16.58%. The hourly reading first went above the old window’s highest value of 15.78% on 8 September, and on 16 October 2026 GeckoTerminal put it at 16.53%. Over the same weeks the pools emptied, from about 5.4% of supply to about 3.2% on daily medians, because the price ended the stretch higher than it started. If the pools explained the level, the figure should have fallen. It rose, and across these six weeks the two levels show no relationship at all, a correlation of about negative 0.13.

The main pool tells the same story on its own. On 16 October 2026 DexScreener reported it holding 22,447,306 coins, about 2.3% of supply, against 3.0% on 5 September. If it is still one of the ten, the other nine addresses held about 14.2% between them on 16 October, against about 12.1% on 5 September. That “if” is the same unconfirmed inference as before, so treat the split as an estimate.

One of the three clues behind that inference has also weakened. On 16 October 2026 the main pool held 75.1% of all liquidity across the token’s pools, according to the DexScreener feed behind the stats page, up from 55.7% on 5 September. If it alone sat in the top ten, the figure’s day-to-day moves should now follow about three quarters of the pooled moves. Since 6 September they have followed about 0.47 of them, so the neat match described above no longer appears, and the inference now rests on the correlation and the size of the pool.

What the log cannot say is who. The ten addresses are still behind GeckoTerminal’s paid tier, and on 16 October 2026 four public Solana endpoints refused the largest-accounts query: one rate-limited it and three required an account. The holder count, also from GeckoTerminal, rose over the period from about 118,000 to about 141,000, and the two levels look linked, but their day-to-day changes show no connection at all. That is what two unrelated lines that both trend upward look like. The rise could be one large wallet, several, an exchange wallet, or something else that holds tokens. This site cannot tell which, and does not guess.

The useful lesson is narrower than the original one. A short jump, especially during a price fall, is still often the pools. A steady climb while the price rises and the pools drain is not the pool effect, and it is the kind of move worth looking at directly rather than explaining away.

What this changes about reading the number

The practical version is short. A top-10 concentration figure on a coin whose trading pools are large enough to enter that top ten is partly a price chart wearing a disguise. When it climbs during a decline, the first thing to rule out is that the pools are filling up, and that is the least alarming explanation available.

It does not make the figure useless. A genuinely dangerous concentration, one address holding a third of everything, would show up as a level so far above this range that no pool effect would explain it. The figure is doing its job at the extremes. It is in the ordinary range, where people read a one-point move as a signal, that it is mostly reporting the weather.

This site corrected one of its own pages on the strength of that. The five checks that actually matter once recorded a top-10 share of about 13% on 9 August 2026 and offered a rule of thumb that under about 15% counts as distributed. By 5 September 2026 the figure read above that line with nothing in this measurement to suggest a group had taken control, so a threshold that routine pool movement could carry a coin across was the wrong tool. On 13 October 2026 that page dropped the fixed threshold. It now treats the figure as a rough signal and links back here.

If you want to see where the figure stands right now rather than on the day this was written, it is on the live stats page. If you would rather stop relying on anybody’s published percentage, including this site’s, checking a Solana token yourself covers reading the underlying data directly.

The questions people ask next