A crowded market can still have an empty side.

A prediction market has two books. Every YES position is matched by someone holding NO, but the two are not mirror images: one side can be thousands of small holders and the other can be four accounts and a rounding error.

Concentration is how unevenly the money on a side is distributed. Low concentration means many holders with comparable stakes — a genuine crowd, and a price that reflects one. High concentration means a handful of accounts set the level, and their exit is the whole book moving at once.

This matters for a reason specific to prediction markets: the price is the crowd’s estimate, so the value of that estimate depends entirely on whether there is a crowd behind it. A 79% chance held by four wallets is not a 79% consensus. It is four opinions.

Below is a real, dated snapshot: the top 100 holders on each side of Will there be no change in Fed interest rates after the July 2026 meeting, captured Jul 12, 2026 from Polymarket’s public holder feed and priced to $5.9M of notional. It has since resolved; the numbers are a photograph, not a feed.

Read as one book it is unremarkable — low concentration, with the largest holder on 12.6% of the money. Split it and the halves stop resembling each other:

Gini runs 0 to 1: zero is every holder with an identical stake, one is a single holder with everything. The YES book comes in at 0.591 — uneven, as every real book is. The NO book is at 0.960, which is close to the theoretical maximum.

Top 1 and Top 5 are blunter and harder to argue with. One account holds 54.7% of the NO side. Five accounts hold 94.5% of it. On the YES side the same five hold 28.9%.

The verdict is a bucketed HHI — the same thresholds a competition regulator uses on market share, applied to a book instead of an industry. Both books together score low. The NO book alone scores high. That gap is the point of this page: the whole-market number is not wrong, it is answering a question nobody asked.

Across both books, 11 of 200 holders own half the money — 5.5% of the accounts. The median position is $7,791; the largest is $745,729.

Predm counts a holder as a whale above $100k of notional, and that bar is absolute rather than a percentile. A percentile always finds whales — the top 1% of any book is, tautologically, the top 1% — which makes it useless for the question people actually ask, which is whether there is real size here at all. A fixed threshold can honestly return none.

In this snapshot, holders above that bar account for 58.4% of both books combined and 86.4% of the NO side.

Gini is a single number for unevenness. It is good at telling you that a book is not a crowd. It is bad at telling you how it fails to be one. A Gini of 0.8 can be one giant holder and a long tail, or a tight oligopoly of five. Those are different trading problems.

HHI— the Herfindahl–Hirschman Index, borrowed from antitrust — squares each holder’s share and adds them up, so a single dominant wallet is punished more than a group of equals. Predm’s verdict word (low, moderate, high, extreme) is a bucketed HHI, using the same thresholds a regulator would recognise, applied to a book instead of an industry.

Top 1 and Top 5 are the blunt instruments, and they are the ones you should be able to say out loud. “One wallet is more than half of NO” does not need a coefficient. The coefficients exist so that two books can be compared when the blunt story is less obvious.

None of these should be quoted for the market as a whole if you can help it. The whole-market number in this snapshot is the one that looked unremarkable. The useful numbers were the split.

A candle tells you what the last price did. Concentration tells you whether that last price is allowed to mean anything. A 12-point bullish day on a YES book that is a genuine crowd is a change in a consensus. The same candle on a YES book that is four accounts is a change in four accounts.

That is why this page sits next to the candlestick guide rather than in a glossary. Technical analysis for prediction markets treats concentration as a qualifier on every pattern, not as a separate product feature. A breakout through 60% is not a breakout until you know who had to be moved.

Whale influence, in this framing, is not a personality. It is a share of notional above a fixed dollar bar. A percentile always finds a whale. A fixed bar can honestly return none, which is the only way the word stays meaningful.

It is a top-N sample, not a census. Polymarket’s public feed returns the largest 100 holders per side, so the tail below them is invisible. That biases every metric here upward: the real books have more small holders than this can see, so true concentration is somewhat lower than these numbers. Predm labels these approximate for that reason, and so does this page.

One wallet is not one person. Concentration measures wallets. A single trader running five wallets reads as five holders, and a custodian holding for many reads as one. Treat it as a lower bound on how concentrated the real ownership is.

It is a snapshot, not a feed. Everything above was true on Jul 12, 2026. A book can change shape in an afternoon, and this market has since resolved.

Concentration is not a signal. It does not say which way a market is going. It says how much weight to put on the fact that it is where it is — which is a different, and usually more useful, question.

What is holder concentration on Polymarket?

How unevenly the money on a side is distributed. Low concentration is a crowd with comparable stakes. High concentration is a handful of wallets setting the level. Always ask it per side: YES and NO are not mirror images.

What is a good Gini for a prediction market?

There is no good number in the abstract. Gini 0 is a theoretical crowd; Gini 1 is one holder. Real books sit in between. The useful comparison is between the two sides of the same market at the same instant, not against a universal target.

Does a whale on Polymarket move the price?

A large holder is a large part of the price. Their exit is the book moving. That is not the same as predicting which way they will go. Concentration tells you the blast radius, not the direction.

Why not just look at the top trader leaderboard?

A leaderboard ranks realised and unrealised P&L across markets. Concentration describes the shape of one book. A profitable wallet can be small in this market; a dominant wallet in this market can be unremarkable on the board.