How much money it takes to move a crypto price by 2%, token by token

Bar centred on the panel median · sort by

RAIN, 0.04 bps on 13 October 2026. $7.5B of market cap across 10 markets, 62% of supply in circulation. It takes $28K to move its price by 2%, which is 0.04 bps of its market cap. The panel median, 40 bps, is 1,072 times higher. 10 markets carry a size of that order.

HYPE, 3.8 bps on 13 October 2026. $18.5B of market cap across 70 markets, 23% of supply in circulation. It takes $30.3M to move its price by 2%, which is 3.8 bps of its FDV. The panel median, 40 bps, is 11 times higher. Less than 30% of the supply circulates, so the ratio is taken on fully diluted value ($79.3B) rather than on market cap; on market cap alone it would read 16 bps.

IO, 619 bps on 13 October 2026. $70M of market cap across 61 markets, 52% of supply in circulation. It takes $4.3M to move its price by 2%, which is 619 bps of its market cap: 15 times the panel median of 40 bps.

These are the two lowest and the highest ratios on this reading, and the names change from one reading to the next. A ratio measures how easily a price moves for the size shown, nothing else.

2% market depth over time · dollars · BTC shown, pick another

ADAAKTATHBCHBNBBTCCCDOGEETHGMXHNTHYPEIOLINKMORPHOPUMPRAINRENDERRIVERSOLTAOTRXUNIVELVETXLMXMRXRPZEC

The record, reading by reading, since 9 September 2026. Each point carries the exact minute it was read, and each asset has its own scale. Drawn on the venue list frozen on 9 September 2026. On this reading, 93% of that list was read for BTC (288 of 309 venues). Every page the source serves is read, so the 21 missing venues are ones the source did not return this time, or returned flagged stale or anomalous and therefore dropped before the sum. 104 readings on the frozen list so far. Hover a point for its value.

Why. A token's market cap is the last price paid, multiplied by the number of coins in circulation. Its fully diluted value is the same price multiplied by every coin that exists, including those not yet in circulation. Both say what the asset is worth on paper. Neither says how much of it can be sold, nor at what price.

How. For every spot market kept in the count, we read how many dollars of resting orders it would take to push the price up by 2%, and we sum across those markets. That figure alone does not compare across assets: $1.20B on Bitcoin and $28K on RAIN cannot be put side by side. So we divide it by the asset's size, market cap or fully diluted value when less than 30% of the supply circulates, and only then do the assets become comparable.

Read it the other way round. An order that is a tenth of the figure shown fits inside the 2% band, if it could be spread across every venue at once, which the first limit below rules out. The figure is the ceiling of what one buyer could take, not what any buyer will get.

This is a ceiling, not a promise. We add up every market as if you could buy everywhere at the same second. Nobody can do that. In practice a buyer hits a handful of venues, so the amount actually available is lower than the one shown.

Spot only. Only spot markets, where the token itself changes hands. Anyone betting on the price without ever holding the token is outside this calculation, and on the large assets there are many of them.

It is a snapshot. This depth evaporates precisely on panic days, which are the days everyone wants to sell at once. A figure measured today says nothing about what will be there on that day.

Nothing is left unread. Every venue the source lists is read, to the last page. On 13 October 2026 that was 14 pages on Bitcoin and 24 on Ether, 98 reads in all. Until 9 September we stopped early, once a page added less than 1.5% of what was already counted, and that cut was wrong twice over: it was measured on a total that included markets we then discard, and it assumed each further page adds less, which the source does not guarantee since it orders venues by volume and not by depth. Ether was losing 19% of its depth to it.

A few venues are counted while stale. The source marks a book stale and we drop it, but a venue that quietly stops publishing keeps an unflagged book. On this reading 5 of the 28 assets carried at least one counted book older than six minutes, the oldest at 6.6 days on ATH. Their depth is counted as if it were current. It is a tail, not the bulk: the median counted book on the panel is 3 minutes old.

Some books never move. Over 7 days and up to 19 readings, 24 Bitcoin and Ether markets changed their depth by less than 1%, while the source kept refreshing them, and 10 returned one identical figure every single time. A book that does not move while it is being re-read is not one we can vouch for. Dropping every market whose 2% book is larger than its whole day of trading, the most conservative test we ran, takes the median of the 28 assets above, on this reading, from 40.4 to 29.5 basis points, and it takes more from the thin assets than from Bitcoin. The figure above still includes them. We are now measuring the difference venue by venue, at the venue, and we will print what it changes rather than quietly restart the record.

This is not a ranking of good and bad. A low figure is not a flaw, a high one is not a merit. We measure how easily a price moves, nothing else. Nothing here tells anyone to buy or sell anything.

Depth history exists. It is sold, not published. Order book history is an institutional product. Kaiko sells market depth computed from order book snapshots taken at least once a minute, and Amberdata sells order book history at granularities from one minute to one day (both product pages read on 9 September 2026). The free sources serve the book as it stands: CoinGecko, which feeds this page, and Binance's public spot API, both checked on 9 September 2026, offer no way to ask for yesterday's book. That is why the $795M we published for Bitcoin on 11 August, in our guide to what a token is worth, could not be replayed: the reading was a photograph, and there was no negative. This record, freely readable, starts on 9 September 2026.

How fresh it is. Computed at each pass, over the books actually counted. On this reading the median counted book was 3 minutes old, and 23 of the 28 assets had every counted book under six minutes. 5 did not: the oldest counted book was ATH at 6.6 days, BNB at 6.6 days, UNI at 5.9 days, and others. The source flags books it judges stale and we drop those before the sum, but that flag does not catch a venue that has simply stopped publishing. That tail is the sixth limit below.

What is 2% market depth? The dollar value of resting orders within 2% of the current price. On the buy side it is what it would take to push the price up by 2%; that is the side Crypto Deep Research sums, across every spot market kept in the count. On 13 October 2026, moving Bitcoin by 2% would have taken $1.20B of resting orders across 348 markets.

Why divide it by market cap? Because dollars of depth do not compare across sizes. Dividing by the asset's size gives a ratio in basis points that does. Across the 28 assets Crypto Deep Research read on 13 October 2026, the median is 40 bps.

Why does RAIN read 0.04 bps? RAIN reads 0.04 bps on 13 October 2026 because $7.5B of market cap sits on 10 markets: $28K of resting orders would move the RAIN price by 2%, 1,072 times less, relative to size, than the panel median of 40 bps.

Where can order book history be obtained? From vendors, under contract: Kaiko and Amberdata hold it at minute granularity. CoinGecko, the free source behind the Crypto Deep Research panel, serves no past book, and neither does Binance's public spot API (both checked on 9 September 2026). Crypto Deep Research keeps its own readings from 9 September 2026 onward at cryptodeep.io/blog/crypto-market-depth/.

The raw series (one line per asset and per reading, JSONL), every reading since 9 September 2026, and a panel built on your own assets are available on request from Crypto Deep Research. How the measurement is made: method.