One hour, four assets, one direction

The complex topped together. Solana and Dogecoin made their highs in the 04:00 hour, XRP made its high inside the 05:00 hour at $1.6999, and then all of them turned in the same sixty minutes.

05:00 UTC hourMoveSpot valueTaker buy shareSession high
XRP8.11% lower$180.4 million50.6%$1.6999
Dogecoin6.78% lower$59.4 million51.8%$0.1008
Solana4.95% lower$164.8 million48.0%$102.74
Bitcoin1.39% lower$294.6 million45.8%$78,828

Read the third column against the second. Dogecoin fell 6.78% in an hour in which more than half of the value crossing the spread was buying. XRP fell 8.11% with buyers taking 50.6%. Bitcoin, which had the lowest buy share of the four at 45.8%, fell the least.

If the taker buy share were a direction indicator, that table would be impossible. It is not one, and this is the clearest demonstration of that we have been able to publish.

For scale on the XRP move: the still open Saturday session has a range of 23.36%. Across the previous 399 completed sessions only four were wider, and the most recent of those was August 20, two days ago, at 23.55%. The median daily range for XRP over that window is 4.20%.

The explanation we expected, and why we dropped it

The first draft of this article said it was a weekend. No US session, no ETF desks, no institutional bid, so a thin book and violent moves. It is the standard line, it is intuitive, and we were about to publish it.

Then we measured it, because the whole point of this site is measuring rather than repeating.

The test is simple. For every hour in the last 720, divide the value traded by the size of the absolute price move. That gives you dollars of volume per 1% of movement, which is a rough proxy for how much money it takes to shift the price. Higher means deeper. Then split by weekday and weekend and compare the medians.

AssetWeekday medianWeekend medianWeekend as % of weekday
Bitcoin$194.5M per 1%$193.3M per 1%99%
XRP$10.3M per 1%$8.9M per 1%87%
Solana$17.1M per 1%$15.7M per 1%92%
Dogecoin$4.0M per 1%$4.4M per 1%109%

Weekends are not meaningfully thinner. Bitcoin is within a percentage point of its weekday depth. Dogecoin is actually deeper at weekends on this measure. The largest gap is XRP at 13% shallower, which is nowhere near enough to turn a normal hour into an 8% hour.

So we deleted that draft. The convenient story did not survive contact with the numbers, and publishing it would have been repeating a cliché with our name on it.

They were deeper than usual, not thinner

Running the same calculation on the crash hours themselves makes the point sharper still.

Asset05:00 hour depthWeekend medianVersus normal
XRP$22.3M per 1%$8.9M per 1%2.5x deeper
Solana$33.3M per 1%$15.7M per 1%2.1x deeper
Dogecoin$8.8M per 1%$4.4M per 1%2.0x deeper
Bitcoin$212.3M per 1%$193.3M per 1%1.1x deeper

Every one of them took more money per percentage point than its own typical weekend hour. These were not air pockets where a small order fell through an empty book. They were large, genuine, expensive moves. XRP burned $22.3 million for each 1% it gave up, against a normal $8.9 million.

Which leaves the question sharper than when we started. If the book was not thin, and the aggressive buying and selling were balanced, what moved the price?

The side of the book nobody measures

Here is the resolution, and it is a property of the metric rather than a property of Saturday.

The taker buy share counts who crossed the spread. It says nothing whatsoever about where the spread was. Those are two different facts and only one of them is in the number.

A price falls when the resting bids get replaced lower. A market maker who was showing a bid at $1.68 pulls it and shows $1.66 instead. Nothing about that decision is a trade, so nothing about it enters our ratio. The next buyer who lifts an offer still counts as a buyer, and the price they buy at is simply lower than the last one. Repeat that a few hundred times in an hour and you get an 8% fall on a 50.6% buy share, with every single trade matched by a willing buyer.

The size of the resting book relative to the flow is what makes this possible, and that ratio is startling. This is a live snapshot taken while writing, at roughly 19:00 UTC:

AssetBids within 1% of midAsks within 1%Hourly value at 05:00
XRP$1.51 million$3.48 million$180.4 million
Solana$2.30 million$3.72 million$164.8 million
Dogecoin$0.67 million$0.98 million$59.4 million
Bitcoin$10.23 million$11.58 million$294.6 million

XRP's entire visible bid stack within 1% of the mid price is about $1.51 million. The hour it fell 8.11% traded $180.4 million. The flow was roughly 120 times the standing bid.

A book that small relative to the flow is not a wall that gets broken. It is a surface that is continuously consumed and rebuilt, and the price is wherever it happens to get rebuilt. The taker ratio describes the traffic. It does not describe the road.

One honest caveat: this snapshot is from 19:00, not from 05:00. Order book depth is not stored in public history, so we cannot show you what the bids looked like during the move itself. The ratio of book to flow is the structural point and it holds on any snapshot we take.

Three limits in nine days, and this one is different

We have now had to document three separate things our headline metric cannot see, and it is worth putting them together because the third is not like the first two.

LimitWhy it is invisibleWhat the number does
ETF creationsSettles off the public book entirelyShows nothing at all
Futures liquidationsExecutes on a different bookDirection right, size understated
Maker repricingNot a trade, so never enters the ratioReads even while price collapses

The first two were about venue. Money was moving somewhere we were not looking, and the fix in principle is to look there too, which is what we did on Thursday by pulling the perpetual book alongside the spot one.

This third one cannot be fixed that way. It is not on another venue. It is on the same book, on the other side, and it consists of orders that were placed and cancelled without ever trading. No taker metric on any venue can see it, because it is by definition the part of the market that did not trade.

That is a real boundary rather than a gap, and we would rather state it plainly than let readers assume the ratio carries information it does not carry.

What we cannot tell you

We would like to tell you whether an even buy share on a large down hour is unusual. We cannot, and the reason is worth showing rather than hiding.

Across the last 720 hours there have been exactly three hours in which XRP fell 2% or more, one for Solana, one for Dogecoin and none for Bitcoin. Several of those are the hours we are writing about. A base of one to three observations cannot establish what is typical.

This site made precisely that mistake on August 3, when we presented five falling gold buy share readings as an ongoing unwind and the pattern reversed the following day. We are not going to repeat it in the opposite direction by presenting three observations as a rule.

The mechanism does not need the sample. A ratio of aggressive buying to aggressive selling contains no information about the depth or placement of the resting orders those aggressors traded against. That is true by construction, on any day, in any sample size.

Bitcoin sat still, and the flows kept coming

Bitcoin is up 0.43% over 24 hours at $77,277, with a range of 3.04% against XRP's 23.36%. Perpetual open interest is 0.09% lower across 26 hours, which is as close to unchanged as this market gets, so neither leverage building nor a further unwind. Funding has now settled at exactly 0.0100% for four consecutive periods, which readers of yesterday's article will recognise as the venue ceiling rather than a signal about positioning.

Friday's ETF figure published while we were writing and extends the run to a fifth session.

DateUS spot Bitcoin ETFsIBIT share
August 17$297.56 million
August 18$189.30 million
August 19$517.19 millionAbout 55%
August 20$606 million83%, $503 million
August 21$307.5 million78%, $239 million
Five day totalAbout $1.92 billion

Yesterday we flagged the concentration in BlackRock's fund as worth watching. It has now run above 75% for a second day. We are reporting that as two days, not as a direction, for the same reason as the section above.

There is no flow figure for today. ETFs do not trade at weekends, which is the one genuine weekend effect in this article, and it makes today's move entirely a matter of what happened on exchange.

Where things stand

AssetPrice24h24h highOff the high
Bitcoin$77,277+0.43%$78,8282.0%
Ether$2,427.72+0.84%$2,546.784.7%
XRP$1.4918+9.39%$1.699912.2%
Solana$94.40+3.78%$102.748.1%
Dogecoin$0.09263+10.62%$0.10088.1%
BNB$698.51+3.68%$726.083.8%
Tokenised gold$4,5870.56% lower$4,6130.6%

The last column is the story of the day. Every major is up over 24 hours and every major is well below where it traded this morning. Total market capitalisation is $2.633 trillion, down 1.49%, which is a fall on a day when Bitcoin rose, because the alts gave back more than Bitcoin gained. Dominance moved back from 59.21% to 58.77%.

Fear and Greed reads 71, Greed, against 72 yesterday. The ETH to BTC ratio is 0.03142, up 0.26%. That is the fourth day we have reported it. Up, up, down, up. We said on Wednesday that two observations are not a trend, and four observations that alternate are not one either.

What we are not saying

We are not saying the alt move is over, or that it resumes. Today's measurements describe an hour that has already happened.

We are not saying the buy share is broken. It measured today's hour correctly: aggressive buying and selling were close to even, and that is a true and useful fact. The error would be reading direction into it, which is why we publish what it does not contain alongside what it does.

We are not setting a level on XRP, Solana or anything else.

We are not saying weekend markets are safe because depth holds up. Our test measured dollars per percentage point of movement in aggregate. It did not measure how that depth behaves under stress, and one Saturday is not a study.

What we are saying is narrow and, we think, worth the space. A metric that reads 50.6% during an 8% collapse is not malfunctioning. It is answering a different question from the one a reader might assume, and the difference between those two questions is the resting side of the order book, which no trade based measure can ever see.