The break that lasted part of a day

SessionHighCloseChangeDaily buy share
Mon Aug 24$80,000.00$78,993+1.62%49.8%
Tue Aug 25$81,273$78,5390.57% lower51.0%
Wed Aug 26, partial$79,252$78,3980.18% lower49.2%

Tuesday is the session worth studying. Bitcoin cleared the round number that had stopped it the day before, ran $1,273 past it, and then spent the remaining twenty one hours giving it back and more. A session that made a new high for the move closed red.

The top itself was undramatic. The five minute candle that printed $81,273 at 02:40 UTC traded $26.3 million at a 55.1% buy share. Nothing about that reading flags a top. Compare it with Monday, where the candle after the $80,000 touch read 19.9%. Tuesday had no such moment. It simply stopped going up and then bled for a day.

That difference is the reason for the rest of this article. Sometimes a high is marked by a violent flow event you can see in the tape. Sometimes it is marked by nothing at all, because the thing that stopped it never traded.

Where highs actually stop

Monday's high was $80,000.00. Not $80,001, not $79,998. Exactly the round number, to the cent. On August 19 the daily high was $70,000.00, also exact. Two exact $10,000 landings inside a week is the kind of coincidence worth measuring rather than remarking on, so we measured it.

We pulled 999 completed daily candles for four assets and asked a simple question: how often does the high land precisely on a round multiple?

AssetRound stepHighs landing exactly on itLows landing exactly on it
Bitcoin$1,00016 of 999 (1.60%)10 of 999 (1.00%)
Ether$1006 of 999 (0.60%)11 of 999 (1.10%)
Solana$105 of 999 (0.50%)13 of 999 (1.30%)
XRP$0.103 of 999 (0.30%)0 of 999

Set that against chance. Bitcoin trades in one cent increments, so there are about 100,000 possible prices between one $1,000 multiple and the next. If highs fell wherever they liked, you would expect roughly 0.001% of them to land exactly on the multiple. The observed rate is 1.60%, which is on the order of a thousand times more often.

And it is not a Bitcoin quirk. Every asset we tested shows it, on both sides of the candle, at its own natural round step. Ether at $100, Solana at $10, XRP at ten cents. The XRP and Solana counts are small enough that we would not read anything into the individual numbers, but the direction is the same in every case.

The shape says wall, not attraction

There are two ways prices could relate to round numbers. They could be broadly drawn toward them, clustering in the neighbourhood. Or they could stop dead on them and otherwise keep away. The distribution distinguishes these cleanly.

Distance of high from nearest $1,000ObservedExpected if uniform
Exactly on it1.60%About 0.001%
Within $5, not exact1.20%1.00%
$5 to $253.00%4.00%
$25 to $10011.91%15.00%
$100 to $25029.13%30.00%
$250 to $50051.55%50.00%

Read the middle rows. In the bands from $5 to $100 away from a round number, highs turn up less often than chance predicts, not more. The median distance from a high to the nearest $1,000 is $270, slightly above the $250 a uniform spread would give.

So there is no general gravitational pull toward round numbers. There is an enormous spike exactly on them and a small hollow just around them. That is what a wall produces. A large resting sell order at $80,000.00 lets buyers trade up into it and fill against it, and nothing prints higher until it is eaten. The high is recorded at exactly the wall. If the wall is big enough to turn the market, price then backs away and does not spend time in the neighbourhood, which is the hollow.

This connects directly to what we published on Saturday. That article argued the taker buy share cannot see the resting side of the book, because orders that are placed and cancelled never trade and so never enter the ratio. The resting book is invisible to our metric. But it is not invisible everywhere. It leaves a fingerprint in the price record, and the fingerprint is these 16 candles.

The book right now

If the explanation is right, the walls should be visible in a live snapshot. We took one at roughly 19:00 UTC while writing, reading the full depth within 2.5% of the mid price of $78,410.

SideLargest resting orderSizeValue
Asks$79,000.0019.658 BTC$1.55 million
Bids$77,500.0021.058 BTC$1.63 million
Bids, third largest$77,600.008.337 BTC$0.65 million

The single largest sell order sits at an exact $1,000 multiple. The single largest buy order sits at an exact $500 multiple. The third largest bid sits at an exact $100 multiple. Every one of the other large orders in the top six on each side is at an ordinary price like $79,268.81 or $78,107.24.

One snapshot proves nothing on its own, and we are presenting it as an illustration rather than as evidence. The evidence is the 999 candles. But it is the illustration the statistics predicted, and it took one query to find.

The competing explanation, tested

Tuesday's high of $81,273 is not a round number, so nothing above explains it. We should say that plainly rather than stretching a finding to cover a case it does not fit.

Reporting elsewhere attributed that rejection to the 50-week moving average. We checked it on our own data.

LevelValueTuesday's high versus it
50-week MA, completed weeks$81,8230.67% below, about $550
350-day MA$81,8690.73% below
200-day MA$69,16317.51% above
50-day MA$65,80023.51% above

That one holds up. The high came within about $550 of the 50-week average, and the 350-day average sits at effectively the same place, which is unsurprising since 350 days is 50 weeks. So Monday and Tuesday produced two highs a day apart with two different candidate structures behind them: an exact round number and a long term average.

We are not claiming either one caused a reversal. Both are levels that market participants watch and place orders around, which is a mechanism, and neither is a prediction. When a widely repeated explanation checks out on our own numbers we say so, in the same way we said on Saturday that a widely repeated one did not.

The squeeze we declined to predict did not happen

On Monday we reported that the Binance long/short account ratio read 0.9459 at the high, further below parity than the 0.9988 that preceded the August 20 squeeze. We then wrote, in as many words, that we were not saying it would repeat, because one prior instance is not a base rate.

It did not repeat. Here is what happened instead.

MeasureMondayNow
Long/short accounts0.94591.0504
Perp open interestRising 2.33% in 20h1.43% lower over 48h
Funding0.0100% for six settlements0.0077%, 0.0059%, 0.0071%

The shorts closed out quietly, open interest came down, and funding dropped off the ceiling it had been pinned to for two days. There was no cascade. Recording the non event matters as much as recording the event, because a site that only mentions its flagged setups when they fire is running a different kind of scoreboard.

The funding detail is worth keeping. Monday's article noted six consecutive settlements at exactly 0.0100%, the venue ceiling, and argued that this was structural rather than a crowding signal. Three settlements since have printed below the cap. Longs are paying less to stay long than they were at the highs.

Where things stand

AssetPrice24h24h range
Bitcoin$78,3980.86% lower$77,633 to $79,252
Ether$2,471.55+0.31%$2,414.64 to $2,475.61
XRP$1.38206.15% lower$1.3574 to $1.4783
Solana$96.721.64% lower$94.95 to $98.60
BNB$698.090.12% lower$688.00 to $708.06
Dogecoin$0.085033.81% lower$0.08385 to $0.08868
Tokenised gold$4,5871.30% lower$4,580 to $4,666

The alts are giving back more than Bitcoin. XRP is down 6.15% and Dogecoin 3.81% against Bitcoin's 0.86%, which is the mirror image of last week when they led on the way up. Total market capitalisation is $2.652 trillion, down 3.51%, with Bitcoin dominance at 59.25%.

ETF flows are the one part of this that has not turned. US spot Bitcoin ETFs took $336 million on August 25, a seventh consecutive day of net inflows, with roughly $2.08 billion across the seven. Fund assets have climbed to about $98.56 billion from $78.67 billion a week earlier, which reflects both the inflows and the price. Today's figure publishes after the US close and we will not estimate it.

Fear and Greed reads 65, Greed, down from a peak of 74 on Tuesday. The ETH to BTC ratio is 0.03150, up 1.03%, the sixth day we have reported it. Up, up, down, up, down, up. Still not a trend, and we have now said so six times.

What we are not saying

We are not saying round numbers are resistance you can trade. A 1.60% rate of exact landings is a strong statistical signature and it is not a strategy. It tells you where highs tend to be recorded, not when they will occur, and the other 98.4% of highs landed somewhere else entirely.

We are not saying $80,000 or $81,273 will hold or fail. Two tests are two tests.

We are not saying the 50-week average caused Tuesday's reversal. We are saying the number checks out at $81,823 and the high stopped $550 short of it, which is a fact about where, not why.

We are not saying the rally is over because the break failed. Bitcoin is still up about 21% from the August 18 close and ETF inflows have run for seven straight sessions.

What we are saying is that our own metric spent this week showing us what it cannot see, and today the thing it cannot see left a mark we could count. Sixteen candles out of 999 ended on a number a human chose in advance. That is the resting book, visible at last, in the one place it cannot hide.