The round number, to the cent

Round numbers are where resting orders cluster, and this one behaved accordingly. Every figure below is a closed five minute Binance spot candle.

Candle, UTCHighCloseValueBuy share
15:20$79,920$79,902$19.9 million56.2%
15:25$79,989$79,896$30.1 million51.2%
15:30$80,000.00$79,814$30.9 million60.7%
15:35$79,910$79,550$48.9 million19.9%

Look at the last two rows together. Buyers were winning the candle that reached $80,000, at a 60.7% share. Then the flow inverted almost completely, and it inverted on 58% more volume. That is not buyers running out of money. That is a different set of participants arriving with size at a price they had decided on in advance.

A 19.9% reading is the most sell skewed five minute figure we have published in this series. For context, our recent examples have been a 71.7% buy share during Thursday's squeeze and a 40.4% reading during Friday's reversal at $79,500. Today went considerably further.

The round number held on its first test. That is a description of what happened between 15:30 and 15:35 UTC and nothing more. We are not treating it as a level, for reasons this site set out on August 3 and has not revisited since.

Six sessions, 26%, and a five point band

Now zoom out from five minutes to whole sessions, because the contrast is the point of this article.

SessionChangeRangeDaily buy share
Tue Aug 18+0.30%1.61%48.9%
Wed Aug 19+7.12%9.09%53.9%
Thu Aug 20+5.32%6.53%50.5%
Fri Aug 21+7.27%8.86%51.9%
Sat Aug 221.61% lower3.04%50.8%
Sun Aug 23+0.86%3.32%53.4%
Mon Aug 24, partial+2.29%4.34%49.9%

The move from the August 16 close of $62,900 to the current $79,512 is 26.4%. Across every session of it, the daily buy share stayed between 48.9% and 53.9%. A day that gained 7.27% and a day that gained 0.30% are separated by three percentage points of buy share.

The obvious reading is that the metric failed to register the largest move of the year. That reading is wrong, and correcting it is the reason for the next section.

A correction we owe you

Over the past week this site published three articles about things our buy share cannot see: ETF creations that settle off the public book, futures liquidations fifteen times larger than the spot flow, and market makers repricing bids without ever trading. Each was accurate. Taken together they could easily leave a reader thinking the number carries no information at all.

So we measured that directly, across 399 completed daily sessions.

MeasureValueWhat it means
Correlation with daily return0.505Explains roughly a quarter of the variance
Full historical range38.0% to 57.5%The metric is bounded and never dramatic
Median session48.9%The typical day is slightly sell skewed
95th percentile53.7%53.9% on August 19 was near the top of the range
Median on gains above 5%52.6%Eight such sessions in 399

A correlation of 0.505 is not nothing. It is a genuine relationship, and it means the buy share does tell you something real about direction at the daily scale. What it does not do is scale with the size of the move. The entire historical range across those 399 sessions spans about twenty points, so there is nowhere for a 7% day to go that a 1% day cannot also reach.

Seen that way, this run behaved exactly as the record says it should. August 19 read 53.9%, just above the 95th percentile. The three biggest up days averaged near the 52.6% median for days of that size. Nothing malfunctioned. The error would be expecting a bounded ratio to shout.

We should have published this measurement alongside the first limitation article rather than after the third. Describing what a tool misses without also quantifying what it catches leaves a misleading impression, and that is on us.

Scale is the whole answer

Put today's two measurements side by side and the resolution is obvious.

Over the full session so far, Bitcoin is up 2.29% and the buy share reads 49.9%, which tells you almost nothing about the drama inside it. Over five minutes, the same metric read 60.7% then 19.9% across the $80,000 rejection, a forty point swing that describes the event precisely.

The averaging is what kills it. A session contains hours of balanced two way flow, and the few minutes that decide the direction get diluted into them. That is why the daily figure sat at 50% while the price went from $77,734 to $80,000 and back.

So the practical guidance we would give, and that we have been circling for a week without stating plainly: read this metric at the resolution of the event you care about. For a specific move, use the five minute or hourly reading around it. For a session, expect a number within a few points of half, and treat a 54% day as genuinely strong rather than looking for 70%.

Positioning at the high, reported without projection

Three readings, and then an explicit refusal.

MeasureReadingNote
Perp open interest, 20h2.33% higherPositions opening into the level
Long/short accounts, 16:000.9459More accounts short than long at the high
Funding, last six settlements0.0100% eachThe venue ceiling, not a crowding signal

The middle row will look familiar to anyone who read us on Thursday. Before the August 20 squeeze the same ratio read 0.9988, and roughly $2.7 billion of shorts were liquidated over the following day. It now reads 0.9459, which is further below parity than that.

We are not saying that repeats. One prior instance is not a base rate, and on Friday we corrected ourselves for exactly this kind of aside, when we implied Thursday's squeeze had exhausted itself and Bitcoin added another 7.4%. The reading is a fact about how accounts are currently positioned. What happens next is not in it.

One more thing worth noting from Saturday. Several outlets framed the weekend as a thin liquidity risk ahead of this $80,000 test. We measured that claim across 720 hours and found weekend depth running at 87% to 109% of weekday depth depending on the asset. The premise was repeated widely and it did not hold up when checked.

Where things stand

AssetPrice24h24h high
Bitcoin$79,512+2.71%$80,000
Ether$2,496.29+2.02%$2,532.95
XRP$1.5141+0.66%$1.5466
Solana$96.99+1.49%$97.61
BNB$709.20+1.69%$716.84
Dogecoin$0.091880.65% lower$0.09453
Tokenised gold$4,655+1.14%$4,673

The price sits about 15.4% above the 200-day moving average of $69,043, a level Bitcoin had not closed above since November 2, 2025 until five sessions ago. Total market capitalisation is $2.696 trillion, up 0.56%, with Bitcoin dominance at 59.15%. Note how modest the alts are today after Saturday's violence: XRP up 0.66% and Dogecoin actually lower, against Bitcoin's 2.71%. The rotation that produced Saturday's 23% XRP range has not resumed.

US spot Bitcoin ETFs drew a net $1.92 billion across the week, the largest weekly total in about ten months. Today's figure publishes after the US close and we will not estimate it.

Fear and Greed reads 73, Greed, its highest of this run. It read 31 on August 17. That is a 42 point move in seven sessions, which says as much about how quickly sentiment surveys follow price as about sentiment.

The ETH to BTC ratio is 0.03136, down 0.85%. Five days of reporting it now: up, up, down, up, down. We declined to call it a trend at two observations and we decline again at five.

What we are not saying

We are not saying $80,000 is resistance. It was touched once and rejected once, in the space of ten minutes, and one instance is an event rather than a level.

We are not saying the accounts positioned short at the high will be squeezed. That is the aside we got wrong on Thursday and corrected on Friday, and we are not repeating the shape of it with a different number.

We are not saying the buy share is a trading signal. A correlation of 0.505 with daily returns is a real relationship and it is nowhere near strong enough to act on alone.

We are not saying the rally continues or stops.

What we are saying is that a metric which reads 49.9% across a session and 19.9% across five minutes inside it is not contradicting itself. It is being averaged. Today gave the clearest illustration of that we have had, and it is the piece we should have published before the three articles about what the number cannot see.