Bitcoin falls, and in the evening the spot ETFs still report inflows. The reading suggests itself: the big money is buying the dip, and the floor is closer than the price looks. Read that way, you buy more the next morning.
The question is simple: did Bitcoin rise after days like that more often than after any other day?
One trading day in four is a "dip with inflows"
The setup is not rare. Of 428 trading days between 11 January 2024 and 25 September 2025, the price fell on 113 days while spot ETFs had net inflows. That is a little more than one day in four.
On 52 more days it ran the other way: price up, ETFs with outflows. Together, flows and price diverged on 165 of 428 days.
Something this common only works as a signal if what follows differs from usual.
At first glance, the story holds
Put flows and price on the same calendar day, and the reading looks good.
| Day | Next day up | all days |
|---|---|---|
| Price down, ETF inflow | 55.8 % | 50.7 % |
| Price up, ETF outflow | 38.5 % | 50.7 % |
After a "dip with inflows", five percentage points more green next days. After a "rally with outflows", twelve points fewer. Both fit the story exactly: the ETFs know something, and following them puts you ahead.
If the story ended here, it would be a usable rule of thumb.
The question that flips it: when was the number known?
A US ETF's daily flow only exists after the close in New York. Whether it had been published by Bitcoin's daily close at midnight UTC is not certain.
Measuring from that close therefore includes hours in which nobody knew the number yet. You could only act on it afterwards.
Measure instead from the next close – the first one at which the flow was certainly known – and the same table looks like this:
| Day | Next day up | all days |
|---|---|---|
| Price down, ETF inflow | 50.4 % | 52.1 % |
| Price up, ETF outflow | 48.1 % | 52.1 % |
The support is gone. After a dip with inflows, Bitcoin rose the next day in 50.4 % of cases – slightly less often than after any day. The warning sign is gone too: 48.1 % against 52.1 %.
The gap between the two readings is larger than any effect that remains. The rule of thumb lived in the clock, not in the market.
The obvious objection: institutions don't think in days
Maybe the buying works more slowly. That is what the second horizon is for: six days.
| After "price down, ETF inflow" | Share up | Median | all days |
|---|---|---|---|
| 1 day | 50.4 % | +0.03 % | 52.1 % / +0.10 % |
| 6 days | 51.3 % | +0.65 % | 56.1 % / +0.65 % |
After six days the rate is not higher but almost five points below all days, with the same median. Even with patience, the setup delivers nothing you would not have had without it.
More time does not turn a coin flip into a better one.
And Ethereum?
The same calculation for ETH and its spot ETFs, on 122 trading days between late March and late September 2025:
| After "price down, ETF inflow" (n = 34) | Share up | all days |
|---|---|---|
| 1 day | 55.9 % | 53.3 % |
| 6 days | 47.1 % | 58.2 % |
Slightly above on the next day, eleven points below after six days – both within what 34 cases produce by chance. The reverse case (price up, outflow) occurred only 15 times for ETH; that supports no verdict and is not interpreted here.
For ETH the statement rests on half a year, not two. Its direction does not contradict the Bitcoin result.
And several red days in a row?
Usually the question is asked bigger: not one red day with inflows, but several in a row. And not the next day, but the next few weeks.
That was not part of the pre-registration. The following calculation therefore only describes; it tests nothing. It covers the whole period from 11 January 2024 to 25 September 2026, 679 trading days, starts at the next close as above and looks 20 days ahead.
| From next close, 20 days | Cases | Share up | Median | Lowest point within the 20 days (median) |
|---|---|---|---|---|
| all days | 665 | 53 % | +0.7 % | −6.4 % |
| price down, ETF inflow | 156 | 48 % | −0.5 % | −6.6 % |
| of which inflow above $500m | 12 | 6 of 12 | −1.6 % | −8.9 % |
| two red inflow days in a row | 26 | 14 of 26 | +0.5 % | −6.8 % |
| three red inflow days in a row | 3 | 0 of 3 | −1.5 % | −4.9 % |
Over three weeks, with large inflows, and in a row, nothing shows up that an arbitrary day would not have delivered too.
Three red inflow days in a row happened exactly three times since January 2024: in May 2024, May 2025 and July 2025. Twenty days later Bitcoin was a little lower each time, by between 0.7 and 2.6 %. Three cases support no verdict, which is why they are listed individually.
The last column matters more than the share. After a red inflow day, the price fell in the following 20 days to a median of 6.6 % below the entry, exactly as after any day. On 7 October 2025, $876m flowed net into the spot ETFs while Bitcoin lost 2.7 %. Anyone who bought at the next evening's close of $123,306 saw a low of $102,000 on 10 October, 17 % below.
The inflow tells you who bought today. It does not tell you how far the price falls first.
How this connects to another observation
That ETF flows tend to follow price rather than lead it has shown up before: In Bitcoin ETFs, flows follow price, not the other way round. The daily calculation here fits. An inflow on a red day describes that day; it does not predict the next one.
What to do with this
- Don't buy more on "the ETFs are buying the dip". After 113 such days, the next day was as open as any other.
- If you buy the dip anyway, plan for the usual pullback. Even after red inflow days, the low of the next 20 days sat a median of more than 6 % below the entry.
- With any "flows vs. price" chart, check the clock first. If it measures from the same day's close, part of the punchline sits in hours when the number was not yet known.
- The flows are still worth reading – as a state: who is buying, who is selling, how much. That is what the ETF flow tracker shows, day by day and by issuer.
How it was calculated
The rules were fixed before the first run: which days count, which close the return starts from, which two cells are tested, and which threshold a finding must clear. The period tested had not been examined by anyone before. Eight comparisons (two setups × two horizons × two assets), hence a strict threshold of p < 0.00625 per comparison. The smallest value measured was 0.19.
A second period has been running forward since 26 September 2026 under the same rules. Its first evaluation is due once 125 trading days are in, around the end of March 2027.
The section "And several red days in a row?" is not part of the study. It is a separate calculation from 28 September 2026 over the whole period, using Binance BTCUSDT daily candles and without a threshold fixed in advance. There too, every cell lies within what the same number of randomly drawn red days produces.
FAQ
Which flow data was used? The daily net flows of all US spot Bitcoin or Ethereum ETFs combined, per trading day. Days with a net flow of exactly zero do not count (13 for BTC, 3 for ETH).
Why does the next day's close count as the start? Because the daily flow only exists after the close in New York. The first Bitcoin daily close at which it was certainly known is the next day's. The calculation from the same day's close is shown for comparison in the first table.
Why only 122 days for Ethereum? The ETH flow series used here starts on 28 March 2025. The ETFs themselves have traded since July 2024; the months before are not in this calculation.
Does this mean ETF flows are meaningless? No. They describe demand and positioning. Only one question was tested: whether a red day with inflows makes the next day, or the next six days, better. It does not.
Can this change? Yes, which is why the second period keeps running forward. If it shows a deviation in the same direction, it will be added here.
Not investment advice, not a recommendation, not a forecast — historical patterns are no promise.
Study the Past — Improve your Future 🥋