±1.83 % 8 of 10 UTC days stay inside this BTC band · Oct-23 00:00Z → Oct-24 00:00Z, from the 00:00Z price · not a forecast of direction, snapshot taken Oct-23 2026, measured and published by askthetape.com.
median move ±0.92 % — half the days move less, half more
🔔 Alert me: the BTC band moves over 50 % in a day 🔔 Alert me: a BTC day closes past its 1-in-10 linefree in Telegram, at the UTC daily close · latest: the band moved 4.2 % for Oct-23, Oct-22 closed at ×1.17 of its line
BTC and ETH, one UTC day at a time (00:00Z → 00:00Z, from the 00:00Z price): a size either way, never a direction.
Engine recount 2026-10-23T21:19Z · 0 h old · as of Oct-23 00:00Z (last closed hour 23:00–00:00Z).
The chart: BTC and ETH at 00:00Z each day from our own hourly candles. The shaded channel is each past day’s 8-of-10 band around the path — a dot inside it is a day that stayed within, a dot outside is a day that left it; the cone ahead is the band for Oct-23 00:00Z → Oct-24 00:00Z from the 00:00Z price, the same width up and down. The marker is the latest closed hour in our data.
Last 365 days (Oct-23 2025 → Oct-22 2026, 365 graded UTC days): BTC — the 8-of-10 band held on 81 % of days (target 80 %); the 1-in-10 band was crossed on 10 % (target 10 %). ETH — the 8-of-10 band held on 82 % of days (target 80 %); the 1-in-10 band was crossed on 8 % (target 10 %).
Oct-23 00:00Z → Oct-24 00:00Z
High-low range: median 2.28 % (the 8-of-10 range line is printed once 30 days graded since Sep-26 2026 show at least 78 % inside; so far 27 days).
| chance of a move larger than | 2 % | 3 % | 5 % |
|---|---|---|---|
| this window | 18 % | 7 % | 1 % |
| all days of the past year | 28 % | 13 % | 4 % |
Options priced a one-standard-deviation day of ±1.99 % (Deribit DVOL 38.0 at the close of Oct-22, divided by √365). How that usually compares with what is delivered: below.
Made at Oct-23 00:00Z from the last closed hour (23:00–00:00Z). Graded at Oct-24 00:10Z, 10 minutes after the window closes.
Oct-23 00:00Z → Oct-24 00:00Z
High-low range: median 3.17 % (the 8-of-10 range line is printed once 30 days graded since Sep-26 2026 show at least 78 % inside; so far 27 days).
| chance of a move larger than | 2 % | 3 % | 5 % |
|---|---|---|---|
| this window | 30 % | 17 % | 5 % |
| all days of the past year | 40 % | 25 % | 10 % |
Options priced a one-standard-deviation day of ±2.73 % (Deribit DVOL 52.2 at the close of Oct-22, divided by √365). How that usually compares with what is delivered: below.
Made at Oct-23 00:00Z from the last closed hour (23:00–00:00Z). Graded at Oct-24 00:10Z, 10 minutes after the window closes.
Measured: how far BTC and ETH move over one UTC day from the 00:00Z price, either way, and how wide the day’s high-low range may get; past days since 2022 are replayed with the same rule and graded against what happened — the 8-of-10 band held on 81 % of 1,755 days for BTC and 81 % for ETH.
Not measured: which way the price goes, and the part of the UTC day already traded when you read this.
Not a price forecast: the band is symmetric on purpose, and the live record of this rule starts on Sep-26 2026; the years before it are a replay.
How far: from the past year’s weeks as they were, not from a model — median ±3.19 %, 8 of 10 weeks within ±6.38 %, 1 in 10 beyond ±9.53 %. On the tail of the week the model is not better than climate, so this band is the past 365 closed weeks.
How loud: the week’s realized volatility (every hourly move, added up) — the model says 4.66 %, the median week of the past year was 5.43 % (R² 0.28 against the past year’s climate).
The past-year band held on 82 % of weeks since Jan-2022; weeks overlap day by day, so this is one pooled number, not a yearly table.
How far: from the past year’s weeks as they were, not from a model — median ±3.86 %, 8 of 10 weeks within ±9.34 %, 1 in 10 beyond ±12.50 %. On the tail of the week the model is not better than climate, so this band is the past 365 closed weeks.
How loud: the week’s realized volatility (every hourly move, added up) — the model says 6.42 %, the median week of the past year was 7.42 % (R² 0.32 against the past year’s climate).
The past-year band held on 82 % of weeks since Jan-2022; weeks overlap day by day, so this is one pooled number, not a yearly table.
Options quote an expected size too. They usually price more than is delivered — but part of that gap would be there even with no premium at all, so the fair comparison is printed next to each number.
BTC, Jan-01 2022 → Oct-22 2026 (1,755 days): the delivered volatility came in below the priced one on 77 % of days — but a fair forecast with zero premium would also come in above the delivered on about 69 %, because a few loud days pull the average up. Median delivered ÷ priced: 0.73, against 0.82 for that fair forecast. The move stayed inside ±1 priced standard deviation on 80 % of days; a perfectly sized one would hold about 77 % (fat tails), not the 68 % of a bell curve.
ETH, Jan-01 2022 → Oct-22 2026 (1,755 days): the delivered volatility came in below the priced one on 74 % of days — but a fair forecast with zero premium would also come in above the delivered on about 70 %, because a few loud days pull the average up. Median delivered ÷ priced: 0.78, against 0.81 for that fair forecast. The move stayed inside ±1 priced standard deviation on 78 % of days; a perfectly sized one would hold about 78 % (fat tails), not the 68 % of a bell curve.
Replayed: each day since Jan-01 2022 re-made by the rule fixed on Sep-25 2026, seeing only the windows closed by then; these bands were not published at the time, the forward line below is the live record. The 8-of-10 band should hold on about 80 % of days; the 1-in-10 line should be crossed on about 10 %.
| year | days | inside the 8-of-10 band [95 % range] | beyond 1-in-10 | high-low inside its 8-of-10 |
|---|---|---|---|---|
| 2022 | 365 | 82 % [79–85] | 10 % | 75 % |
| 2023 | 364 | 80 % [76–83] | 10 % | 78 % |
| 2024 | 366 | 80 % [76–83] | 10 % | 81 % |
| 2025 | 365 | 82 % [78–85] | 10 % | 79 % |
| 2026 (so far) | 295 | 80 % [76–84] | 10 % | 82 % |
| 2022 → Oct-22 2026 | 1,755 | 81 % [79–82] | 10 % | 79 % |
Last 90 graded days (Jul-25 → Oct-22): 81 % inside the 8-of-10 band. Over 90 days a well-sized band lands anywhere in 73–87 % by chance alone.
Forward record since Sep-26 2026 (windows made after the rule was frozen on Sep-25 2026): 27 days graded — 85 % inside the 8-of-10 band, 7 % beyond the 1-in-10 line, high-low inside its 8-of-10 on 85 % (too few days to judge yet).
Not equally good in every stretch: on the quietest third of days (by the band’s own width against the past year) it held 77 %, on the middle third 81 %, on the loudest third 83 %. How much of the day-to-day change in volatility the model catches beyond what the past year alone would say: R² 0.29 against the past year’s climate.
| year | days | inside the 8-of-10 band [95 % range] | beyond 1-in-10 | high-low inside its 8-of-10 |
|---|---|---|---|---|
| 2022 | 365 | 81 % [78–84] | 12 % | 76 % |
| 2023 | 364 | 82 % [78–86] | 9 % | 77 % |
| 2024 | 366 | 79 % [77–81] | 11 % | 78 % |
| 2025 | 365 | 79 % [77–82] | 10 % | 80 % |
| 2026 (so far) | 295 | 82 % [78–86] | 9 % | 81 % |
| 2022 → Oct-22 2026 | 1,755 | 81 % [79–82] | 10 % | 78 % |
Last 90 graded days (Jul-25 → Oct-22): 83 % inside the 8-of-10 band. Over 90 days a well-sized band lands anywhere in 73–87 % by chance alone.
Forward record since Sep-26 2026 (windows made after the rule was frozen on Sep-25 2026): 27 days graded — 85 % inside the 8-of-10 band, 11 % beyond the 1-in-10 line, high-low inside its 8-of-10 on 78 % (too few days to judge yet).
Not equally good in every stretch: on the quietest third of days (by the band’s own width against the past year) it held 78 %, on the middle third 82 %, on the loudest third 83 %. How much of the day-to-day change in volatility the model catches beyond what the past year alone would say: R² 0.34 against the past year’s climate.
Every forecast with its grade, both coins, one row per day (3,514 rows; our bands, whether each day stayed inside them, and the move and range it had, no price candles): A row without a grade is either still open or a window the exchange left with missing hourly bars (Mar-24 2023); those are left out of the record.
After Bollinger bands were at their narrowest tenth of the past year, the next 7 days’ range was wider than the past year’s median week on 27.3 % of 451 such days for BTC (35 separate quiet stretches) and 33.1 % of 411 for ETH (38) — against 44.5 % and 44.4 % for random days drawn in the same clusters. A quiet stretch was followed by another fairly quiet stretch more often, not by a storm.
Only for BTC, and only a hint: after a day in its top 5 % with no other such day in the week before, the next day landed in the top tenth 16.7 % of the time against 9.4 % for any day (72 cases; counting days inside a cascade would inflate it). For ETH, no: 8.5 % against 10.6 % (71 cases).
Yes: in 2022–2026 a Saturday or Sunday carried 0.61 of a weekday’s volatility for BTC and 0.69 for ETH.
Not something to count on. In 2022–2026 release days (FOMC, CPI, the jobs report) looked louder for BTC (×1.18), but part of that is 2022: without it, ×1.11 (1.01–1.20) for BTC and ×1.05 (0.97–1.15) for ETH — the ETH range includes no difference at all.
These four are one study (Sep-25 2026, our own code; placebo days drawn in the same clusters as the real episodes), not recomputed daily.
How it is made: the volatility of the last 1, 7 and 30 days (from hourly candles) predicts the volatility of the UTC day that follows (00:00Z → 00:00Z) on a log scale (HAR); the band is that volatility times the spread of past outcomes. Refit every day at 00:00Z: the coefficients on all closed 24-hour windows since Oct-2017, the band on the last 365 closed windows. The track record is this same rule run day by day since Jan-01 2022, each day seeing only windows closed by then. A window is graded 10 minutes after it closes; until then it has no grade. The band is symmetric on purpose: it measures size and says nothing about which way. Checks at every recount: the model must still catch volatility clustering, must find a planted signal and must find nothing in shuffled inputs — all passed on this recount. Options volatility (DVOL) is shown as what was priced, not used as an input: it added too little to pass the pre-registered test. Pre-registration 4373c8e4 · engine recount 2026-10-23T21:19Z.
Where the numbers come from: our own band on Binance hourly candles — Binance spot BTCUSDT and ETHUSDT public klines, pulled by our own box since Aug-2017 and checked against Coinbase hourly and Bitstamp daily closes; the bands, grades and record are ours. DVOL is Deribit’s public index, captured hourly by our own poller.
Free: the 00:00Z band for both coins, the full record and the file of every forecast. On the Tape plan the same band is re-made every hour from the last closed hour (as of 21:00Z in this build).
export: whole history · on the Desk plan also as one file, /export/next-move-size/record_daily.csv.gz: the same record as the CSV above (our bands, grades, each day's move and range), one row per coin per graded UTC day since Jan-01 2022, origin stamped inside; the hourly candles under the bands are the exchange's and are in no file
the daily facts page · A size, never a direction.