Key findings
- Across 8,591–8,593 matched OKX spot hours (October 2025–September 2026), median absolute hourly returns were 0.2004% for BTC, 0.2588% for ETH, 0.3129% for SOL and 0.3164% for DOGE. Evidence →
- Sample hourly betas against BTC were 1.2168 for ETH, 1.2706 for SOL and 1.2473 for DOGE. Evidence →
- Empirical P99 hourly moves (1.7751% for BTC to 2.8403% for DOGE) describe historical tail frequency, not a guaranteed loss boundary or causal trigger. Evidence →
| Asset | Closed bars / expected | Valid returns N | Median |r| (%) | P99 |r| (%) | BTC beta | Matched returns N |
|---|---|---|---|---|---|---|
| BTC | 8596 / 8598 | 8593 | 0.2004 | 1.7751 | 1.0000 | 8593 |
| ETH | 8596 / 8598 | 8593 | 0.2588 | 2.4855 | 1.2168 | 8593 |
| SOL | 8595 / 8598 | 8591 | 0.3129 | 2.6685 | 1.2706 | 8591 |
| DOGE | 8595 / 8598 | 8591 | 0.3164 | 2.8403 | 1.2473 | 8591 |
What the sample shows
The median absolute hourly return is 0.2004% for BTC and 0.3164% for DOGE. The corresponding P99 values are 1.7751% and 2.8403%. A fixed percentage move therefore occupies a different place in each asset’s empirical distribution. That supports reporting an asset-specific baseline alongside an alert; it does not establish an optimal alert setting.
The gap between central and tail observations is large. P99 is an empirical order statistic, not a guaranteed future loss boundary. These figures alone are not a formal test against a normal distribution, and no false-positive rate is estimated because there is no independently labelled set of abnormal events.
Sample construction and missing hours
We recalculated public-market candle data held in an existing research archive. The common observation window is 1 October 2025 00:00 UTC to 24 September 2026 06:00 UTC, with 8,598 possible hourly candles per asset. Only confirmed, closed candles enter the sample. Timestamps, duplicate keys, finite positive prices and OHLC consistency were checked. The local archive ends before the initially considered October cutoff, so the published period stops at the shared available endpoint.
Returns are 100 × (current close / previous close − 1), and only exactly adjacent hourly closes are paired. Gaps are neither forward-filled nor treated as zero returns. Quantiles use the sorted absolute returns at index floor((N − 1) × p), without interpolation. The table gives the remaining return count as well as candle coverage.
How to read Bitcoin beta
Beta is covariance(asset return, BTC return) divided by variance(BTC return), using signed returns at matching timestamps on the same venue and in the same quote currency. ETH, SOL and DOGE have sample betas above one here. This is a historical linear association; it does not mean every BTC move is multiplied by that number, or that the remaining return is caused by a coin-specific news event.
A beta calculated from a broad historical window can differ from a recent rolling estimate. Market composition, volatility and the sample period change. Hourly estimates also cannot be transferred directly to a minute-level move. Mixing USD and USDT series would add a quote-currency effect to the comparison.
What this changes when reading a price alert
A useful alert separates the measured move, its historical context and the explanation being proposed. Robust summaries can reduce the influence of isolated outliers, but they do not identify a cause.
What is altcoin beta to Bitcoin?
Bitcoin beta describes how an asset’s returns co-move with BTC in a specified historical sample. It differs from correlation and from the asset’s own volatility. It is not a fixed multiplier for every future BTC move.
Do altcoins have the same hourly volatility as Bitcoin?
Not in this four-asset sample: median and tail absolute hourly returns differ across BTC, ETH, SOL and DOGE. The comparison covers selected surviving assets on OKX spot, not every altcoin or trading venue.
Limitations
Four selected surviving assets on one spot venue are not a representative sample of all crypto assets. Missing hours remain excluded. There is no regime-stability test, causal identification, profitability backtest or out-of-sample validation in this article.