Bitcoin to Altcoin Correlation

How each column is calculated

Beta

Slope of the asset's daily log returns regressed on BTC's daily log returns over the window. 1.40 means it moved 1.4% for every 1% BTC move, on average.

Correlation / R squared

Pearson correlation of the same two return series, and its square. R squared is the share of the asset's daily movement that tracked with BTC. Below about 0.3, beta is a weak description of the asset.

Volatility

Standard deviation of daily log returns, annualised by √365.

Upside / downside beta

The same regression run only on days BTC closed up, and only on days BTC closed down. A downside beta well above the upside beta means the asset has historically fallen harder than it has risen.

Beta z-score

Today's beta measured against that asset's own previous 90 readings. Past ±2 it is flagged as unusual for that asset, not merely high.

vs BTC / Alpha

vs BTC is the asset's window return minus BTC's over the same days — the plain answer to "did it beat Bitcoin". Alpha is the regression intercept scaled to the window: the return left over once BTC's influence is removed, over the same days the beta was measured on. (The annualised figure is stored and exported, but on a 30-day sample it reads as hundreds of percent and says nothing a trader can use.) A high beta with negative alpha means the asset reacted hard to BTC and still lost ground to it.

Robust beta

The same regression after clamping both return series at ±3 robust sigma (median and MAD, so the clamp is not set by the outliers themselves). A wide gap to the plain beta means a few extreme days are carrying the figure.

Top-3 share

The three largest daily moves in the window, summed, divided by the window's whole move. 50% means half the return came from three days; above 100% means those days delivered more than the entire window and the rest went the other way. Flagged as "few candles" when it is above 80% and those days alone moved the asset by 20% or more.

Burst

The largest return over any seven consecutive days inside the window, and what share of the whole window's move it accounts for. Above 70% the window is the burst. Alongside it: how far the asset now sits below its highest close in the window and how many days ago that high was. A burst with the high in the last three days is still in progress; one holding within 10% of an older high is paused ("holding"); one 10% or more below it is fading. Top-3 share catches a three-candle spike; this catches the more common one-to-two-week vertical run.

Persistent leaders

A screen, not a signal. An asset qualifies when it beat BTC over both the 30-day and 90-day windows, its 30-day alpha is positive, R squared is at least 0.3 (so the relationship the numbers rest on is real), and neither the few-candles nor the burst flag is set. The streak counts consecutive daily readings on which the 30-day return-versus-BTC was above zero; "of last 90" is how many of the newest 90 readings were. Long streaks describe what has kept going; they do not say it will continue.

Up / down capture

Cumulative asset move on BTC-up days divided by BTC's cumulative move on those days, and the same for BTC-down days. 1.00 matched Bitcoin, 2.00 doubled it, 0.50 halved it. Unlike up/down beta this is built from what actually compounded, not a slope.

Data

Settled daily candles from the OKX public API, USDT pairs. Only days where both the asset and BTC settled are used. Results are stored in Supabase, so the table you see is a saved snapshot rather than a live tick.