What the risk score is
One number between 0 and 1 answering: how stretched is Bitcoin right now compared to every moment of its own past? We take the distance between price and its 374-day moving average (in log terms) and rank today's reading against every previous day since 2011. A score of 0.95 means price has been this far above its yearly trend only 5% of the time in history.
The methodology, in full
risk = percentile rank of ln(price / SMA374) among all prior days. That is the whole model. No curve fitting, no secret weights, no adjustments after the fact — each day is ranked only against days before it, so the history you see never repaints. Similar "risk level" charts elsewhere are typically proprietary; ours is reproducible from public data in a spreadsheet.
How to use it
The score is a position-sizing input, not a buy/sell signal. Low readings (below 0.2) have historically clustered in late bear markets when nobody wanted coins; high readings (above 0.8) mark euphoric stretches — which can run for months before mattering. Pair it with the drawdown chart for a "how bad / how hot" picture, and with position sizing for what to do about it.
FAQ
How is the risk score calculated?
We take the log of price divided by its 374-day moving average, then rank today's value against every previous day since 2011. The rank, from 0 to 1, is the risk score. The formula is fully disclosed — no proprietary weights.
What do the risk zones mean?
0–0.2 deep value (price far below its yearly trend, historically rare), 0.2–0.4 accumulation, 0.4–0.6 neutral, 0.6–0.8 elevated risk, 0.8–1.0 euphoria (price stretched further above trend than almost ever before).
Is a high risk score a sell signal?
No. It measures how stretched price is relative to its own history, not direction. Markets can stay in the euphoria band for months. It is a position-sizing input, not a timing signal.
Why the 374-day average?
Roughly one year of daily closes — long enough to smooth a full market season, short enough to adapt across cycles. Halving the window or doubling it changes levels slightly but not the shape.
Does the score repaint?
No. Each day's score is ranked only against days that came before it, so history never changes when new data arrives.
Related: Cycle comparison · Power law model · Market structure