CryptoWeeklies / Glossary
What is logarithmic regression?
Logarithmic regression fits a straight line through price plotted on a log scale. The fitted line is the fair value estimate, and the deviation from it is how far price has strayed.
Why log scale
On a normal axis, an asset that went from $1 to $100 makes its first ten years invisible. On a log axis, equal vertical distances mean equal percentage moves, so a 50% drop in 2015 looks the same size as a 50% drop today. For assets that have moved across several orders of magnitude, that is the only view that shows the whole history honestly.
A straight line on a log chart is exponential growth at a constant rate. Fitting one is a way of asking: what steady compounding rate best describes this asset's history, and where is price relative to it?
Reading the deviation
The deviation figure is the gap between today's price and the fitted line, as a percentage. Positive means above the long-run fitted trend; negative means below it.
Where it breaks
Log regression assumes growth continues at roughly the historical rate forever. Nothing does. As an asset matures its growth rate falls, and a line fitted to its explosive early years will keep reporting "undervalued" as the real trend flattens beneath it. The fit also depends heavily on the start date — moving it a year changes the answer.
It is a useful description of the past. It is a weak forecast.