Three ratios — and how much they disagree
Return alone ranks assets by how much they went up, which says nothing about what holding them felt like. Risk-adjusted ratios divide return by some measure of risk, and the choice of measure changes the answer. This page computes three of them for the largest crypto assets and, more importantly, shows how far apart their rankings are.
Sharpe divides by total standard deviation. It is the most quoted and the least suited to crypto: it assumes a normal distribution that fat-tailed daily returns violate, and it treats a violent rally as risk. Sortino divides only by downside deviation, which removes the second objection. Calmar divides annual return by the worst drawdown actually endured, which avoids distributional assumptions entirely but rests on a single historical episode.
None of the three is right in general, so the rank-spread column is the point of the page. A coin near the top under one ratio and far down under another is not a discovery about the coin; it is a demonstration that any single ranking you were shown elsewhere was a choice of definition presented as a measurement.
⚠️ Not investment advice. Every ratio here is computed from past returns over a chosen window, and past return in particular does not persist. The risk-free rate is taken as zero. All decisions and risks are your own.
It divides return by total standard deviation. Two things make it a poor fit here: it assumes a normal distribution that fat-tailed crypto returns violate, so the worst days are understated, and it treats large gains as risk in exactly the same way as large losses.
It divides only by downside deviation, so upside volatility is not penalised. The denominator still uses the total number of observations rather than only the losing ones, which prevents assets with rare losses from being unfairly punished.
Annual return divided by the worst drawdown actually endured. It makes no assumption about the shape of the return distribution, which is its advantage, but it rests on a single historical episode — so a coin that happened to avoid a crash scores well without being safer.
Because they encode different definitions of risk. A coin whose losses are rare but severe looks better under Sortino than under Calmar; one with steady mild volatility looks better under Sharpe. The rank-spread column shows how many places each coin moves, and a large spread means the ranking depends on which definition you accept.
Because against crypto return magnitudes it is within rounding error, and fixing it to a specific figure would distort comparisons between periods when rates differed. Setting it to zero keeps every column on the same basis.
No. Volatility persists to a useful degree, so the denominators carry some information about the future. The numerator is past return, which does not. Treating a high historical ratio as a forecast is the main way these numbers get misused.