The MAR ratio uses the whole history rather than a rolling three years, so it includes every bad period the strategy ever had. That makes it harsher and more honest: a manager cannot age out of a 2008 drawdown.
Interpretation is similar to calmar-ratio but the bar is lower because the window is longer. Long-running trend-following programmes typically land between 0.3 and 0.8 over multi-decade histories; anything above 1 sustained over twenty years is exceptional, and figures of 3 or 4 in marketing material almost always come from short samples or from track-record-selection.
Its weakness is shared with every drawdown-based ratio: one observation in the denominator. A strategy that has simply not yet met its worst environment reports a flattering MAR right up until it does.
Related: calmar-ratio, max-drawdown, track-record-selection, sterling-ratio