M-Squared Risk-Adjusted Performance

Told that Fund A has a Sharpe ratio of 0.9 and Fund B has 0.7, most people can't say how much money that difference is actually worth — a Sharpe ratio has no intuitive units. M-squared (M²) fixes that by translating the ratio back into something everyone already understands: a percentage return, expressed on a like-for-like risk basis, so a manager can say "this fund would have returned X% if it had carried exactly the benchmark's volatility."

Levering to match the benchmark's risk

$$M^2 = r_f + \frac{\sigma_{\text{benchmark}}}{\sigma_p}\big(R_p - r_f\big) - R_{\text{benchmark}} .$$

In words: take the portfolio's actual excess return over the risk-free rate, $R_p - r_f$, and scale it by the ratio of the benchmark's volatility to the portfolio's own volatility, $\sigma_{\text{benchmark}}/\sigma_p$. That scaling is exactly what levering up (if the portfolio was less volatile than the benchmark) or delevering with cash (if it was more volatile) would do to the portfolio's actual historical return — mixing in risk-free borrowing or lending doesn't change the Sharpe ratio, only the volatility, so this is a fair way to make two differently-risky portfolios directly comparable. Add back the risk-free rate to get a normal-looking return figure, then subtract the benchmark's actual return: the leftover is M², in ordinary percentage points, positive if the portfolio would have beaten the benchmark at matched risk.

Worked example 1 — a lower-vol fund, levered up to compare

A fund returned 11% with volatility of 8%; the benchmark returned 9% with volatility of 12%; the risk-free rate is 3%. Scaling factor: $12/8 = 1.5$. Levered excess return: $1.5 \times (11 - 3) = 1.5 \times 8 = 12\%$. Adding back the risk-free rate: $12 + 3 = 15\%$. M² is $15\% - 9\% = 6\%$ — at matched risk, this fund would have beaten the benchmark by 6 percentage points, a number a client can immediately understand, unlike a bare Sharpe ratio.

Worked example 2 — a higher-vol fund that looks good raw but isn't

A second fund returned 14% with volatility of 20%, against the same 9%-return, 12%-volatility benchmark and 3% risk-free rate. Its raw return of 14% beats the benchmark's 9% by 5 points — but its volatility is far higher. Scaling factor: $12/20 = 0.6$. Delevered excess return: $0.6 \times (14 - 3) = 0.6 \times 11 = 6.6\%$. Adding the risk-free rate: $6.6 + 3 = 9.6\%$. M² is $9.6\% - 9\% = 0.6\%$ — at the benchmark's actual risk level, this fund barely beats it at all, despite its much larger headline return; nearly all of its outperformance came from taking more risk, not from skill.

What this means in practice

M² is popular in client-facing performance reports precisely because it avoids the "what does 0.9 mean" problem — a positive M² of, say, 2% is immediately legible as "two percentage points of outperformance at equal risk," which a Sharpe ratio alone never states directly. It ranks portfolios identically to the Sharpe ratio (it's a monotonic transformation of it), so nothing new is learned quantitatively — the value is entirely in making the number speak the same language as an ordinary return.