Sharpe Ratio Choices: Excess Returns, Cash, and Fund Flows

Summary

The document raises practical questions about calculating and annualizing a strategy’s Sharpe ratio when interest rates are no longer near zero. It contrasts a simple mean-profit-and-volatility calculation with approaches that subtract a risk-free return, and asks whether the adjustment should use profit or percentage returns and whether volatility should be measured on excess returns. It also highlights cash that cannot earn the assumed bill rate, changing fund size from investor flows, and financing or margin effects.

The material is a set of questions rather than a resolved methodology: it gives no recommended formula, data convention, or worked example. It nevertheless identifies choices that can materially change reported performance, including the return series, benchmark horizon, treatment of idle cash, and handling of flows. Any implementation would need a consistent portfolio return series and benchmark aligned with the reporting interval; the document does not specify how to make those choices or settle its questions.

Key ideas

• A Sharpe ratio based on excess returns requires a clearly specified risk-free benchmark.
• The return series and volatility measure should use consistent definitions and units.
• Cash that earns less than the assumed benchmark can affect measured performance.
• Benchmark maturity and rolling conventions matter when the reporting period spans months.
• Fund inflows, outflows, margin, and funding require explicit treatment, but the document offers no final formula.

Tags

# Formal Sharpe Ratio Calculation


# Formal Sharpe Ratio Calculation












Would appreciate clarity from senior quants on the correct way to calculate sharpe

Back in the zero interest rates days, I saw some senior quants would calculate sharpe ratio as avg(pnl)/std(pnl) and then annualize depending on strategy freq

- Now that interest rates are > 5%, I'm very skeptical of this quick calc. If systems are too hardedcoded, would you just sythentically do ( avg(pnl) - (3m t-bill total pnl) )/ std(pnl)? Frankly I do not like this method, and I've seen people argue over whether it should be divided by std dev of excess returns over t bills

- The other way I saw was calculating returns (%-wise) and doing the same for 3m t-bills, then doing excess return.

- what if you are holding cash that you can't put into t-bills, (so you need to account for this drag)?

- if your reporting period is 6 months to 1 year, would you roll the t bills or just take the 6m/1y bill as the risk free rate?

- To account for increasing capacity and <3/4>, I start out with the fund's total cash, then do the daily value of the holdings + cash, take the avg of that pnl, minus the cash return from 3m to get the numerator. I take the avg of the time series above to get the denominator.

- But if the fund size changes do to inflows or outflows, how would you account for that? what about margin or funding considerations?

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.