Sharpe Ratio does not take risk-free rate as a parameter

[afzalmushtaque]

Backtesting.py Line #954

backtesting.py/backtesting/backtesting.py

Line 952 in 7010d68

s.loc['Sharpe Ratio'] = mean_return / (returns.std() or np.nan)

Expected Behavior

s.loc['Sharpe Ratio'] = (mean_return - risk_free_return) / (returns.std() or np.nan)

Actual Behavior

s.loc['Sharpe Ratio'] = mean_return / (returns.std() or np.nan)

Additional info

Sharpe Ratio

[kernc]

With interest rates effectively zero or negative, and with short-term bond yields close to zero, assuming the risk-free rate at 0% was an honest, intended, and not unreasonable simplification.

Does your Sharpe change significantly if you amend it so?

[afzalmushtaque]

Yes it does. Risk free rate here in India is ~6%.

[kernc]

As a quick workaround, might this fix your Sharpe?

def adjusted_sharpe(stats, riskfree_rate=.06):
    mean_return = stats['Avg. Trade [%]'] / 100
    return stats['Sharpe Ratio'] / mean_return * (mean_return - riskfree_rate)

You can also use it to optimize: Backtest.optimize(..., maximize=adjusted_sharpe)

[afzalmushtaque]

Haha thanks for going the extra mile but that's pretty much what I am already doing. Not a deal breaker for me. However most users kind of expect a risk-free-rate parameter for a Sharpe function coming from other quant packages so I guess adding it would avoid some confusion and time spent scouring through the source code to ensure it hasn't been hardcoded to something like 2%.

[kernc]

It is since v0.3.2 possible to use backtesting.lib.compute_stats and recompute the statistics with a custom risk-free rate:

stats = Backtest(GOOG, MyStrategy).run()
new_stats = compute_stats(stats=stats, data=GOOG, risk_free_rate=.02)