How to Compare Sharpe Ratio, Returns, and Drawdown
Summary
The discussion considers how to judge algorithm performance when strategies have different returns and Sharpe ratios across instruments. It argues that Sharpe alone is incomplete: it relates excess return to total return variability, but a high value can be practically irrelevant when absolute returns are negligible or leverage is constrained. Sortino and Calmar ratios are offered as alternatives that focus respectively on downside variability and drawdown relative to return.
The answers also emphasize actual returns, position sizing, and the possibility that strategies fit different instruments or their characteristics at trade entry. Maximum drawdown matters because severe losses may be unacceptable, while examining a return distribution can add context beyond a single statistic. The document provides qualitative guidance, not a universal ranking or empirical comparison; the appropriate metric depends on risk limits, leverage, and the strategy’s intended use.
Key ideas
- Sharpe ratio measures excess return relative to total variability, but does not capture every practical constraint.
- Sortino and Calmar ratios emphasize downside variability and drawdown, respectively.
- Absolute returns and feasible leverage matter even when a strategy has a high risk-adjusted ratio.
- Interpret performance in light of position sizing, instrument characteristics, and the distribution of returns.
Tags
Full text
# Sharpe Ratio vs Net Profit vs max drawdown # Sharpe Ratio vs Net Profit vs max drawdown When evaluating the performance of an algorithm, what should hold more importance? Sharpe Ratio , Net profit or max drawdown? For instance, I have two algorithms one performs very good on Stocks with high rate of return like AAPL, FB but the second one that predominantly gives low profit but slightly higher sharpe ratio in general seems to outperform the first one on stocks like INTC , AAMC , IBM. What should I infer from this? ## Answer by NBF (score 1) https://quant.stackexchange.com/a/16969 Sharpe ratio alone is not, IMHO, a great measure. It measures the slope of a line to the Riskfree rate in a mean-stdev plot. If leverage were free and unlimited, then Sharpes would count more. Many traders look at Sortino ratios (looking specifically at downside stdev), or Calmar ratios (excess return/max drawdown), as more reasonable measures of an algorithm's performance. Since leverage is limited, none of these ratios replaces looking at actual returns. If the Sharpe is 3 but the returns are 0.01bp per day, it is likely to remain a completely theoretical excercise. ## Answer by dkhokhlov (score 0) https://quant.stackexchange.com/a/16988 RE: Sharpe Ratio , Net profit or max drawdown? Definitely not Net profit. It is primitive and suspected to luck or spurious results. To make a choice from the rest of two we need additional information about the risk management - position sizing. Assuming we do not have that info - i would choose Max drawdown, loosing all money is not an option. ## Answer by Nathan S. (score 0) https://quant.stackexchange.com/a/17000 How does it perform on stocks according to the characteristics you would know at trade entry? For example, you're not really gaining insight if you recognize a momentum strategy worked well on stocks that had momentum characteristics during the trade. But you might mean high beta or high volatility stocks. Without more detail I will say that some strategies are suited to particular instruments and it helps to recognize that. In some sense you need ALL of those measures, but you can go about measuring any way you like. I think something like Sortino works well if you had to pick just one. Just keep in mind that before it was the Sortino it was just a metric some guys used to supplement their intuition about strategy perdormance. I personally like to look at a histogram of daily returns more than anything else.
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.