After the many examples of the dangers of the Sharpe Ratio, this question gets us back to reality: The Sharpe Ratio might be the least bad option for basic performance evaluation and communication to relatively non-technical parties. Communication to non-technical parties may be the bulk of investment performance communication.

In the real world, the options are rarely the Sharpe Ratio on the one hand and sound and the comprehensive statistical and qualitative analysis this board advocates on the other. Usually, the options are the Sharpe Ratio on the one hand and the (even more dangerous) analysis of historical returns, unadjusted for any measure of variability on the other. If most users of the Sharpe Ratio did not use it, they would probably not use a battery of tests instead (max drawdown, tests for normality, estimates of tail risks). They may simply pick a strategy with the highest mean return.

There may also be historical and philosophical reasons why imports from physics and signal processing were used when quantitative finance was emerging in the ‘60s. Conventions, even expensive ones, are very hard to overturn.

If practitioners can, they should try to use more robust methods. But, they may still want to communicate externally using the Sharpe Ratio and other simple and widely-accepted metrics.

Yesterday, I was looking at Sharpe formula. I agree with Greg and Christopher that it is a good measure of past performance of the system but I still don't see how it can tell us what system will do in the future. Anyway, a crystal ball that works has to be invented yet...

I think, we established that Sharpe is not as good as we want it to be but it is one of the best measures among the worst...

Christopher's combination of Sharpe ratio with maximum drawdown is very appealing to me. The next question is: what is maximum tolerable drawdown for us?

Everyone likes a Sharpe to be greater than the value of One, most look at an excellent value to be 1.5 some strategies fair well with less than value of 1, but any value less than .5 is not a Strategy that offers confidence. I look at Sharpe like the PH value where 1 stands for balanced strategy less then .5 not worth it, and greater than 1.5 something has to give, large draws to profits, care must be taken almost to good to be true.

I'd like to add the hint that the sharpe-ratio may not be a perfect measure but maximum drawdown isn't either. The maximum drawdown from a historical simulation is not a robust measure at all. If you would like to find out something about the pdf of the return of your strategy it is nice to have a look at the drawdown but it will not at all be a good estimator for the future.

I think a few important points have not been mentioned yet regarding the question if a trading strategy is acceptable having a specific Sharpe (or other measures in general).

You may look at the Sharpe of a backtest - fine! But in order to decide if the trading strategy is acceptable you have to take effects into account like:

1.) trading frequency

2.) modeling of your market impact

3.) modeling of the order execution probabilities

4.) adverse selection

( 3.) and 4.) are only relevant in case of limit orders )

If you designed a trading strategy where you have just the signals for a few trades it will probably be overfitted. You cannot find out if your signals are really statistically significant.

But if you do have a lot of trades and the Sharpe is high the signals might really be good.

The higher your frequency the better you may find out if your signals are statistically significant.

But in case of a high trading frequency your trading strategy might be very sensitive to your market impact or the way you simulate limit orders (if you use limit orders at all).

So I think the decision if a trading system is reliable is a problem that depends on the properties of the trading strategy itself and on the assumptions that you used to simulate the trading process.

You might also be interested in checking these two links, they start by elaborating on the weaknesses of Sharpe Ratio and then give some solutions for those weaknesses:

1) https://www.crystalbull.com/sharpe-ratio-better-with-log-returns/

2) http://allaboutalpha.com/blog/2014/07/30/what-is-right-what-is-wrong-with-the-sharpe-ratio/