Isn't it quite useless if the returns are not normal? It favors strategies that consistently profit but have a rare but catastrophic loss. It seems to be a very popular metric so I can only assume most strategies would have this bias. Why is the Sharpe ratio used, and are there any alternatives to it?
http://www.tangotools.com/ui/ui.htm has a comparison of the Sharpe Ratio and the Ulcer Performance Index.
Why would the non-normality of returns make the Sharpe ratio useless? If you have two strategies that both yield 20%/year and one has Sharpe 1.2 and the other 2.0 I know which one I would pick. That the return distributions are fat tailed doesn't change that the return is accomplished for less risk on average.
You under-appreciate the meaning of lower average risk. Basically, being in a drawdown sucks since you either need to wait it out or exit with a loss. Higher sharpe ratio means fewer drawdowns. In addition, it tends to be the case that a higher Sharpe strategy will simply have a (potentially significantly) lower maximum drawdown than a lower Sharpe strategy. The largest drawdowns are typically caused by longer sequences of bad trades, not by a single bad return (but there can be exceptions depending on strategy, for instance a highly levered high beta strategy getting killed by a gap). As you observe, the worst case tail event probably most certainly didn't happen during your backtest and something truly horrible can always happen. This is not related to the Sharpe ratio though. It really boils down what you are using the Sharpe ratio for. If you are using it to say X percent of returns will be below Y in absolute magnitude or something similar, then you're in deep water. However, for comparing one strategy to another it is great, and is typically reliable for that provided sufficiently large number of samples.
You nailed it- the Sharp ratio is not a good metric for traders. The Sharpe fails to distinguish between upside versus downside volatility and assumes the markets follow a normal distribution around a curve which is dead wrong. The Sortino Ratio, Gain-to-Pain, Calmar, Profit Factor, & closed NAV with worst draw down are more practical metrics for traders. The Sharpe ratio was created to measure mutual fund performance - a whole different animal. http://www.redrockcapital.com/Sortino__A__Sharper__Ratio_Red_Rock_Capital.pdf https://oxfordstrat.com/ideas/sharpe-ratio/ https://unovest.co/2017/08/sharpe-ratio-problem/
While you're correct, Sharpe ratio is still a pretty good measure of risk/return. I doubt you'd want to trade something that had one big equity spike and did nothing since. I believe this would have a high Sortino.