Does Sharpe Matter if Sortino is High?
From my understanding of sharpe and sortino, sharpe discounts the ratio for all volatility and sortino discounts the ratio for downside volatility only. I trade a system of algorithms in my personal portfolio and all of them are high sortino (2+), but sharpe ranges from .7-1.2. I have been pleased with this and love the way that it performs. Why is there so much focus on sharpe? Is high sortino valued in HF?
Don't know but my guess is Sharpe was made first and became popular as the first way to measure risk adjusted returns. Sortino is probably better but just never gained as much traction b/c it may have came later.
Sortino focuses on downside volatility, which is a bit more convoluted to calculate compared to regular standard deviation (in Python that would mean manually coding for sum of ([negative return - average return]^2)/n for downside volatility while regular volatility would just use the .std() function).
Also there’s less datapoints for downside vol compared to regular vol, which may pose a problem when it comes to having a big enough sample size to accurately measure downside risk.
Hopefully hedge fund guys can code that super easily. It can be done in one line.
Hedge funds make outsized returns because they lever up on high sharpe strategies. If your sharpe is below 1 then it’s way too risky to lever up on.
Why would I lever up on high sharpe if I could just not lever high sortino ratios and get better returns? Also, I’m not sure if it will stay this way, but through the live testing of my system, I have a sharpe above 2.4 and a super high sortino. My strategies are running on 10+ uncorrelated underlyings and the strategies themselves take advantage of different setups that happen in all market conditions. My equity curve is very straight and I feel like the strategies complement each other very well to balance out to a pretty consistent equity curve. Would it make sense for a diversified system of algos that individually don’t have high sharpes, but together have a combined high sharpe?
Sharpe is a more comprehensive view of risk, so that’s part of the standard that a allocator (whether investor or risk team) will consider, along with things like net, VaR, etc. I’m not saying a sortino isn’t useful at all, it’s just one side of the picture.
It makes sense that you can bundle low sharpe strategies into a larger portfolio with higher sharpe. That’s the diversification benefit.
for those who took one basic statistics/risk class, actually understand the course content and apply the knowledge outside the exam period - sharpe in your case does not matter and it should not. sortino is a superior measure for risk adjusted returns. simple example from my work experience: I once worked for a macro hedge fund with a sharpe > 2.0. in 2020, we were up high double digits which killed our sharpe ratio simply b/c we "deviated" but to the *upside*. our sharpe ratio (includes standard deviation) went down while sortino (includes downside deviation only) went up massively. we started reporting the sortino ratio instead and had to explain to seasoned asset allocators why this statistical measure is more accurate...remarkable who advances in this game.
This. Sortino is objectively better than Sharpe because of the exact situation you described. If you outperform drastically your Sharpe will be penalized
Short answer is... SHARPE is what is in the investor mandates/prospectus.
We all know the pros/cons of diff risk metrics. But sharpe is what investors understand and have used for decades so that is what is important. If you pull a term out they don't recognize they get cold feet and think you're trying to pull a fast one.
It depends, Sharpe is better than Sortino in many contexts. The question to ask is what is the driver of the upside volatility? In many strategies (especially algorithmic) the largest upside swings are rand not skill e.g. the best trading days may be holding through a catalyst that isn't explicitly being bet on - and while it may have "worked out" historically, including it in the sample when measuring risk is often a helpful measure of true risk taken as it could have just as easily gone the other way. For most strategies (with some exceptions) investing is a game of a ~51-55% hit rate where skill is shown in the averages and not the right tail, so both sides of the distribution have useful information on the true risk taken.
Whenever someone presents Sortino over Sharpe I am immediately skeptical - the choice to pick the less conventional metric is almost always taken after the fact and primarily because it looks better, so there is already a lot of selection bias. Sortino > Sharpe when the upside vol has historically been higher than the downside - unless the track record is long or the strategy has a good reason to have that skew, priors would suggest that at least some, and probably a lot, of that skew is rand (keep in mind in your case, had it been the other way around with the Sharpe being higher than the Sortino, you probably would not be asking this question in the first place).
There are of course some good examples of where Sortino is a clearly better metric, especially in right tailed / long volatility style strategies (buying options with capped downside, shorting rates at the start of 2022 where there is a logical floor on losses, etc.).
My two cents below.
Sortino is a better measure, but slightly more complicated to use and thus less adopted. For example, do you zero out the positive returns or do you remove them? What is a good sharpe? Everyone have an opinion on good sharpe while most investors don’t have on sortino.
If you zero out positive returns you might end up with not enough datapoints so sortino would be less robust and you need much more data (especially if high sharpe negative skew strat).
If you have one big outsized return (cocoa this year or long vix call during covid), which is the dream of every money manager, your sortino improves dramatically but your sharpe suffers, which is counterintuitive given what you are actually asking to a “risk adjusted return” measure.
Another way is incorporating sharpe with skew. Most of the very high sharpe strategies (stat arb, vol selling, etc) have negative skew that may blow up your portfolio if the leverage is too high in the wrong moment. If you look to develop strategies with skew around zero or positive skew the blow up risk is lower. You can still close down the fund due to return bleeding though because you are seeking positive skew strat only (think long vol or CTAs during some period of time).
Jack Schwager (market wizards author) proposes a measure such as sharpe / sqrt(2), this measure becomes, according to him, comparable to the sortino. You will see that often if sharpe/ sqrt(2) is > sortino then strategies have negative skew and viceversa, i found it very interesting and useful.
Both measures are fine to be used but be ready to defend the sortino as everyone is much more familiar for sharpe due to earlier adoption and easier computation.
Strategies with 0.7-1.2 sharpe are fine if the skew is not that negative, honestly positive skew would give me much more confidence. You can also for robustness check the daily pnl: average positive days should be higher than avg negative days and the most positive day ideally is higher than the most negative one (in abs). All these go in the same direction.
This is my summarized view on the topic, hope helps
The negative skew is a good point, especially because it often cannot be measured from the history of returns and is hard to quantify. This is also the problem with the VAR or tail ratio, it looks more reasonable than Sharpe on the surface but cannot be measured accurately from historical data. Even "good Sharpe" means different things to different people, a smart beta strategy versus an HFT market making strategy will look very different.
The guy above is right, this is primerily about skew. Right skew strats are preferred bc optimal leverage is higher for the same variance. But Sharpe has a bunch of profound meanings / dualities which you loose with Sortino so most just stick with Sharpe.
Fugiat eius dolorem ab itaque enim minus. Possimus cumque ea et voluptatibus incidunt voluptatem. Tenetur nulla quod aspernatur vel esse. Qui voluptatem reprehenderit laudantium. Non cumque illo cum doloremque architecto omnis temporibus. Debitis tempora inventore dolorem in nihil dolor.
See All Comments - 100% Free
WSO depends on everyone being able to pitch in when they know something. Unlock with your email and get bonus: 6 financial modeling lessons free ($199 value)
or Unlock with your social account...