What do you think of the UPRO/TMF Strategy that beats the S&P 500?

[murray t turtle]

With the looming recession, I was considering: 25% UPRO / 25% ZIV/ 45% TMF/ 5% EDZ

Although, I have no way to back-test this. Any ideas how?

%% I like the name UPRO+ liquidity until about $ 52area or 4th quarter[OCT-DEC]whichever comes first.With first of JUNE + last day of MAY month being down+ summer rally being weakest, weak buy volume; may do a short term SQQQ. Small, small cap ETF/dividends position;took profits off DON, semi liquid mid cap/dividends.

 

[Magic]

Um, given that Rf is one of three terms in a Sharpe ratio... yes.

Alright, what does your math look like? That was the only parameter I could think of that would make a similar error but after digging in a little more I found even that wouldn't get us close to 1.

St. Dev from Jun 1995 - Present is about 18.8 (daily close to close vol). Even using annual periods for compounding which is the most generous for the sharpe calc, we get ~9.1% return. Even before backing out risk free rate, we're going to be south of 0.5 here.

If there's no math error and instead you're asserting sharpe of 0.5 is about 1; I don't think that's a very accurate statement for discussions of investment performance over such long periods of time. Running a 1.0 sharpe portfolio at the same vol or $ risk as a 0.5 portfolio is going to lead to a significant different in wealth generated inside even a decade.

 

[Sig]

If you pick 1995-28 I end up with an average of .84

Year Sharpe
1995 4.01
1996 1.48
1997 1.53
1998 1.14
1999 0.89
2000 -0.66
2001 -0.69
2002 -0.9
2003 1.59
2004 0.84
2005 0.16
2006 1.08
2007 0.07
2008 -0.92
2009 0.94
2010 0.81
2011 0.09
2012 1.19
2013 2.82
2014 1.18
2015 0.08
2016 0.86
2017 3.04
2018 -0.37

And given what Sharpe is measuring, if you stop and think about it qualitatively, Sharpe should converge on 1 for something that is representative of an optimized portfolio which the S&P 500 should.

None of which actually has anything to do with my point in any way, which is that the strategy in question doesn't have a risk adjusted return that is anything special, it's just a leveraged product, and it does have some definite tail risks, some of which are nearly identical to those that caused famous blowups in the past like LTCM. No amount of discussion about how many angels fit on the head of a pin change that, and as it happens that's the topic of discussion on this thread.

 

[Magic]

Averaging the sharpe distorts it beyond what it is intended to measure or convey.

Regarding generalities though; I agree with you that there are more hidden tail risks once you move off the beaten path. And that there are a lot better ways to hold risk than a basic levered stock/bond. This historically did out-earn SPY in part because it has more embedded tail risk—which didn’t end up realizing in the window of observation. That doesn’t show up on the sharpe, so we’re on the same page that optimizing for historic sharpe isn’t a special discovery.

I don’t see how conceptually sharpe should converge to 1 though. There’s no maxim that says we should optimally get paid risk premium equal to the st. dev. of the risk asset we’re holding. It makes sense to me that st. dev. will be ratioed higher than risk premium for massively scalable and passive investments, and diminish as a proportion to yield as those parameters diminish.

 

[Sig]

Averaging the sharpe distorts it beyond what it is intended to measure or convey.

Regarding generalities though; I agree with you that there are more hidden tail risks once you move off the beaten path. And that there are a lot better ways to hold risk than a basic levered stock/bond. This historically did out-earn SPY in part because it has more embedded tail risk—which didn’t end up realizing in the window of observation. That doesn’t show up on the sharpe, so we’re on the same page that optimizing for historic sharpe isn’t a special discovery.

I don’t see how conceptually sharpe should converge to 1 though. There’s no maxim that says we should optimally get paid risk premium equal to the st. dev. of the risk asset we’re holding. It makes sense to me that st. dev. will be ratioed higher than risk premium for massively scalable and passive investments, and diminish as a proportion to yield as those parameters diminish.

I'll kind of take back my qualitative statement. I think that if you think about what a standard deviation represents it makes sense that excess returns would have to be close to the same for the concept of investing in a diverse portfolio to make sense, but I wouldn't fault you if you called BS on that.