Stomaching the Standard Deviation

[mindofmush]

From Investopedia:
Modern Portfolio Theory states that adding assets to a diversified portfolio that have correlations of less than one with each other can decrease portfolio risk without sacrificing return. Such diversification will serve to increase the Sharpe ratio of a portfolio.

The seasonal strategies don't follow "Modern Portfolio Theory" therefore the Sharpe Ratio is inapplicable. Seasonal strategies are not portfolios because they invest in only one "asset" at a time; there is no adding negative correlating assets to the holding to reduce risk. Since we are not adding any assets to a "portfolio", we can't use the Sharpe Ratio to determine which "nothing" will be the best addition to that portfolio.

You can plug the numbers into the Sharpe ratio formula and the daily seasonal strategies with the lower "std deviation" will show higher(better) Sharpe ratios but it's the wrong tool for evaluating a seasonal strategy.

[MakeMe$$$$]

... but I can also put together "Rolling Strategies".

A rolling strategy is this. Say we are looking at a 5-year rolling:

For your 2005 strategy, you use the strategy that would have yielded the highest possible return for 2000-2004. For your 2006 strategy, you use the highest possible for 2001-2005. And so on, and so on. I would compare a 3-year rolling to a 5-year rolling and a 7-year rolling....

I'm kinda stupid on some things but...

I'm not sure I would strictly use a specific date range or even subscribe to it. My reasoning is that these strategies are nothing but brute force analysis of history without context in either technical or fundamental factors. Instead I like the current set up with quick date ranges but also the time bar you can drag around.

That doesn't mean a rolling notification system is bad, just not something I think is best...at least for now.

[bamafamily]

I have to respectfully disagree...the Sharpe Ratio is just a calculation of some average annual return (minus the risk free return) divided by the standard deviation....It does not have to apply to a MPT to be a qualifying calculation.
Something with a Standard Deviation of 25% and only a 3% average annual gain would be a terrible, volatile investment...and the Sharpe calculation would rate it so. (.12 in this case)

I agree that most uses for the Sharpe calculation can be found inside a single portfolio with variably correlated instruments.
I use it that way in addition to each of my portfolios having their own global Sharpe number. Each of my portfolios is variably correlated against the other portfolios..
So I have an external Sharpe number (my 5 portfolios) which contain strategies of instruments with their own Sharpe numbers....

Since the seasonals are either 100 or 0, I use the Outer portfolio Sharpe for keeping things correlated (<.75) the way I like....

Hope that make sense and I did not ramble off into oblivion.....

Bama

From Investopedia:
Modern Portfolio Theory states that adding assets to a diversified portfolio that have correlations of less than one with each other can decrease portfolio risk without sacrificing return. Such diversification will serve to increase the Sharpe ratio of a portfolio.

The seasonal strategies don't follow "Modern Portfolio Theory" therefore the Sharpe Ratio is inapplicable. Seasonal strategies are not portfolios because they invest in only one "asset" at a time; there is no adding negative correlating assets to the holding to reduce risk. Since we are not adding any assets to a "portfolio", we can't use the Sharpe Ratio to determine which "nothing" will be the best addition to that portfolio.

You can plug the numbers into the Sharpe ratio formula and the daily seasonal strategies with the lower "std deviation" will show higher(better) Sharpe ratios but it's the wrong tool for evaluating a seasonal strategy.

[bamafamily]

Anne,
The Sharpe ratio is calculated only using the CAGR(Mean) and the Std Deviation.
It looks for the highest return with the least amount of deviation...CSI has nothing to do with the calculation. Sharpe does not care if you attain your numbers by being in the market 99% of the time or 10% of the time....(although you probably get a lot less data points by being in the market less...not sure on that one..would have to think on it)
While 12367 is very good, 12216 edges it out based on the formula criteria...

If Sharpe Ratio is a measure for calculating “risk-adjusted return” I don’t understand it given the 2 mixes and why 12216 would be better than 12367? Looking at the data:

G . F .. C . S . I CSI . . .Mean . StDev Sharpe
24 25 . 4 32 15 50.80 32.14 . 12.83 . 2.35 http://tspcalc.com/seasonal.php?ID=12367
16 29 . 4 37 14 55.20 31.58 . 11.39 . 2.59 http://tspcalc.com/seasonal.php?ID=12216
12367 has 4.4% less exposure to CSI stock funds, 4% less in F Fund with 216% greater return. It has 1.44% Greater Standard Deviation and 2 more IFTs but if your goal is to sleep at night, it seems like 12367 is less risky for a greater return

[mjedlin66]

... but I can also put together "Rolling Strategies".

A rolling strategy is this. Say we are looking at a 5-year rolling:

For your 2005 strategy, you use the strategy that would have yielded the highest possible return for 2000-2004. For your 2006 strategy, you use the highest possible for 2001-2005. And so on, and so on. I would compare a 3-year rolling to a 5-year rolling and a 7-year rolling....

I'm kinda stupid on some things but...

I'm not sure I would strictly use a specific date range or even subscribe to it. My reasoning is that these strategies are nothing but brute force analysis of history without context in either technical or fundamental factors. Instead I like the current set up with quick date ranges but also the time bar you can drag around.

That doesn't mean a rolling notification system is bad, just not something I think is best...at least for now.

I think that once you see what I put together, your perspective will change.

[bassinger]

Yea I've been studying it for a while now. I'm very impressed.

[evilanne]

Anne,
The Sharpe ratio is calculated only using the CAGR(Mean) and the Std Deviation.
It looks for the highest return with the least amount of deviation...CSI has nothing to do with the calculation. Sharpe does not care if you attain your numbers by being in the market 99% of the time or 10% of the time....(although you probably get a lot less data points by being in the market less...not sure on that one..would have to think on it)
While 12367 is very good, 12216 edges it out based on the formula criteria...

I don't know how all the complicated formulas work, but if it is the Std Deviation that is driving the "risk-adjusted return" results of the Sharpe Ratio, it doesn't make logical sense IMO. Maybe it would make more sense with more data over a longer period of time but higher percentage of the time in the market would increase the risk and volatility.

[mindofmush]

The very low yields on T-bills and other bonds over the last 7 years have changed how institutional investors invest: with no safe alternative more money is going toward stocks than bonds.

I think mj's rolling strategy will incorporate this factor and automatically adjust for the higher bond yields as the Fed raises the rates over the coming years.

[Rome26]

I follow 7474. It has a StDev of 17.76 and CAGR of 35. It's worst year produced 18.6 with a best of 87.9.

[hmarkway]

I follow 7474. It has a StDev of 17.76 and CAGR of 35. It's worst year produced 18.6 with a best of 87.9.

Holy cow! I like the looks of 7474. Good find! I was seriously considering 7980 (http://tspcalc.com/seasonal.php?ID=7980 ... I-jahbulon), but now I'm also looking at 7474 (http://tspcalc.com/seasonal.php?ID=7474 ... marks=7980). 7474 has less time in I Fund earlier in the year, and a slightly higher mean (24.46% already for the year!). But for the remainder of the year, they are the same, so I'm going to just go with it, and decide what to do before they diverge in March. But the worst year and the best look pretty awesome.

[Rome26]

I follow 7474. It has a StDev of 17.76 and CAGR of 35. It's worst year produced 18.6 with a best of 87.9.

Holy cow! I like the looks of 7474. Good find! I was seriously considering 7980 (http://tspcalc.com/seasonal.php?ID=7980 ... I-jahbulon), but now I'm also looking at 7474 (http://tspcalc.com/seasonal.php?ID=7474 ... marks=7980). 7474 has less time in I Fund earlier in the year, and a slightly higher mean (24.46% already for the year!). But for the remainder of the year, they are the same, so I'm going to just go with it, and decide what to do before they diverge in March. But the worst year and the best look pretty awesome.

Yeah I'd happily take it's worst year for the next 17 years. That would put me just north of 3 million assuming I make the same contributions.

[EZlife]

I'm with you on that thought.