No. of Recommendations: 113
I've been working the Screen Optimizer spreadsheet do death for the last few days, trying to come up with a new blend of screens for my portfolio. I'm not quite there yet, but the first phase is completed and can be shared.

Over the last two years I used a blend of five screens, buying the top four stocks in each, and rebalancing monthly. I plan to continue in a similar pattern, but am not sure yet whether I will hold 20 stocks as before, or maybe 24 or 25. There are several combinations of screens that could work. I could go with 5 screens, top 4 of each, as before. Or I could go with 4 screens, top 5 of each. I could go with 6 screens, top 4, for a total of 24 stocks, or 5 screens, top 5, for a total of 25. The table below reflects those possibilities. I also tested a 3 screen blend, top 5, as a courtesy to others who make seek a portfolio with fewer stocks.

In the past I optimized my blend to find the highest Sharpe ratio. This time I've added other measures - Sortino Ratio, Information Ratio, Upside Potential Ratio. Each optimization target produced different blends.

Here are brief definitions -
Sharpe Ratio - Average differential return divided by stdev of differential return, where the differential return is the difference between the screen and the risk-free rate.

Sortino Ratio - Uses the same numerator as the Sharpe Ratio. The denominator is the downside deviation, a form of stdev that subsitutes zero for all positive deviations.

Information Ratio - Calculated the same as the Sharpe Ratio, except that instead of using a risk-free rate it uses a benchmark investment rate. I used the S&P500 as the benchmark.

UP Ratio - The ratio of upside deviation to downside deviation. The formula for these deviations is not exactly the same as the one in the Sortino Ratio. The deviation here us the average of the squares, which rewards large upside deviations and strongly penalizes downside deviations.

I'll let the results speak first, and then add some observations.
                Positions 1-4                                     Positions 1-5                 
Max Sharpe
Screens 5 6 3 4 5
PIH_CSO PIH_CSO PEG-Minimalist PEG-Minimalist PIH_CSO
PEG-Minimalist PEG-Minimalist LOWPE5ZLTD H52EarnPS H52EarnPS
H52EarnPS H52EarnPS YLDEARNYEAR LOWPE5ZLTD LOWPE5ZLTD
LLTD YLDEARNYEAR YLDEARNYEAR YLDEARNYEAR
YLDEARNYEAR2 YLDEARNYEAR2 Screamers
LOWPE5ZLTD
Sharpe 2.18 2.20 2.09 2.16 2.21
Sortino 5.21 5.36 5.21 5.24 5.66
Info Ratio 1.94 1.82 1.85 1.88 1.91
CAGR 42.96 40.60 38.60 39.11 41.42
GSD 15.76 14.62 14.67 14.32 14.87
UPR 2.46 2.49 2.52 2.48 2.61


Max Sortino 5 6 3 4 5
PIH_CSO PIH_CSO LOWPE5ZLTD LOWPE5ZLTD LOWPE5ZLTD
YLDEARNYEAR LLTD YLDEARNYEAR YLDEARNYEAR YLDEARNYEAR
RS4WKT12 YLDEARNYEAR Screamers YLDYEAR YLDYEAR
YLDEARNYEAR2 YLDEARNYEAR2 Screamers Screamers
REV GAR4 LPE_YLD
REV

Sharpe 2.07 2.10 2.05 2.10 2.11
Sortino 5.99 6.08 5.82 5.91 5.94
Info Ratio 1.83 1.86 1.84 1.70 1.77
CAGR 45.22 43.93 41.27 36.96 36.11
GSD 17.39 16.66 16.01 13.80 13.40
UPR 2.91 2.89 2.81 2.82 2.84


Max UPR 5 6 3 4 5
LOWPE5ZLTD LOWPE5ZLTD LOWPE5ZLTD LOWPE5ZLTD LOWPE5ZLTD
YLDEARNYEAR YLDEARNYEAR YLDEARNYEAR YLDEARNYEAR YLDEARNYEAR
RS4WKT12 RS4WKT12 Screamers YLDYEAR YLDYEAR
YLDEARNYEAR2 YLDEARNYEAR2 Screamers Screamers
GAR4 GAR4 LPE_YLD
REV

Sharpe 2.01 1.97 2.05 2.10 2.11
Sortino 5.85 5.85 5.82 5.91 5.94
Info Ratio 1.82 1.83 1.84 1.70 1.77
CAGR 45.16 49.27 41.27 36.96 36.11
GSD 17.93 18.75 16.01 13.80 13.40
UPR 2.91 2.92 2.81 2.82 2.84


Max Info Ratio 5 6 3 4 5
PEG-Minimalist PEG-Minimalist PEG-Minimalist PEG-Minimalist PEG-Minimalist
LLTD LLTD EGPLOW_PE LLTD LLTD
RS4WKT12 YLDEARNYEAR ZLTDA Screamers ValueRatio
LPE_YLD RS4WKT12 LPE_YLD Screamers
SLS_RS26 LPE_YLD LPE_YLD
RSWEPS

Sharpe 1.94 2.05 1.92 1.98 2.07
Sortino 5.05 5.48 4.88 5.35 5.57
Info Ratio 2.21 2.32 2.02 2.13 2.15
CAGR 44.69 45.65 45.73 42.18 39.76
GSD 18.71 17.94 19.42 17.08 15.25
UPR 2.52 2.67 2.45 2.63 2.69


You'll note that there are some relatively new screens in these blends. As I was working on this a couple of new screens surfaced and I decided to bite the bullet and start over each time. I think the extra effort this one time may be worth the payoff in the coming year(s). At least I hope so.

The screens in each blend are listed by descending Sharpe ratio. It an almost dizzying array of choices. Which parformance measure is best? Which combination of screens and positions? I don't really know yet. Statistically all the results are so close as to be indistinguishable. That's a good thing. There are many equally good choices in this smorgasbord, increasing the chance that people who use this table will end up with different screens and not all pile up into the same stocks.

As I executed the optimization runs, some of them more than once, I saw that the slightest change in conditions can produce a different blend even when the optimization goal is the same. For example if I added a few screens to the evaluation I might get different results, even if the added screens did not in fact participate. It goes to show that there are many more blends under the covers with virtually identical backtested results. The optimization is a heuristic process, and different starting conditions can produce somewhat different results.

The fact that in many cases the blends with 4 stocks per screen produced different lists than the blends with 5 stocks per screen is an indication, IMO, of how limited the value of the backtests might be. I don't think a screen can really be inferior as a top 4 and much better as a top 5, or vice versa. These differences are statistical noise.

As noted, this was only stage one of my selection process. The next steps will be to evaluate the frequency of overlaps in the blends. I don't like a lot of overlaps, and obviously from the screen list in some of them you can tell that there will be plenty of overlaps. The backtester likes the YLDEARNYEAR family so much that it tends to put two of those in a single blend. I will also analyze turnover, the less the better. And I will analyze the effectiveness of HTD. I hope to come up with my new blend in time for next week's rebalance.

This information is provided for anyone to use as they see fit. I'd rather that you each make your own decisions, and we get some diversity, than having too many follow too closely in my footsteps. Should I let people know what I decide to do or leave you with what I've done so far and let everyone come to their own conclusion?

Any other comments are welcome.

Elan
Print the post  

Announcements

What was Your Dumbest Investment?
Share it with us -- and learn from others' stories of flubs.
When Life Gives You Lemons
We all have had hardships and made poor decisions. The important thing is how we respond and grow. Read the story of a Fool who started from nothing, and looks to gain everything.
Contact Us
Contact Customer Service and other Fool departments here.
Work for Fools?
Winner of the Washingtonian great places to work, and Glassdoor #1 Company to Work For 2015! Have access to all of TMF's online and email products for FREE, and be paid for your contributions to TMF! Click the link and start your Fool career.