Time to move on to the next quant strategy I want to highlight. This post will cover a strategy called Trending Value, aka the value stocks on the mend strategy. This strategy is the top ranked strategy by risk adjusted return (sharpe ratio) in the book What Works On Wall Street . This strategy shows the power of combining the three market factors that have been proven to outperform over time; size, value, and momentum. Actually you get two strategies out of this one since you start out with a powerful value screen. With the basics we learned in the previous posts we can easily implement these strategies now.
The Trending Value strategy consists of two parts. The first part involves doing a composite value screen on the universe of all stocks. By using the all stocks universe it opens the strategy to all companies with market caps of $200M or greater which provides a smaller cap benefit to this strategy. The value composite used for this screen, value composite 2 (VC2), is the same one I described in my last post on the Utilities strategy. It basically is a ranked list of stocks based on a composite score of P/E, P/B, P/S, P/FCF, EV/EBITDA, and Shareholder Yield. This part of the strategy is one of the most powerful strategies in the book. It’s called the VC2 high 25 strategy. It simply buys the top 25 stocks ranked by the VC2 score. From 1964 through 2009 this strategy returned 18% a year with a sharpe ratio of 0.7 and a max drawdown of 55.6%. My screen results as of June 10 are below.
Pretty compelling value scores; pe and pfcf of less than 9, pb and ps of less than 1, ev/ebitda less than 6. If you’re a tried and true value investor maybe you just stop there. You’ll do great. The only issue with pure value strategies is that they can under perform for long periods of time. Remember the late 90s? Would you have been able to stick with such a strategy when companies with no business plans, no earnings, valued on clicks were ripping higher year after year? These long periods of under performance are reflected in the base rate stats. This VC2 strategy has a base rate of 88% over rolling 3 year periods. That means 12% of the time your rolling 3 year return would not have beaten the market. These are great numbers, by the way, but just one the factors affecting an investor’s ability to stick it out. But we can do even better by applying a momentum screen to this powerful value stock screen.
The second part of the strategy takes the stocks in the top decile (top 10%) of the value ranked list. In excel you would do a PERCENTRANK function on the total composite value score column then do a custom sort for a value of 0.1 (10%) or less. Now you have the top decile of value stocks from the all stock universe. The last step is simply to sort the top decile of value stocks by six month price change, highest to lowest. The top 25 stocks on this final list would be the Trending Value portfolio. Now you can see the rational in the name of this strategy. It represents the top value stocks with the best momentum or price trends. Below are the results of my Trending Value screen that I ran at the beginning of the week.
As you can see from the table above the VC2 scores aren’t as good as the pure value screen but are still quite good. The benefit you get is higher returns, 21.08%, a higher sharpe ratio, 0.91, a lower max drawdown, -51% and higher base rates. This strategy has a 3-year rolling return base rate of 97% vs 88% for the VC2 strategy. Pretty powerful stuff. Now the hard part if you’re a value investor. Actually pulling the trigger and buying into the portfolio. Look at the last column in the table. How would you feel about buying a group of stocks that already went up 74% on average over the last 6 months? Not easy for most investors. But momentum is one of the most powerful forces in markets and it tends to persist. That’s what this strategy tries to take advantage of by buying a group of value stocks that the market has acknowledged, through price appreciation, are already on the mend. Yes, they are up big but from very low levels.
In summary, the Trending Value strategy combines the best of the three long term out performance factors (size, value, and momentum) to yield superior risk adjusted returns. Now you have all the tools you need to run the strategy yourself. I would recommend learning to run the screen, run it once a month for a while, and keep track it so you can see how it performs before committing any capital to it. I have one more screen to review, the Enhanced Dividend Yield screen, before I move on to some other quantitative investing topics.
59 thoughts on “ Quantitative Investing – Trending Value Strategy ”
Too bad that no one is offering these portfolios. Although these spreadsheets and calculations are up your alley, for me not so much! 🙂
It does make it harder to implement. And what keeps most individuals away. There are single factor strategies, like the consumer staples strategy using dividend yield which are screenable in the FINVIZ screener directly without using excel.
Paul: I reported a thoughtful review of O’Shaughnessy’s book ( http://wp.me/p1LlDo-aM ) and found it to be impractical, perhaps dangerous, for use by ordinary investors. 1. His winning portfolios require monthly or yearly rebalancing of 25-50 holdings, very expensive! 2. He is most confident about portfolios of 50 holdings, least confident about holdings of 10-25 portfolios. Ordinary investors desire about 25 holdings. 3. Readers should be cautioned that applying his winning strategies to 5-10 year holding periods might not achieve the same fantastic results as 40-52 year time periods. I continue to appreciate your blog and ideas. Thanks, Doug
Hi Doug, I think the strategies in the book can be quite useful for certain active individual investors. For “ordinary investors” you are probably right. But there are many active individual investors already implementing active strategies (active trading, options, individual stock picking) that could do much better with some of the quant strategies in the book and with a lot less effort.
As far as the strategies being expensive, it all depends on how much capital is employed. At $7-$10 a trade, fully re balancing a 25 stock portfolio costs $175-$250 a year. So, to keep fees at a super reasonable 0.25%, the portfolio would have to be $70K to $100K. Investors pay a lot more than that for many value, growth, or active ETFs (DVY for example is 0.4%). Not to mention the exorbitant fees charged by mutual funds.
As far as returns over short holding periods you are correct. All these strategies are about the long term. That’s why I highlighted base rates as being so important and why the book spends quite a bit of time on them and why investors should choose strategies with base rates. There are strategies with base rates of 100% over 5 year periods, and even a few with greater than 95% base rates over 3 years.
Thanks for the input.
Have just started the book but your blog has been the impetus to get me started down the road in that direction. I appreciate the details regarding the stock screeners and have started paper trading some of your ideas. Should be fun. Especially when I get to using real dollars. Keep up the good work.
Once we are living off of our investment income, should we still choose the best risk-adjusted return or focus more on high short term base rates and low max drawdown?
Jeff, it depends. I look at all three factors. high short term base rates make the strategy easier to stick with. Low max drawdowns are more important in retirement and also make strategies easier to stick with. I try and look at all three factors.
Have you started implementing this strategy with real money? If so, what percentage of your portfolio? I’m planning to once I have enough capital to make the balancing fees worth it. I’m in Vanguard index funds for now.
Nevermind, I see that someone asked this on your previous article. Thanks Paul
Sorry for all of the posts, I am very grateful for what you are doing.
I cannot view your tracking portfolio at FINVIZ in this and previous posts, which send me to the FINVIZ website, but not a portfolio. Maybe if on FINVIZ in the top right, you go to “my account” > “my profile” > privacy settings > share presets/portfolios. I’m not sure if I am the only one having this problem.
If the above doesn’t work, to be sure that we can screen correctly, could you post a screen shot of the final portfolio in excel based on FINVIZ with dividend yield and no EV/EBITDA? Hopefully it doesn’t matter too much based on intraday prices. It would be much appreciated.
Also your second screen shot above looks like you included price to sales (psps), but the text doesn’t mention this. Was this a part of the ranking?
Hi Jeff. I changed my profile settings. Let me know of that works.
And thanks for the catch on the text. Yes, the screen includes the P/S ratio.
The link still doesn’t work, maybe update it in the post. It might be a different address now that your FINVIZ portfolio is shared. Thanks Paul
That didn’t work either. I contacted FINVIZ to see what I can do. I just track mu portfolios on FINVIZ. I screen with Stock Investor Pro.
Jeff, FINVIZ says portfolios can’t be shared for tracking purposes. But you can get access to my portfolios as a registered user. They will open up in the screener. You log-in, go to collaborate and look for my user name paulnovell. You’ll see all my portfolios listed there. When you click on one it will automatically open up in the screener.
I’ll have to find something else for performance tracking. Or just publish results every so often.
Thanks Paul. Looks like we only have 10 stocks in common despite creating our portfolios a few days apart, but I think I did it correctly. It is comforting to see that we have additional ones in common when I go back to your April portfolio and the fact that O’Shaughnessy removed the time dependence of when you buy in. Sorry to bother you with this, but I don’t plan to buy Stock Investor Pro until its price plus trading fees are 0.5% of my portfiolio.
No problem Jeff. The screeners use different data sources, that’s part of the difference. Also, using dividend yield v shareholder yield, and not using EV/EBITDA is going to cause differences. I’m actually quite impressed with 10 stocks in common out of 25 considering these differences. With all these screens its important just to keep the basic premise in mind. We’re trying to generate a list of cheap stocks based on a composite of value scores (adding momentum for trending value). And those composite of scores have been shown to outperform over time.
Totally agree on waiting on Stock Investor Pro. That’s what I did.
Ohhh, I see. I was trying to make sure I did the Excel steps properly and thought I was comparing my FINVIZ exported to Excel screen to your FINVIZ exported to Excel screen and that only 10 stocks in common was based on the prices from different days. Now I am confident that I did the process correctly if we had that many in common and yours was from Stock Investor Pro. The confusion was that I knew the 25 stocks in the post was from Stock Investor Pro that you track with FINVIZ, but I incorrectly thought you also had a FINVIZ portfolio screened with FINVIZ. Case closed, thanks!
What are your thoughts on the tax efficiency of these quant. strategies vs something simpler such as index funds?
Hey Conrado, these quant strategies are not tax efficient compared to passive index funds. Its safe to assume you have 100% turnover a year in the portfolio. But there is enough extra return in these strategies to still make them well worth it. And compared to more active strategies and funds they fare better.
Did you test the strategies by some of the backtester? e.g. http://www.portfolio123.com
Hi Ellis, I saw no reason to backtest as that is what the book is all about. It has data going back to at least 1965, and 1929 for many strategies. Also, sites like Portfolio123 are expensive and only have backtest data going back 10 years which is of limited value.
The value I see in Portfolio123 is that they use the Compustat database. If I’m going to use a strategy on my family’s retirement portfolio, I want to be sure that my stock screen data is accurate.
Adam, I agree. Compustat is the way to go and that is why I made the with to P123 a while back.
How do you like P123 then? I’m seriously considering some of the O’Shaughnessy screens and wonder how the Trending Value method has done over these last few years. Do you still run the screen or use it at all?
I like P123 quite a bit. Once you get the screens set up it’s pretty simple and efficient. Yes, Trending value is one of my core holdings. Not too long ago I did an update on the trending value portfolio. See here .
Hi there, you might like the discussion about trending value on reddit:
and here’s a matlab code that performs the trending value algorithm by the author of the above thread (I didn’t try it out):
Cool. Very interesting thanks.
Hello again, the matlab code inspired me to create a web-based version. It’s open source, alpha and might contain mistakes (needs review). If you feel like taking a look here it is: http://finance.nmr.io .
Will definitely take a look. Paul
Thank you for the answer. I just created a new snapshot yesterday. I’d like to ask you for a favor: if it’s not too much of a hassle, could you take a look whether the result is similar to your excel system? Otherwise no problem, thank you for this inspiring guide
My last Trending Value Portfolio is as of July, 26. It had these stocks:
I wanted to see if we could run our screens on the same day and then compare results. I’d like to look into some of the databases differences. Yahoo seems to use Capital IQ as its main database where SI Pro use the Thomson database.
I figure we could run the screen after the closing of the market tomorrow Aug 9th and compare. I’ll post a couple of versions of my results.
I’ll do the same with the matlab script, which gets some data from Yahoo and some from FINVIZ, which I think uses Thomson. I’ll run 4 screens: blanks placed neutral or bottom and p/cf or p/fcf.
Sounds like a plan.
Here is the link to my 4 versions of the screens. I included the top 50 stocks in each version and highlighted the top 25.
From my results, ranking blanks or NAs in the middle or the bottom doesn’t make much difference. FCF vs CF makes more of a difference but prob
not enough to make much of a performance difference. From these results I would stick with ranking blanks or NAs in the middle and using CF vs FCF as O’Shaughnessy does although I’m going to dig more into the CF vs FCF differences.
Scratch the last part. Just downloaded the methodology paper from OShaughnessy. He uses FCF. Here is the formula he uses for the composite value score.
comp_value = sum(0,coalesce(pct_cfp_free,50),
Okay let’s do this! Since I’m in Switzerland that would be quite late so I’ll do it on Saturday – it doesn’t matter, right?
For the experiment I added an other page from where to see all 3000+ stocks ordered by rank (probably won’t leave it there forever since it’s a lot of data). I’ll add the links here as soon as it’s done.
As long as the databases have time to update after Friday afternoon’s close it shouldn’t be a problem. I usually have to wait until Sunday with the Stock Investor Pro databases.
Okay and here it is:
all data as csv
curious about what comes out
Thanks Nicolas. I’ll review and see what if anything different comes up.
Jeff I compared our results and there are some differences which I think could be fixed if we solve this: why is prchg_26w of TMUS (the second stock in your list) 93.65%? on finviz and yahoo it’s 27.42% http://finviz.com/quote.ashx?t=TMUS&ty=c&ta=1&p=d&b=1
Paul, not Jeff
Nicolas, this seems to be one of those special situations with the databases. TMUS did a 1-2 reverse-split in May 2013 and paid a $4/share (pre-split) special dividend. Cant get more convoluted than that. The Yahoo price does not include the special dividend – a hole in Yahoo data that I’ve often found. I think the FINVIZ an the SI Pro data assume the re-investment of the special dividend. Morningstar shows a 67% 6 month return, as of 7/31, and my hand calculation gives me about an 80% 6 month return. Anyway, the FINVIZ data is much closer to reality since you have to include the special div.
I created a spreadsheet similar to yours and found similar results, although we have different stocks with a decent amount of overlap.
by similar results, I mean in relation to the 4 different scenarios
Agreed. I’m pretty happy the results were as consistent as they were.
so yeah, I found the same patterns and effects as Paul, but I think the differences in specific stocks are due to database differences and maybe because I use tiedrank in MATLAB. That means if there are 10 stocks and 3-5 have the same rank, it would be 0,10, 30, 30, 30, 50, 60 ,70, 80, 90, 100.
That’s one of the reasons I used percentrank in excel.
Jeff, another difference is that I’m using a min market cap of $250M. Not sure why I did this since the book calls for a $200M min. I think I was trying to inflation adjust the $200M. But that’s a bit much. The $200M in 2009 is around maybe $212M today. Not a big deal but def is affecting the rankings somewhat.
Hi Paul, In doing a bit of surfing this morning I visited O’Shaughnessy’s website. His funds do not approach the kind of returns that you are getting with your screens.
If there is any interest in some registered, paying, subscriber/users for your proprietary screens that can be shared I would certainly raise my hand as a paying member.
Portfolio123 has reduced their base membership down to 30 bucks and if you calculated a fee for the monthly membership for you, it could really work out.
Does any other service offer the screening, backtesting, and subcriber capabilities that Portfolio123 has?
Very true Doug. His mutual funds suffer from 2 huge negatives – very high fees, and fund flows. No matter what the manager thinks a fund is subject to investor outflows so he’s forced to sell holdings to meet redemptions. I don’t think the open mutual fund structure is a good vehicle for these quant screens. A CEF would be a much better structure.
The AAII pre-defined stock screens are the closest I’ve found to what your looking for in a paid service. They could have a larger choice of screens but what they offer for the value is not bad. I agree it could be better and there is a niche to be filled out there. I’ve thought about doing maybe a basic quant screen paid service but its a lot of work I’m not ready to sign up for yet….
Portfolio123 is probably the best choice out there. The cheapest $30 option is per month and doesn’t give you the capabilities you need to run the screens I run here. You’d have to go up to the $83/month option to run your own screens. At $30/month, you need a $72K portfolio to keep the expenses at 0.5% per year.
Thanks for the research and food for thought.
Thanks Paul. Yes, think about it. It would only take a handful of us to get you in the green on your own costs of the developer levels. It appears that you are doing this work anyway. And of course new screens can then be developed as we learn.
Please don’t take this in any other way then the highest compliment. What you have shared and continue to share with us has been of great, great value. Thank you.
Comments are closed.