Use access key #2 to skip to page content.

TMFAleph1 (95.98)

Request for Ideas: A Quantitative CAPS Strategy



September 30, 2007 – Comments (3)

2:18 PM

My copy of Barron’s never showed up yesterday, so I had to pay $5.50 for it at the CVS. After that indignity, there were only a couple of articles that actually caught my attention, but none that seemed like “must-read” material (although the interview with Mr. Grundlach, a bond fund manager at TCW, looks instructive).

On to a different topic. Today I am soliciting from the CAPS community suggestions for a quantitative approach to CAPS. I’d like to put together a systematic strategy for making CAPS picks based on a sorting algorithm. This is what I’m looking for: I’d probably start with the 500 constituent stocks in the S&P 500 (a good starting point, in my opinion, since the CAPS benchmark is the S&P 500). The idea is then to rank the 500 stocks based on one or multiple criteria. Finally, I would pick the top decile to ‘outperform’ and the bottom decile to ‘underperform’. Of course, the key in this strategy is the choice of your ranking criteria. I’m leaning towards a fundamental approach, rather than a technical approach (although there are momentum effects in the short-term), but I’m ready to look at all suggestions.

I have found a few books on Amazon on quantitative approaches to equity portfolio management that may be useful, but I’d be interested to know what CAPS members would suggest. I’d be glad to get references for academic papers, books or ideas that the members have toyed with.

Thanks for your suggestions!

Total: 250 words

Time: 10.75 minutes (w/o spellcheck)


*** The above text was written (and spell-checked) in ten minutes. As a result, some of it may not stand up to rational scrutiny. I apologize preemptively for any errors, omissions and misrepresentations. ***


3 Comments – Post Your Own

#1) On September 30, 2007 at 4:02 PM, TMFAleph1 (95.98) wrote:

*** UPDATE *** 

I wanted to make it absolutely clear that this is not an official request from the CAPS team; it is purely in support of my personal project.

Thanks again,


Report this comment
#2) On September 30, 2007 at 7:56 PM, bubuzim (95.70) wrote:

One of my previous advisors, Tomaso Poggio @ MIT, has had a good deal of success using support vector machines (svm) to decide which stocks to pick and when to buy/sell.  It's a very technical-based approach, but you could insert fundamentals and caps player results (whatever data you can think of) into the algorithm as well and try different weightings.
I would let the algorithm do the screening for the top 500 (running a binary SVM for each candidate stock with buy/sell as the output and choosing the top 500).

I'd love to fool around with the data as well.  It would be interesting to know how the final weights end up... that could put a value on the ability of the caps system to predict the market.

Report this comment
#3) On October 01, 2007 at 10:25 PM, StockSpreadsheet (74.73) wrote:

Among the criteria I look at are PE < 50, PEG < 1.5, falling debt-to-equity ratio (or if it is already at zero), sales CAGR, EPS CAGR and dividend yield (and whether the dividend has been raised over the past 5 years).  I have always thought of this as a technical approach but others have told me it is more fundamental, so I don't know if it would fit into your desired approach.  I also like looking at the 50 and 200-day moving averages to see if they are converging or diverging, though I don't know how you would put that into your formula without calculating each yourself, (which would be a pain), as I don't know where to get them as anything other than a graph, (like on Yahoo! Finance).  I also like looking at the S&P star rating and the Morningstar ratings, but since not all of the stocks I look at have ratings, (for instance, Morningstar rates NO ADRs and S&P does NOT rank most small cap stocks), then these inputs are often non-existent.  Anyway, if you plugged in those variables into your equation, you would basically have my approach, (minus the shorting aspect, which I don't do), so you could look at my results to see the results of those inputs into a formula, (except that mine is done all by hand, so it is labor intensive.  If you could automate it, so it could run on a daily/weekly/monthly basis, you could get more up-to-date data to make decisions on, if you wanted to trade more frequently than the 6-month minimum timeframe that I use, (though my timeframe is usually years, not months)).  Anyway, hope this helps.  If you get all of my inputs automated, I would be interested in how you do it to see if it would be applicable to my purposes.   Good luck, take care and may all your picks be winners.


Report this comment

Featured Broker Partners