Magic Formula Portfolio123 Screener

In addition to ranking stocks by the two factors in the Magic Formula, there are several criteria that Joel Greenblatt has developed to screen out companies before any ranking occurs.

From what I have gathered through The Little Book That Beats the Market, stocks from the utility and financial sectors are eliminated.  This is primarily due to the unique capital structure of these industries.  In addition, ADRs (foreign stocks) are also filtered out.  My guess is that these stocks are filtered out because of accounting inconsistencies and the unreliability of data regarding foreign stocks in many databases.

I have also set up the screener to eliminate stocks that have less than a $50 million market capitalization.  I’ve done this to come close to the 3,500 highest market capitalization stocks that Joel Greenblatt used when he backtested the Magic Formula. This last step is not a requirement, and in fact, the Magic Formula Investing site let’s you specify the minimum market capitalization of your choice.

I entered the following criteria into the Portfolio123 screener:Portfolio123 screener criteria

  1. Universe($ADR) != 1
  2. Sector != Utilit & Sector != Financ
  3. MktCap > 50

The final screening criteria concerning Rank just returns the stocks with a rank higher than the one I specify.  I adjusted the Rank up and down to get a list of about 100 stocks, so I could compare these 100 stocks with those produced at Magic Formula Investing website.

I ran the screen and I got a list of just over a hundred stocks.  I then visited the official Magic Formula Investing site, and I brought up the 100 stock list of companies over $50 million in market capitalization.  I discovered that 39 out of 100 stocks on the Portfolio123 screen were on the Magic Formula Investing list.  That is not a very good correlation.  I was hoping to get at least a seventy percent overlap between the two lists.  After the initial disappointment, I decided to try a few simple things.

First, I decided to remove the excess cash calculation out of the equations, since I was most uncertain about this calculation.  I replaced my excess cash estimate (cash minus 5 percent of sales) with total cash.  The results were not much better.  I got 40 out of 100 stocks to match the Magic Formula list.

I also decided to try removing the cash calculation all together from the formula.  This results in an equation that assumes that no company has excess cash.  That resulted in only 32 of 100 stocks overlapping with the official Magic Formula Investing results.

Finally, I tried to use Joel Greenblatt’s alternative formula for people who don’t have access to a screening tool that has the data required for the Magic Formula.  The alternative formula simply looks at earnings divided by price per share for earnings yield and return on assets (ROA) for the ROC quality component.  That formula was real easy to set up in Portfolio123.  However, this time I only got 18 out 100 stocks to match up with the official Magic Formula results.

I’m puzzled by my results.  Am I missing something?  I wish I had access to Joel Greenblatt to find out exactly where I am straying from his formula. I’m having a hard time believing that it could just be a difference between the two underlying fundamental datasets being used.  The differences are just too great.

I’m not giving up quite yet.  I’ll be trying a few more things out and sharing them here with you over the next couple of days.  I’ll be looking closely at some of the top ranked stocks on the Portfolio123 screen but not on in the Magic Formula list.  I also plan on backtesting my original Portfolio123 version of the Magic Formula and comparing the results with the backtest numbers published in The Little Book That Beats the Market.

14 thoughts on “Magic Formula Portfolio123 Screener

  • April 10, 2006 at 2:29 pm

    Have you looked at the stocks which appear on Greenblatt’s screen but not on yours? If you picked a sample of 10 of those, you could figure out whether the deviation seems to have more to do with the ROIC calculation or the EY calculation. That would allow you to narrow down the cause of the discrepancy.

  • April 10, 2006 at 2:49 pm

    Also, the difference could come from different data sample. Joe’s data may not be the same as yours (some older, some newer)

  • April 10, 2006 at 3:30 pm

    One other thought: I wouldn’t be surprised if one of the biggest drivers of the discrepancy is the EBIT number itself, which drives both Magic Formula metrics. Even if both you and Greenblatt’s team are using EBIT – TTM, there might be a significant difference in how “one-time” or “special” items are dealt with (either by the two different databases — Compustat and Reuters, right? — or by Greenblatt’s team). Such items are so common these days that they can be a major factor.

  • April 10, 2006 at 9:52 pm

    David – I’m working on going through some of the top stocks on each of the lists to see why they might not be on both lists. I’ll post what I find in the next few days.

  • April 10, 2006 at 9:54 pm

    Jim and David – It is likely that the data might be somewhat different between the two databases. I’m having trouble believing it is that different, but it might be. David, the EBIT numbers from Reuters I don’t believe adjust for “one-time” or “special” items. Does anyone know if the Compustat Point in Time database does?

  • April 13, 2006 at 11:52 am

    You might want to check out some of the posts at the Yahoo group on the Magic Forumla. Some folks there have gone through a lot of work to figure out screening parameters for different databases, especially AAII’s Stock Investor Pro. I believe that one group member was able to get about 70% overlap with the MFI site. Unfortunately, the posts with that info may be a bit hard to find!

  • April 13, 2006 at 8:18 pm


    not to nit-pick. a couple of comments:

    does it really matter if your screening method gives you low correlation to the stocks from the MFI websitte ? Since the basic formula is ” low valuation vs high roc “. The stocks will have good prospects for being cheap while exhibiting high management effectiveness (quality).

    Assuming the “little book” stays on the best seller list for the whole 2006 and sells one million copies, of which 3 of of 10 readers actually goes out and buy a basket of stocks. That’s 300k investors looking at a “list” of stocks to buy. Assuming the list is constantly evolving and rolling, that’s still alot of investors buying the same stocks. How is this Intelligent Investing ? As a clever person once said, you can’t make money from the obvious.


  • April 14, 2006 at 5:29 am

    Hi Dave – I have spent a good amount of time over at the magicformulainvesting Yahoo Group. I’ve gotten quite a few good ideas from them. I don’t think anyone has gotten quite as high as 70 overlap with the MFI site but I do remember reading some posts that discussed getting an overlap of sixty something percent.

    raytoei – I do agree that it doesn’t really matter how close my correlatoin to the MFI screen gets. Actually, I just ran a backtest of my results to 2001 and they are very similar to Greenblatt’s results. I just wanted to get a high correlation so as to develop some confidence with the formula.

    I am not concerned about how popular the Magic Formula is getting. I’m not currently investing using the formula, but I am evaluating its effectiveness to determine if it is something I want to add to my personal toolbox. By the time I convince myself that the Magic Formula really works well, it probably won’t even be popular anymore. ;-)

  • October 8, 2006 at 6:59 pm

    Here is a site using advanced statistical mdoeling (pattern recognition) to forecast stock prices. This model will narrow down the magic list based on the model ability to generte profit based on past foredcast adn evaluating its performance in the last year and last 3 months. If the model failed to generate profit no smart invstors could do it either

  • April 14, 2009 at 11:18 am

    What is the status now on the Magic Formula? Is 123 still off by such a large margin? How has the portfolio fared? I know the little book says there is NO 3 year period when it has lost compared to Mr. Market and the blog post was almost 3 years ago :D

  • July 25, 2009 at 11:55 pm

    I’ve used ROA in place of ROC in the past and that is a mistake. Assets on the balance sheet comprise the objective and the subjective (and can be greatly overvalued). You can find a nice formula on the web allowing you to calculate ROC by knowing a company’s ROE and it’s debt to equity ratio. Since ROC is not commonly presented on most financial pages (again, ROA is useless), try the conversion number and see where your rankings move. The formula is ROC = ROE / (1 + DTE), where DTE is debt to equity.

  • March 23, 2015 at 1:23 pm

    guys ,
    any success in replication .. if not the returns has someone been able to match the daily screens from magic formula ??

Leave a Reply

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.