1. How to use GuruFocus - Tutorials
  2. What Is in the GuruFocus Premium Membership?
  3. A DIY Guide on How to Invest Using Guru Strategies
Mark Lin
Mark Lin
Articles (212) 

Imperfections with the Piotroski F-Score

November 11, 2012 | About:

The F-Score, developed by university professor, Joseph Piotroski, scores companies on a list of nine parameters related to profitability, operating efficiency and leverage.I screened for stocks with a perfect score of 9 for the F-Score and a P/B ratio of less than 1. Three stocks, MOD-PAC Corp. (NASDAQ:MPAC), P&F Industries Inc. (NASDAQ:PFIN) and Nortel Inversora S.A. (NYSE:NTL) passed the test with flying colors.

As a result, I found imperfections with the F-Score that I would like to share with readers.Firstly, the F-Score focuses on improvement, rather than absolute levels. MPAC improved its gross profit margin from 17.14% in the previous year to 17.63% this year. The increase of 0.49% in gross margin and 17% gross margin by itself is nothing to shout about. Also, PFIN's debt/equity decreased from 49% last year to 34% this year. Despite a significant reduction in leverage, a debt/equity ratio of 34% is still considered high. By focusing only on improvement instead of absolute levels, firms with low profitability and high leverage can still get a perfect F-Score. Secondly, the F-Score is not timely for contrarian value investing. If you wait for the robins, spring will be over.

For many stocks especially the depressed cyclicals, the best time to enter into a position usually happens when the fundamentals of the companies are at their trough. The F-Score will pick up only these depressed stocks when there is an indication of improvement in financial metrics. By that time, the stock prices will have priced in the recovery, making valuations unattractive. Lastly, the time frame for the F-Score is too narrow.

The F-Score typically compares this year's financial metrics against those of last year and could create distortions in perception. For example, the F-Score requires a positive divergence between cash flow from operations and net income before extraordinary items. It will be a fairer assessment if the divergence between cash flow operations and net income is evaluated on a multi-year basis. Cash flow from operations could fall below net income in any single year as a result of the timing in recognition of net income.

About the author:

Mark Lin

Rating: 3.8/5 (9 votes)


Systematic Value Investor
Systematic Value Investor - 5 years ago    Report SPAM
Something does not have to be perfect to work.

As long as it is a strategy that makes sense and has been proven to work over a long time... If the principles are implemented through a disciplined strategy, it will give you an edge over time. It has been proven to work over the long term when you hold a diversified portfolio (30+) of these highly ranked stocks chosen within the lowest p/b ratios pool of stock (lowest 20% p/b).

That's why it is important to be well diversified. Quantitative screening won't just magically find the three best bargain stocks for you right now. It's just helping you find above average opportunities, on average. I would not give it too much weight to the F-Score alone unless it is implemented into a long term, diversified strategy with strict, unemotional buy sell decisions (1 yr holding period).
Jhodges72 - 5 years ago    Report SPAM
Excellent summary. Same conclusion I came to a few years ago during the height of the recession. Buffett warned investors to be cautious of people advocating formulas in determining if an investment is worthy. That's simply not what value investing is about.

Lastly, it's a silly notion to state one is best to be diversified in order to protect yourself in case a basket of F-Score stocks don't work out. That simply isn't rational thinking. It's far more rational to know what you're doing in a few stock selections than to rely on a formula based on someone else's work which very few have the mathematical knowledge to understand.
Davidchulak premium member - 5 years ago
There are imperfections with any tool including P/S, P/E, EV/EBIT, etc. You must recognize that these are only screening tools that should not have too much importance attached to them. They must be looked at as a whole.

Interestingly, AAII shows the following results for the Piotroski screen:


Not too shabby
Davidchulak premium member - 5 years ago
Not sure why the first column was deleted, however; in order they are YTD, 3 YR, 5 YR, 10 YR and since inception
Marklin premium member - 5 years ago
There is nothing right or wrong with quantitative value investing per se. However, one should understand the reasoning behind the quantitative tools and formulas before using them. Quantitative tools can be a useful starting point, but not necessarily an end point.
Rrurban premium member - 4 years ago
The Piotroski score will give you a list of small cap stocks with very irratic histories and unpredictable futures. Using the Piotroski score as a "starting point for further research" is a total waste of time and effort. Fundamental analysis will be of no help and even the CEOs of these companies cannot predict their companies futures. These companies are unpredictable and thus are priced accordingly. Graham and Schloss used to buy low P/B and Net/Net stocks for years, usually holding over 100 at a time. they also did this with little fanfare and/or competition from other investors. That strategy no longer exists so Piotroski is the next best thing, for now until the whole world catches on. I wouldn't spend much time trying to "analyze" these companies. Just buy with lots of diversification (1% per stock) then monitor the stocks to ensure financials don't deteriorate (P-score stays at or above 8), which is another reason the AAII monthly screen works so well. I would sell upon a 50% gain like Schloss used to do. Once the whole world is using the Piotroski method then we can all go find something else that works since this will no longer work.
Mitochon premium member - 4 years ago

Sample size, independence, "expectation" (statistically) over long time frames are essential concepts here. Each stock is an event, and each event is independent. The F-score doesn't say anything about any one event, only the sample (N=n, where n=your basket of stocks) expectation over some holding period. For example, sampling a fair coin with n=3 as the trial size might yield H, T, T. This doesn't mean that the coin is biased in favor of "H", with 33.3% as a long run expectation. At n=50 trials, say, the apparent expected value would be much close to 50%. Treat the F-score as an indication of the statistical bias (in your favor) of sampling a basket of N=n stocks. The reason the score works is that it has some fixed bias in favor of the "sample", not the events, and it has a relatively large variance on the sample that may have its own bias...and the variance may not be fixed...it may change over time. The bias is not random, that's fixed (in so far as we have sampling data).  The only thing that appears to be predictable is the bias (again, in your favor) fo the long run expected returns of the sample (basket of stocks0.  The variance, however, may still be a problem; confidence intervals on the basket may be too high for many investors, and "tolerance" intervals for any one year's returns, as a single "event" may be too high for many, as well (see below).  Tolerance intervals show us the bounds for any one event (n=1)...they will be much higher than the bounds on the sample of N=n stocks. 

That said, it is this high variance, this high individual company (event) risk and high variance in annual returns, that drives people away from using the approach, and likely means that it has long term utility.   The Piotrosky stock portfolio margin of safety, or purchase discount, is due to high variance in the basket -- few institutions would be able to sell this to clients, so they don't try.  Few individuals have the temperment to see high drawdowns in their portfolio.  Arguably, there's no reason not to expect very bad back-to-back years. One may well see, in future, two years of, say, 50% losses, leaving a portfolio at 25%, or why not 3 years of 50% losses, leaving the return over 3 years at 12.5%?   So, if you're uncomfortable holding a sample that requires bad companies to ensure a long run return bias in your favor, and also has high variance that could have you selling at a nadir in response to a possible scenario like the above, try using it only as a portion of your investment approach, for, say, 20% of your investment.  This reduces the contribution to the variance on your total portfolio.  

Note: Trying to analyze any one stock event based upon other company information may or may not improve results.   Certainly picking only what appear to be the "safer" companies is likely to reduce returns, because which individual companies that work out is random, and in fact this compensates you for the ones that don't work out.  Picking perceived "safer" companies introduces a second unknown bias that combines with the Piotrosky bias, and while it may shrink the variance, it will alter the total return bias in unexpected ways, perhaps giving low variance, but ~ zero return bias, negative return bias, or just diminished positive return bias, say.  Still improving the bias toward positive returns and/or shriking the variance without altering the bias, using information external to the company, such as Guru holdings, or insider buying, can't be ruled out.   

Finally, Piotrosky portfolios represents a form of statisitcal "value" investing, as you derive value only from the expectation on the bias in aggregate, but can't do traditional value investing on the company based upon what is known publically, apparently.  This apprears not to be the case with some forms of "special situations" investing, such as "spin offs".  In the case of spin offs, there appears to be both a statistical bias in aggregate for the class of special situations, and also proven methods to remove some candidate spin offs from one's holdings to improve the overall return; a blend of traditional- and statistical value.  The additional research in traditional value may also improve the positive bias, and shrink the variance.  



Please leave your comment:

Performances of the stocks mentioned by Mark Lin

User Generated Screeners

HOLKLSUTest First Group Q1 Trending 2
jerkolberGraham 1
cspunarDividend Cover
doniemaherScreen #28 - Profitable, Lo De
HOLKLSUSmall & Mid Cap 2018 Rising Ra
FranktheTankKramer XGrowth
FranktheTankH Kramer
HOLKLSUTest First Trade Late Stage
Get WordPress Plugins for easy affiliate links on Stock Tickers and Guru Names | Earn affiliate commissions by embedding GuruFocus Charts
GuruFocus Affiliate Program: Earn up to $400 per referral. ( Learn More)

GF Chat