The Best Piostroski Screen Combination

February 14, 2012 | Comments (15)

I have gone through in detail the Piotroski criteria and backtested the results to see whether it really did achieve what the academic papers claim. My conclusion? It has resulted in outstanding performance.
While all my other value strategies did poorly in 2011, the Piotroski 2011 performance was very respectable at -1.3%.


Written by

Jae Jun

follow me on

Facebook

Twitter

Piotroski Screener & Piotroski Spreadsheet

I currently have a Piotroski screener which I update weekly or so as well as a free Piotroski spreadsheet that makes use of the excellent SMF excel plugin.

I have gone through in detail the Piotroski criteria and backtested the results to see whether it really did achieve what the academic papers claim. My conclusion? It has resulted in outstanding performance.

While all my other value strategies did poorly in 2011, the Piotroski 2011 performance was very respectable at -1.3%.

Best Piotroski Screen Combination Test

The Piotroski score is a great mechanism for filtering stocks, but I want to see whether better results can be achieved. So I am going to share with you my test results of numerous combinations utilizing the 9 point scoring system derived by Piotroski. My goal is to try and nail down the best combination that will outperform the standard Piotroski screen.

I did something similar with Graham’s stock selection checklist. Graham originally had a 10 item checklist which I felt were not all necessary. By going through  the list and eliminating criteria such as a stock having to be priced at 2/3 of NCAV, the modified screen turned out to be much better.

The Initial Test Filter

With 9 criteria, there are hundreds of possible combinations, but there is no way of testing each one.

In order to cut the time and labor, I back tested one criteria at a time against the past 5 years – from 2007 to 2011.

Then I sorted the criteria based on the return it made. If you look at the image below, criteria #6, the latest current ratio being greater than the prior year current ratio, performed the best.

But using a single criteria as a screen itself is too broad and results in far too many stocks in the final output.

The Best Piotroski Criteria Combination

Using the findings from above, I tried combinations such as the top three, top five, positive only, bottom three etc, but results were not what I expected. Thus, the process of elimination began.

Rather than wasting your time explaining every thought process, I’ll just dumb it down.

I started with the best two criteria as the foundation and added one criteria at a time. If the new combination beat the previous combination, I continued with the new combination. If not, then the criteria was eliminated and so on.

The combinations were back tested over a one year, 3 year, 5 year and 10 year period and then ranked. The lower the total rank, the better.

As you can see, the best combinations (#6,9,1) and (#6,7,1,2) both tied for first place. However, I am awarding combination (#6,9,1) first place as it protected the downside better in 2011.

The original Piostroki screen placed 10th place and is the only one that posted positive numbers across all time periods. Impressive indeed.

In the end though, the best Piotroski combination is a very simple selection criteria which doesn’t resemble Piotroski at all.

  • Current ratio should increase
  • Decrease in shares outstanding compared to prior year
  • Positive net income

The results adhere to the rule of keeping it simple.

20 Stocks for the Best Piotroski Screener Combination

(stock price is from Feb 12, 2011)

Disclosure

None

About Jae Jun


Jae Jun is the founder of Old School Value. He is on a mission to provide practical and actionable value investing tools, tutorials and educational material to help empower the individual investor. Keep in touch with Jae via any of the methods linked below.

Get Tips and Strategies to Achieve Higher Returns
Bonus: Get 9 FREE Investing Spreadsheet Calculators
  • Matt

    I’ve enjoyed reading your company analysis posts but this article is excellent and very informative. Thank-you!

  • Mike

    Jae,
    Great job testing these various Piotroski components…very informative! One question though..In Piotroski’s research, did he not start the entire process with a basket of stocks that are in the bottom decile (or some similar metric) of P/B values…then use the “Piotroski” criteria to find the best stocks out of this group?? It may be interesting to add such a criteria to this screen (ie. P/B<0.8) or even a similar tangible book value criteria. Thanks again for the great work Jae!!
    Mike

  • http://www.oldschoolvalue.com Jae Jun

    @ Matt,
    Thanks

    @ Mike,
    Oh right, I forgot about that. Will have to try some things again on the side.

  • Jim

    Very informative information Jae. I’d like to see an article written by you concerning your techniques in back-testing. What devices do you use to accomplish the task and such.

    ValueInvestorToday

  • http://ratioratingranking.blogspot.com Patrick

    Remarkable study to separate the winners from the loser’s criteria.
    I things the 4th criteria is quite interesting, haven’t you test combination such as 6,9,4 or 6,9,4,1.
    A different weighting of each criteria should also improved the performance.

  • somrh

    In the original paper (http://www.chicagobooth.edu/faculty/selectedpapers/sp84.pdf) he looked at how the variables (including market returns) correlated with each other. The order (in magnitude of correlation with market returns) is as follows: (see Table 2… pg 16 of pdf)

    1, 2, 5, 4, 7, 8, 3, 6, 9

    6 and 9 are at the bottom of the list here. So I wonder how much of this is a product of the particular 4 year period you looked at compared to the different period Piotroski looked at.

    Part of it will also be due to the fact that the above Piotroski info focused on high book to market stocks whereas I’m assuming you’re looking at a much broader universe.

    In any event, interesting stuff. (I also would be interested in VIT’s question as well.)

  • Andreas

    Hi Jae!

    Great post! Why don’t you use Discriminant Analysis or Neural Network to find the best combination of those variables. Relatively easy if you use a statistics program.
    Cheers,
    Andreas

  • http://www.oldschoolvalue.com Jae Jun

    @ Jim,
    Going to have to disappoint you in how I go about doing this but I think there is a commercially viable solution. I’m going to test it out and if it is good, I’ll do a review on it.

    @ Patrick,
    I’m sure I did try that combination as I tried the remaining 7 possible 3 criteria combination starting with 6,9.

    @ somrh,
    Interesting feedback. I wonder whether the results are so much better because of the fact that is is less correlated.

    @ Andreas,
    I forgot all my math from 7 years ago lol. I had to look up the formulas for permutations, let alone try to remember how to go through statistics again.

  • Andreas

    if you need help just send me a mail with the data…the good is that with Discriminant Analysis that I have used, you get also different weights for the different variables.

  • http://viewonvalue.blogspot.com Konrad

    Hi Jae,

    I love your breakdown, pretty informative. One thing perplexed me though, why is there a difference between 6,9,7 and 6,7,9 in the results? Aren’t they the exact same thing? Or there’s a real practical difference in implications of searching for companies meeting criteria 7 in the universe of those meeting 9 versus the other way around.

    Great work, this sounded like a big job

  • http://www.oldschoolvalue.com Jae Jun

    @ Andreas,
    Send me some numbers and I’ll see what I can do. Maybe a followup would be good.

    @ Konrad,
    Good catch I didn’t notice. They should definitely both be the same though.

  • valueinvestortoday

    Jae, that’s fair enough. I do back testing but it’s very tedious. Was just wondering if you had an easier way. Take care.

  • Lars Bech

    Very interesting article – I really admire your work. I liked the original piece by Mr. Piotroski however not the part where he shorts the lowest scoring stocks found. That’s not possible for me to do. I am looking into your findings and combining them with an indicator to get me out of the market if it is in a downtrend – something like Mebane Fabers work. That would have gotten you out of the market already 1½ year before the financial crisis. That strategy has been proven to return about 100 per cent in 3 years. Cheers and keep up the very good work.

  • Robin H

    Hi Jae, it’s really awesome what you are doing on your site. I am still at awe. I always thought there was a way to backtest fundamental strategies, but you’re not only backtesting it but optimizing it.

    May I know what program you are using to perform these backtests and optimization? I would like to point out also that the criteria for Shares Outstanding in Piotroski’s F-Score is awarded not only when there is a reduction, it is also valid if the number of shares stays the same with the prior year. Or at least that’s what I got from the other articles I have read on it.

  • Michael

    Hi, can you tell me what your holding period, rebalancing, and how many holdings you applied in your backtesting? Thanks.

x
Receive Email Tips and Strategies on Achieving Higher Stock Returns

Bonus: FREE Set of 9 Investing Spreadsheets