Discussion of Karel's MrMarket approximation

[Jaws57]

Karel

Appreciate all the work you have done. I followed your MMM picks on yahoo. I downloaded the excel files to my computer to my Microsoft spreadsheet except the formulas wont convert (I dont have Excel). This is not a major problem because the screens you use I saved on the stock screener. Those work fine. Basically I have two questions:

1. How do you weight the 3 columns (12mo sales,hist PE, fwd, PE)?
I used multipliers to equal them out.

2. R-squared ? This one I cant figure. I assume it is based on a tight line from start point to end. Is the formula easy to set up?

Thanks Jaws57

[Karel]

Hi Jaws, it would really be much easier to get Excel or a compatible spreadsheet program. In MrMarket's description of his method, he says he uses several columns for ranking purposes, so the program has to support the ranking function.

I know for certain that StarOffice 5.2 supports this function and opens the spreadsheet without problems. You'll find a free (and newer) version on http://download.openoffice.org/1.1.0/
This a 60 MB download. If you don't have broadband, ask someone to download it for you andburn a CD, or find a CD distributor (see CD link on the page mentioned)

An alternative could be 602PC Suite http://www.software602.com/products/pcs/trial.html
The free trial is unlimited, and the 'limited' spreadsheet might very well work. This is a 20 MB download. I have no experience with this program.

Regards,

Karel

[Jaws57]

Downloads

Karel,

You are HUGE!!!. I went for the 20mb download 602PC Suite. I get it free for 30days and then I can register free. I works great. The r2 works too except I had to trnasfer to my old worksheet before I got to the mimic one. I ran it tonight and got CCBI ECPG PMTI CMTL GGB CMT TSO DHB GI FPIC. This was the rank before r-squared. I havent had time yet. I did test it on WHC and it calculated .949.

Thanks,
Jaws57

[Karel]

Hi Jaws, you're on your way! I just ran the screens and got

CCBI ECPG PMTI CMTL TASR GGB CMT TSO GI FPIC

as the last ten. The difference is immaterial; Moneycentral probably has done some update or other. I also get .949 as r^2 for WHC (a very nice value BTW).

When you do your DD, it may be good to now that MrMarket likes positive earnings, and solid earnings and revenue growth. Good luck!

Regards,

Karel

[scottieahmed]

testing so far

Hey Carl,

I emailed you in December to talk about your model. I have started testing it now. So far I have three picks going:

FPIC
PMTI
ECPG

It looks like waiting a week between each pick might not be long enough to get seperation and/or new stocks with the highest R-squared value. Anyway, i just thought that I would let you know that i am enjoying this exercise. I am tracking each stock and looking for 13-15 percent growth as a target. I will keep you posted on the testing, if you are interested. Have you run your screen at all this year? Do my stocks look like familiar finalists to you?

Good luck with your investing@

[scottieahmed]

Question on your R-Squared formula

I ran the Rsquared formula from the mimic spread sheet on HAR, the MLK winner. While Ernie got a value around .91 or .92, i got 0.504894846. I got a years worth of historical prices from yahoo, and then pasted them into column B of the R-Squared section three times and got the same value. Am i doing something wrong? I scrolled through the data on yahoo and noticed that a couple of times a dividend was paid and that there was a 2:1 stock split on Dec 5, 2003. Was i suppossed to do something different with the data if there was a stock split? Anyway, thanks for any help that you might have.

Scottie

[mrmarket]

Re: Question on your R-Squared formula

I ran the Rsquared formula from the mimic spread sheet on HAR, the MLK winner. While Ernie got a value around .91 or .92, i got 0.504894846. I got a years worth of historical prices from yahoo, and then pasted them into column B of the R-Squared section three times and got the same value. Am i doing something wrong? I scrolled through the data on yahoo and noticed that a couple of times a dividend was paid and that there was a 2:1 stock split on Dec 5, 2003. Was i suppossed to do something different with the data if there was a stock split? Anyway, thanks for any help that you might have.

Scottie

Scottie,

Use the "Adjusted Quote" column that Yahoo has...when the stock splits, you have to make sure the historical prices are adjusted, otherwise the download will thing that the stock suddenly dropped 50% in price, which would be very unhealthy for its r^2 value.

[scottieahmed]

Thanks for the Help, Mr. Market

The values match now. I am having a lot of fun testing this and learning from your site. It is a great place to read what people have to say, and then go out and learn on your own. I am very impressed with your model and your results. I hope to start investing in about 6 months, when I can save up the money to open an account. Hopefully, I will also be able to pay for an IBD subscription, so that I can run your 1st, 2nd, and 8th screens as well, as I want to be able to test both. It is fun to see which stocks your screens turn up vs the "approximation" screens. I noticed that one of your top five was not in my dump, and that none of my final 10, ranked by highest R-squared value, were on your top five.

Anyway, this has been very educational and a lot of fun. I truly appreciate your site.

All the best! Keep rocking your portfolio!

Scottie

[hawley]

Karel spreadsheet vs MM magic

Hi Jaws, you're on your way! I just ran the screens and got

CCBI ECPG PMTI CMTL TASR GGB CMT TSO GI FPIC
[...stuff deleted...]

Karel

I run your spreadsheet model about once every 3 months or so, get frustrated on one particular issue, and file it away until next time. Now is the time to address what I have always thought was something I was doing wrong.

Last night, I ran the screens and used the spreadsheet and basically got the same list of 10 you got a few days ago. I recognize that some of the screens are MM approximations from lack of IBD data on money central.

The part I don't understand is why none of our top 10 or the MM top 5 from a few days ago seem to match. His 5 are on our spreadsheet with a much lower rank. Ours are on his data dump but must obviously be lower ranked since they don't make his top 5.

If we are all doing the ranking based on the following criteria: Price Appreciation, Price Appreciation divided by trailing 12 month P/E, Price Appreciation divided by forward 12 month P/E... shouldn't they be more closely matched if not exactly the same?

The approximations on a few screens seems to be the only difference. So I can only see how this affects our "universe of about 200 stocks" but I can't figure out the discrepancy with the rankings.

[mrmarket]

once I find the high r^2 stocks, I throw out the ones with fuindamentals that I'm not crazy about.

[scottieahmed]

Question for Karel

Karel,

When you sort on column K, to find the stocks that you will get R-squared values for, what is the header for that column? Is it xxx4 or !!!!!!! ?


Thanks, Karel

Scottie

[hawley]

once I find the high r^2 stocks, I throw out the ones with fuindamentals that I'm not crazy about.

E,
Was this intended to be a reponse to my post regarding the discrepencies between Top 5 posts and Karel's spreadsheet model?

If so, this does not make a lot of sense to me unless you have some automated way to grab a bunch of R^2 values. Some of your last top 5's ranked in 30s and many much lower in the spreadsheet. You can't tell me that you calculate dozens of R^2s to get down to these lower ranking picks. Also, I'm pretty sure that I have seen you use different lengths of time for R^2 calcs (i.e. not always 1 year).

I think your picks are bubbling up higher in your process than they are in Karel's spreadsheet. I was trying to point this out and see if we can determine what might be missing so the spreadsheet would become a bit more true to your model.

[mrmarket]

once I find the high r^2 stocks, I throw out the ones with fuindamentals that I'm not crazy about.

E,
Was this intended to be a reponse to my post regarding the discrepencies between Top 5 posts and Karel's spreadsheet model?

If so, this does not make a lot of sense to me unless you have some automated way to grab a bunch of R^2 values. Some of your last top 5's ranked in 30s and many much lower in the spreadsheet. You can't tell me that you calculate dozens of R^2s to get down to these lower ranking picks. Also, I'm pretty sure that I have seen you use different lengths of time for R^2 calcs (i.e. not always 1 year).

I think your picks are bubbling up higher in your process than they are in Karel's spreadsheet. I was trying to point this out and see if we can determine what might be missing so the spreadsheet would become a bit more true to your model.

Well...I guess it's not an exact science...I'm sure the stocks you are generating will do fine. Have you backtested your picks?

[Karel]

Re: Question for Karel

Karel,

When you sort on column K, to find the stocks that you will get R-squared values for, what is the header for that column? Is it xxx4 or !!!!!!!

It is xxx4. The !!!!!! column shows the rankings, but should never be used for sorting.

Regards,

Karel

[Karel]

Re: Karel spreadsheet vs MM magic

... If we are all doing the ranking based on the following criteria: Price Appreciation, Price Appreciation divided by trailing 12 month P/E, Price Appreciation divided by forward 12 month P/E... shouldn't they be more closely matched if not exactly the same?

The approximations on a few screens seems to be the only difference. So I can only see how this affects our "universe of about 200 stocks" but I can't figure out the discrepancy with the rankings.

Well, the fact that the universe is slightly (or more than slightly) different may pick up other stocks that bubble higher. Currently there are two big differences in the universe selection filters: the use of approximations, but also the use of same-quarter growth figures. I picked that option from the screener, because I didn't know how often the yearly figures get updated. If those get updated every quarter, it would be best to use that, but I have been too lazy to find out.

Regards,

Karel