<THE FOOLISH FOUR>

The Birth of RP
by Bob Price
(TMF Sandy)

Houston, TX. (Oct. 6, 1998) -- The RP system has had the best performance of the Dow selection systems we follow. You can see some tables showing its performance here. You might be interested in how I came across it.

Sometime early in 1996, I obtained some computer software that could search thousands of stocks for those matching certain criteria. I had learned about Beating the Dow here in the Motley Fool on America Online, and I wanted to generalize the Beating the Dow system to a larger universe. The problem was that this software couldn't do rankings. Because of this, I couldn't do something like Beating the S&P (which probably hadn't been invented yet).

One day, someone posted on the Fool's AOL Dow message board the idea of looking at the ratio of yield over price. While MF Dowman (as Robert Sheard was known then) didn't think much of this, it struck a spark with me. So I got out the Knowles and Petty Dividend Investor book and went through their historical tables of the Dow stocks. That approach did okay -- not great, but better than the Dow. I began using it in my software as part of the screens for my value picks. I called it the "Dividend Price Ratio," or DPR (to myself only, at this point). I really didn't expect to beat the Dow systems anyway; I just wanted something that would perform about as well as them, and then I figured I'd do better by having more stocks than just the Dow stocks to look at.

I got the Dow spreadsheet early in April and started playing with it. I thought I'd try to do the best I could with DPR. I set up a function that divided yield by price, but it didn't do too well. I knew that dividend ITSELF divided by price did fairly well, so I set up a formula on the totals page of the spreadsheet. This was lost in one of my disk drive upgrades, so I can't tell you exactly how it worked, but basically it took the dividend and divided it by price, taken to a factor that I could control on the last page. When that factor was 1, the ranking was dividend divided by price, or yield. That gave the same results as (for example) the Beating the Dow 10 using 10 stock picks. When the factor was 2, so that the dividend was divided by price squared, that was the yield divided by price formula I had tried before.

The next step seemed fairly obvious at this point. I simply varied the factor continuously and looked at how the stocks chosen this way performed. There was a large "flat spot" in the performance around a price factor of 1.5. That is, varying from 1.47 to 1.53, say, didn't make much difference in the performance. If you've played with graphs, you know that when you are near a smooth top or bottom, you get this flattening out, so 1.5 appeared to be very near a maximum. And, to my shock, this strategy outperformed the Dow methods I knew about by a wide margin. I had only hoped to nearly equal them -- but I had surpassed them. The only other refinement was to duplicate an idea from the old Fool Four. It turned out that if you dropped the highest ranked stock by the formula, the method did better. That was the final refinement I made.

What did the formula mean?

Well, dividend / (price ^ 1.5) is the same as yield / (square root of price).

"Square root of price" rang a bell. I had read in Norm Fosback's Stock Market Logic (mostly an advertisement for his newsletter, but it does have certain interesting things in it) that the square root of price was a better predictor of future beta than even past beta. In English, 1 / (square root of price) is an excellent predictor of future volatility.

What the formula seemed to do was pick the high yielding stocks that were likely to have future volatility -- weighing the volatility with the yield. If you have stocks that are likely to go up, volatility is a good thing. Since it was a ratio, it didn't have the unpleasant "non-linearity" that the regular Dow methods do. For instance, if a stock ranked #10 in yield is low in price, the Fool Four may pick it. However, if the price goes up a tiny bit, it won't move a position or two in the Fool Four -- but it might totally fall off the high yield list. Since this formula ranks the stocks smoothly, such "non-linearity" is less likely to happen.

At this point, I made what in retrospect was an error. I promised a friend of the family that I wouldn't "give it away" but would try to get paid for it. So I didn't just post it on the AOL boards. I decided to try to get the Fools interested in it.

It had been a simple thing to discover. I was very concerned that someone else would think of it. I thought the meaning of "DPR" was too easy to guess. So after a bit of thought, I came up with "Ratio Procedure." "RP" was not going to help anyone guess how it worked, I thought -- especially since they would probably think it was my initials.

In April of 1996, I wrote an e-mail offering RP to the Fools, citing how it had performed but not how it was done. I even mentioned in that email that "RP" are my initials (although, in fact, I don't use them as my initials), which may be why Robert Sheard apparently still thinks that's the source of the name.

Robert Sheard asked that I share the method with him, promising to keep it secret, so that he could independently verify that it worked. He did, and in fact he was the first to point out that the Sharpe Ratio of RP4 (as we were calling it) was also much better than anything else known at the time. Another contribution he made was that he thought that the square root would cause confusion. Since the formula was being used to rank the stocks only, its square worked just as well. That's the way you usually see it given these days -- as yield squared / price.

Over the next two years, for one reason or another, RP was not published. Robert invented UV, which comes closer in performance to RP, so that removed a bit of the pressure to bring out RP.

Then, what I had expected to happen much sooner finally happened. Elan Caspi rediscovered the system and posted it on the Web message boards. He called it ER, for Elan's ratio (see his first post here).

Robert Sheard was told it was out on the web. He told me about it, and I then published the info on the AOL boards and added a follow-up to Elan's excellent article on the Web.

Current Dow Order | 1998 Dow Returns

What Happened to Robert Sheard?

10/06/98 Close

Stock Change Last -------------------- UK - 5/8 41.69 IP - 1/4 45.56 MO +1 3/4 49.25 EK +1 5/16 76.19

Day Month Year FOOL-4 +0.91% -0.14% 10.68% DJIA +0.22% -1.27% -2.09% S&P 500 -0.05% -2.85% 1.81% NASDAQ -1.68% -10.80% -3.79% Rec'd # Security In At Now Change 12/31/97 206 Eastman Ko 60.56 76.19 25.80% 12/31/97 276 Philip Mor 45.25 49.25 8.84% 12/31/97 289 Int'l Pape 43.13 45.56 5.65% 12/31/97 291 Union Carb 42.94 41.69 -2.91% Rec'd # Security In At Value Change 12/31/97 206 Eastman Ko 12475.88 15694.63 $3218.75 12/31/97 276 Philip Mor 12489.00 13593.00 $1104.00 12/31/97 289 Int'l Pape 12463.13 13167.56 $704.44 12/31/97 291 Union Carb 12494.81 12131.06 -$363.75 CASH $754.73 TOTAL $55340.98