In Part II we did historical testing of an efficiency signal on today’s S&P 500 data going back to the year 1980. Basically we bought highly efficient stocks on the first of the month and held them with a 25% trailing stop. Once the portfolio had 25 stocks (i.e., with 1% risk we were fully invested), it only bought more stocks when we were stopped out of a loser. The net result was a compounded annual ROI of 37%. It took 674 trades and rejected (i.e., we were fully invested) 701. 56.7% of our trades made money and the average win was 3.87 times bigger than the average loss. We also spent 378 days in a winning trade versus 80 days in a losing trade. And we spent nine million four hundred dollars in trading costs, which amounted to 1% going in and 1% going out, so you can’t say that low cost influenced our results.
In Part III, we took a look at some of the bugs in our coding. These included a change in the way the position sizing was calculated using the total cash variable in Mechanica. This resulted in a decrease in our returns of about 5%. The second change was to our smoothing function (which tended to favor low-priced stocks). This made a significant impact on our results because 1) the days in the trade were increased and 2) totally different stocks were purchased.
At this point, I’m not sure that our smoothing algorithm is giving us stocks that I would normally buy using my discretionary methods of looking at charts. As a result, there is still more research to do. First, I need to look at the charts of some of the stocks bought to see how the algorithm is doing. I haven’t had the time to do that yet. As a result, in this article I’m going to look at the effect of buying and holding the stock in our two databases:
1) today’s S&P 500 going back to 1980, and
2) the actual S&P 500, including all of those dropped from this list, going back to 1990.
Buying and Holding the April 2005 S&P 500 (from 1980 or on the date when they first came out as stocks).
In our first study, we simply bought $200 worth of today’s S&P 500 on October 3, 1980 or whenever they came out as stocks. Thus, we were still purchasing $100,000 worth of stock, but once we bought we didn’t sell unless 1) the stock stopped trading or 2) the database ended on April 22, 2005. There were no exits except those two. Thus, this is a real buy and hold situation. However, we are basically buying the BEST American companies. We were also buying them either on the start date in 1980 or when they first came out as stocks. That’s right, our initial entry (e.g., for a stock like DELL or MSFT) was not when it became part of the S&P 500, but when it first came out as a stock. In addition, we are buying and holding them through the longest bull market of the 20th century and into the secular bear market starting in 2000. Although we probably would have trouble finding the best stocks in the U.S. 25 years from now, its gives us some idea of the absolute best performance that we could expect from a buy-and-hold philosophy under ideal conditions.
Table 1 shows a listing of the years and the number of stocks purchased during that year. Since the last year is 2003, we were actually not fully invested until 2003.
| Table 1: Stocks Added Purchased by Year |
| Year |
Number Stocks |
Number Losers |
Total Stocks |
% Invested |
| 1980 |
249 |
6 |
249 |
49.90% |
| 1981 |
8 |
0 |
257 |
51.50% |
| 1982 |
9 |
0 |
266 |
53.31% |
| 1983 |
19 |
0 |
285 |
57.11% |
| 1984 |
23 |
1 |
308 |
61.72% |
| 1985 |
9 |
0 |
317 |
63.53% |
| 1986 |
18 |
0 |
335 |
67.13% |
| 1987 |
9 |
0 |
344 |
68.94% |
| 1988 |
15 |
0 |
359 |
71.94% |
| 1989 |
10 |
0 |
369 |
73.95% |
| 1990 |
12 |
0 |
381 |
76.35% |
| 1991 |
16 |
1 |
397 |
79.56% |
| 1992 |
12 |
0 |
409 |
81.96% |
| 1993 |
19 |
2 |
428 |
85.77% |
| 1994 |
8 |
0 |
436 |
87.37% |
| 1995 |
11 |
0 |
447 |
89.58% |
| 1996 |
13 |
3 |
460 |
92.18% |
| 1997 |
6 |
3 |
466 |
93.39% |
| 1998 |
6 |
1 |
472 |
94.59% |
| 1999 |
9 |
4 |
481 |
96.39% |
| 2000 |
8 |
3 |
489 |
98.00% |
| 2001 |
4 |
0 |
493 |
98.80% |
| 2002 |
3 |
2 |
496 |
99.40% |
| 2003 |
3 |
0 |
499 |
100.00% |
Only 499 stocks are listed, and I’m not sure what the missing stock is or why, but notice that we only lost money in 26 stocks out of 499. Also notice that we were only 49.9% invested in 1980, 76.56% invested in 1990, 89.58% by 1995, and 98% invested by the end of 2000.
Let’s look at the overall statistics of our little experiment. We started with $100,000 and ended up with $ 3,025,960. Our gain amounts to a compounded return of 14.89%. We made money on 94.78% of our trades and the average gain was 71.35 times the average loss. Sounds like an ideal system doesn’t it and perhaps a strong statement for buy and hold, but remember that we were buying stocks like MSFT and DELL when the were first issued simply because they later became part of the S&P 500.
However, there was also bad news because we had a maximum drawdown of 51.27%, which occurred on October 9, 2002. And we were in a drawdown from March 27, 2000 until the end of the run on April 4, 2005. I’m not sure whether we’d even be out of the drawdown in 2007.
Next I decided to call 1R the full investment amount of $200. This allowed me to calculate R-multiples for the trades and also the expectancy. Using this calculation, the mean R-multiple (expectancy) was 29.31R, the standard deviation was 99.96R, and the ratio between the two was 0.2936.
Our compounded ROI with the efficiency algorithm was 28.59% with the “close minus close” smoothing algorithm on the same database. And it was 14.58% with the “close divided by close” smoothing algorithm. And I’m not convinced that either of these came close to what I trade with a discretionary judgment of what is an efficient stock. I didn’t get any volunteers to search out trades for me and I have not yet had time to look over the stock charts myself.
You might be interested to know what the best and worst stocks were in the database. For example, what stocks in the 2005 S&P 500 database have actually lost money? And what American stocks have been the best since their inception? Both of those questions can be answered with this study.
Table 2 shows the big losers in our study. Who would have thought that if you had bought one share of AT&T in 1980, that you’d lose money over the next 25 years? Who could predict the government breakup of AT&T in 1983. In addition, old AT&T holders got shares of all of the companies that AT&T broke into and that’s probably not accounted for in this database. And there may be other such examples in the data that are not so obvious – again, more data problems. 2
Incidentally, the new AT&T is what used to be SBC Communications (or Southwestern Bell). It’s not the same as the AT&T that was around in 1980, but it has reacquired much of the old AT&T. And Lucent, which used to be Bell Labs, the research arm of AT&T, had more patents than any other company. It was a powerhouse of invention. But when it separated from AT&T and went out on its own (in 1996) it failed miserably after a nice start.
| Table 2: Top Losing Stocks In Our Database |
| Symbol |
Name |
Entry Date |
Exit Date |
$ P/L |
Shares |
R-multiple |
| T |
AT&T |
10/ 03 1980 |
04/ 20 2005 |
($146) |
1 |
-0.73 |
| UIS |
Uinsys |
10/ 03 1980 |
04/ 20 2005 |
($131) |
9 |
-0.66 |
| DAL |
Delta Airlines |
10/ 03 1980 |
04/ 20 2005 |
($130) |
17 |
-0.65 |
| DYN |
Dynergy |
11 12 1993 |
04/ 20 2005 |
($129) |
22 |
-0.65 |
| LU |
Lucent |
04/09 1996 |
04/ 20 2005 |
($133) |
25 |
-0.67 |
| CIEN |
Ciena |
02/11 1997 |
04/ 20 2005 |
($170) |
11 |
-0.85 |
| DPH |
Dephi |
02/09 1999 |
04/ 20 2005 |
($150) |
10 |
-0.75 |
| JNS |
Janus Capital |
06/28 2000 |
04/ 20 2005 |
($129) |
5 |
-0.65 |
| AV |
Avaya |
09/20 2000 |
04/ 20 2005 |
($120) |
10 |
-0.6 |
| EP |
El Passo Corp |
01/09 2002 |
04/ 20 2005 |
($140) |
4 |
-0.7 |
Table 3 shows the big winners in our study. The big winner, of course, is DELL computer. If you had purchased $200 worth of Dell in June 1988 when it came out, today you would have an investment of $356,099. That’s an R-multiple of 1780R. That’s about five times bigger than the 9th largest winner Microsoft, which only became a 240R winner.
| Table 3: The Biggest Gainers in the 2005 S&P 500 |
| Symbol |
Name |
Entry Date |
Exit Date |
$ P/L |
Shares |
R-multiple |
| CSCO |
Cisco |
02/21 1990 |
04/20 2005 |
$42,343 |
2500 |
211.72 |
| MSFT |
Microsoft |
03/17 1986 |
04/20 2005 |
$47,952 |
2000 |
239.76 |
| PGR |
Progressive Group |
10/03 1980 |
04/20 2005 |
$50,674 |
571 |
253.37 |
| BMET |
Biomet* |
12/22 1982 |
04/20 2005 |
$52,558 |
1428 |
262.79 |
| UNH |
Unitedhealth Group |
10/19 1984 |
04/20 2005 |
$60,556 |
666 |
302.78 |
| BEN |
Franklin Resources |
09/27 1983 |
04/20 2005 |
$66,827 |
1052 |
334.14 |
| GPS |
GAP, Inc. |
10/03 1980 |
04/20 2005 |
$68,299 |
3333 |
341.5 |
| MYL |
Mylan Labs Inc |
10/03 1980 |
04/20 2005 |
$81,572 |
5000 |
407.86 |
| HD |
Home Depot |
09/24 1981 |
04/20 2005 |
$176,018 |
5000 |
880.09 |
| DELL |
Dell Computer |
06/24 1988 |
04/20 2005 |
$356,099 |
10000 |
1780.5 |
You can tell what happened to the price of the various stocks by looking at the number of shares purchased when we initially bought the stocks. For $200 we were only able to buy one share of ATT back in 1980. But for the same $200 (split and divided adjusted, of course), we were able to buy 10,000 shares of DELL.
In Part V of this research, we’ll look at our accurate S&P 500 database and see what happened with buy and hold. In this case, we’re only buying DELL when it becomes part of the S&P 500 and that should make a big difference.
Other research we’ll do in this series will include the following:
1. Determining what happens when we allow ourselves to take as many as 250 trades (i.e., half the S&P 500 database) at any one time with the two smoothing functions. With 1% risk and a 25% trailing stop we are limited to 25 trades. With a 0.1% risk and a 25% trailing stop, we are will be limited to 250 trades. We’ll simply increase our starting equity to $1M so that we’ll be investing the same amount ($4000) with each trade.
2. Research to convince me that I’m really buying the stocks I’d normally buy when looking at a chart will follow that. One way would be to look at charts of the 100 trades from both smoothing algorithms to determine how many of them look like “efficient” stocks. This will allow us to determine if we are looking at efficient stocks or not. If any of you would like to do that and save me some time, I’d appreciate it. Please let us know and we’ll send you the data. And if there are a number of you, we’ll simply split them up.
3. We’ll also try both 1) the 180 day channel breakout and 2) the linear regression to pick our trades.
4. And, lastly, when I feel I have some of the answers I’m looking for, we’ll then move to the real S&P 500 database that we have.
Notice at this point I still have not yet 1) looked at the effect of any trend following algorithm and compared it with efficiency, 2) looked at the data on a S&P 500 database that added and subtracted stocks as the index did, or 3) made position sizing adjustments to see what’s really possible with this sort of trading. All of that is still to come in subsequent articles and it looks like this series might continue for some time.
Until next week, this is Van Tharp.
_____________________________________________
1 By the way, if you have some interest in Mechanica, which we are using in these tests, then visit the Mechanica web site -- http://www.mechanicasoftware.com. Mechanica is the new Windows version of Trading Recipes.
2 We checked this with our more accurate database from Bloomberg. It shows AT&T with share price was $3.3656 (adjusted for splits/dividends) at our start date in 1980. The last trade was $20.35 in 2005, at which point it presumably merged with SBC and disappeared. Again, this points out the huge problem you have with any database you might have – accuracy. In this example, we have a winner (which goes from 3.3656 to 20.35) that turns into an 73% loss just because of a data problem. And the only way I know that is because it turned out to be the biggest loser and we decided to check that out. How many data problems are there? And how can you trust any historical study when such data problems exits?
