12. COMPUTERS AND COMPUTERS
You may think I am putting you on about Albert and his computer, and about Railroad Bill. The strange thing is, the Railroad Bill sort of thing is actually happening. Railroad Bill is, of course, not Railroad Bill at all; he is a computer. In order for you to see how Railroad Bill the Computer works, let us take a sidelong glance at the whole computer phenomenon.
When computers first came in, a lot of people on Wall Street and in investment management palaces signed up for them. Computers were new, scientific, and the wave of the future. The reason everybody signed up for a computer was that everybody else was signing up for a computer. Gustave Le Bon in the Electric Age. The first thing everybody did with their computers was to give the machine the straight-line back-room stuff to do, i.e., the payroll, figuring out margin accounts, and so on. This was on the noninvestment side.
The computers got bigger and faster and pretty soon the computer could do this stuff and still take the afternoon off for a ball game. So the Computer People were hired. The Computer People soon affected the whole scene. “Facts” went out; “bits” came in. A bit is one little zot of information and a computer can not only remember millions of bits, it can remember where each bit is, rearrange the bits in any specified order, add them, subtract them, divide them, play with ratios of them—and still take the afternoon off for the ball game. (All this spare computer time is expensive, which gives rise to “time-sharing,” in which the same computer can be used by different people for different purposes.) Security analysts used to walk around with a slide rule poking from one pocket. Now using a slide rule is like starting a fire with a flint, so the analysts have taken to using computer words like “input.”
The next step for the computers was called screening. Screening is just what it sounds like; you just saw Albert do a little bit of it. Having ingested all these bits, the computer could be called upon to arrange them in any order it was asked. When interrogative systems, like Albert’s, arrived, the analyst could simply sit down and say, “Computer, give me the fifty lowest price—earnings ratios of any stocks,” and the computer would print out, carefully ranked, the price—earnings ratios, or it would display them on its television-like display screen. Then the analyst could say, “Computer, of the stocks you have given me, which ten have the greatest return on invested capital?” And the computer could give them. Then the analyst could play with these ad infinitum. He could say, “Computer, adjust the current price—earnings ratios of these ten stocks to moving three-year averages of their earnings,” and the computer would do that.
All of the computer’s calculations were what good analysts were doing anyway, but no analyst could take the time to perform thousands of calculations. What the computer did was to screen millions of bits of information and create different patterns from them, something no analyst would have the time, physically, to do.
So today anybody can subscribe to a service that will give profiles of a stock over ten years, rank companies within industries by various characteristics, or rank the industries themselves against other industries.
Meanwhile the analysts and programmers have gone on to other adventures. One of these is projection, which is also just what it sounds like. A model of an industry is created, and then the profits of the industry are tried out against differing sets of circumstances. This is the adaptation of input-output analysis. The computer can even adjust bits for seasonal variations.
Finally, the analyst can just sit there playing with the toy, trying moving averages, exponential smoothing, and all kinds of different sets of series and ratios, just to see if any one of them will fit a hypothesis he has thought up. (This is why Wall Street bookstores can rack up big sales of books with titles like Smoothing, Forecasting and Prediction of Discrete Time Series.) The analyst may just try for a “feel” of the statistics, or he may try correlating various measures of price—earnings ratio, say, with the growth rates of sales and earnings and the deviations from these sales and earnings, isolating through this multiple-regression analysis—for that is what they call it—the variables that seem to affect the price—earnings ratio. Again, this is what analysts have always done, but they have done it by sight, feel, and slide rule, and hence only approximately and not for very many stocks. Not only can the computer print out all these things, but if it is really equipped it can change this linear information to graphics and draw you a chart.
(If you get a chance sometime, try playing with a computer equipped for graphics. You can take your light pencil (like a pocket flashlight), draw a circle in light on the screen, and the computer will correct it to a perfect circle. This is old stuff to aerospace engineers, but for amateurs it is the most fun since Mother put you in the attic with a tracing board that rainy afternoon.)
All of this kind of computer work deals with Fundamentals and factors related to Fundamentals: the business, sales, profits, profit margins, and so on. The real fun comes when the computer is used for Technical Analysis, i.e., what is everybody else doing? You saw a bit of this when Albert asked the computer to rank the momentum of stocks—to come up with the stocks that had just moved a greater percentage in price than all the other stocks.
I have a friend called Irwin the Professor at one of our nation’s leading universities who must rank as one of the top architects of computer-based technical analysis. Irwin is the professor of technology, mid-sixties style par excellence, which is to say that he is busy doing sixteen other things besides teaching the youth of the nation, and his university salary is only about a third of his earnings. Two thirds comes from consulting and various commercial ventures Irwin gets into, and of course he has all these bodies chained to the oars, the graduate students who are writing theses for him. I went by to see Irwin not so long ago. When you go by to see Irwin you do not have to come within three miles of the university. Irwin has a snazzy set of offices in a nearby office building, complete with Jens Risom furniture and a receptionist. This is where three of his companies are based. I can’t tell you the names of the companies, but they all have words in them like “computer,” “decision,” “application,” “technology,” and so on. Vice-presidents at big companies who are faced with some decision hire Irwin at a fat fee. Irwin and his computer models then tell the vice-president, after some complicated model building and computer work, that the new licorice toothpaste will not work because it is black, Americans do not want to have black teeth, and the people who do want black teeth are already chewing betel nuts and comprise only 4.6623 percent of the potential toothpaste market.
Irwin’s computer system was bankrolled by a couple of interested institutions and it is on line, real time, and all that. It is hooked up to the tickers of the New York and American stock exchanges, and it doesn’t even have to read the tapes optically; it picks up the original electrical impulse which drives the stock tickers and whisks it right into the memory. To Irwin’s computer, screening, projecting, and multiple-regression analysis are like Fish in the Ocean to a champion bridge player. I wanted to know how Irwin’s computer worked on the Technical side of the market.
“The first thing it does is to monitor every stock transaction, the price, the volume, and the percentage move,” Irwin said. “We have a Behavior Pattern for every stock. When a stock is behaving out of its pattern, the monitor flashes on. It says, ‘Hey, look at this.’”
(Like many computer people, Irwin tends to think of his computer as a large, faithful talking dog, and the objects to be scanned as sheep which are always getting out of line.)
Irwin pressed a couple of keys, but the large display screen on his table stayed dark.
“Nothing on the air at the moment,” Irwin said. “Let’s play for a while.”
We were happy playing some multiple-regression analysis game—I think it was whether if you bought every morning before eleven and sold every afternoon before two-thirty you could make money by outflanking the specialist—when Irwin’s computer flashed a monitor signal.
MONITOR
it said. We held our breath.
DIGITAL DATAWHACK
EXCEEDED LIMIT
38-½ 2:14 500
58/56 54/52 12/12/12 47/47 42/56
“Digital Datawhack is out of its pattern on the upside,” Irwin said. “This happened sixty seconds ago, at two-fourteen, on five hundred shares.”
“What’s all the rest of that stuff?” I wanted to know.
“Parameters,” Irwin said. “Don’t worry about it. Let’s see the trades in Digital Datawhack today.”
DIGITAL DATAWHACK, said Irwin’s computer.
200 |
10:12 |
36-¾ |
100 |
10:15 |
36-⅝ |
600 |
10:27 |
37 |
200 |
11:38 |
36-¾ |
500 |
1:51 |
37-¼ |
1100 |
1:59 |
37-¾ |
3000 |
2:05 |
38 |
1000 |
2:07 |
38-¼ |
500 |
2:14 |
38-½ |
“That’s every trade today in Datawhack and what time it happened and on what volume,” Irwin said.
“This sounds to me just like charting,” I said. “Upside breakout and all that.”
“Most of the patterns Chartists use are sheer myth,” Irwin said. “The principle of monitoring the stock movement is the same, except of course the computer can monitor thousands of stocks simultaneously and our patterns are statistically tested.”
Then Irwin sat bolt upright. The display screen was still flashing away cheerily about Digital Datawhack.
“Hey!” Irwin said. “There’s another computer on the air! There’s another computer on the air!”
As if sirens had gone off and a loudspeaker had barked “Battle Stations!” two graduate students rushed in from the next room. Irwin did not actually mean “on the air” like a broadcast, but discernible as if it were.
DIGITAL DATAWHACK
54/52 52/52/52 61/65 99/99/99
said Irwin’s computer.
One of the graduate students was busy toting in bulky computer print-outs from the library shelves.
“I bet it’s that IBM 360/50 in Minneapolis that bought those blocks of Boeing the other day,” said the graduate student.
“That’s unscientific speculation,” Irwin said. “Call up the big computer.”
“The big computer?” I said.
“Our computer can’t store everything,” Irwin said. “When it gets a problem it can’t handle, it calls up an IBM 7094 we share. We gave it an open phone line to the 7094. The 7094 has all the patterns there are.”
“You mean computers are buying and selling?” I asked.
“Most buying and selling is still done by individuals and institutions,” Irwin said, “just as it was in the old days.” (The old days to Irwin are 1962 or so, when computers were still doing only clerical chores.)
“But,” Irwin went on, “there are a couple of sophisticated funds that have computers like ours on the air. Then it really gets to be fun. Our computer scans the pattern of the other computer on the air, what its buying and selling programs seem to be. Once we get its pattern, we can have all kinds of fun. We can chase the stock away from it. Or even better, we can determine where the other computer wants to buy. Let’s say this computer is just getting its feet wet at thirty-eight and a half on Datawhack, but it will really open up and buy at forty-two. Maybe we can buy enough at forty and forty-one to bump the other computer into its major buy pattern at forty-two, and its buying will run Datawhack right up to forty-six. Then we’ve got a hell of a trade.”
“Just like Fourth and One, let’s go for it, on the charts,” I said.
“Same principle,” Irwin said, “only the game will be over by the time the Chartists get around to making their little marks. The Chartist has to sit down with his quill pen and let the ink dry and stare at it for a while, and besides, the Chartist’s pattern is probably wrong. It’s not scientific.”
“Is the other computer looking for your computer?” I asked.
“It might be,” Irwin said. “We haven’t really gotten into defensive strategy yet, because there aren’t enough computers on the air. But there will be soon. Unfortunately, not everybody believes in computers yet. Even our own subscribers drag their heels a lot of the time. They insist on following their own judgments, their intuitions, all their old ideas from the archaic pre-computer days. That means our computer doesn’t have a fair chance, because it’s not used consistently.”
“But your computer does buy and sell,” I said.
“Unfortunately, a person still has to give the order,” Irwin said, “because the stock exchange won’t accept orders from a computer, although the computer could certainly give the order directly. But the answer is, yes, our computer is running one portfolio.”
“How is it doing?” I asked.
“When we first put the computer on the air, we asked it what it wanted to buy and we couldn’t wait to see what it reached for. It said, ‘Treasury bills. Cash.’ We couldn’t get it to buy anything. So we checked out the program again, and while we were checking out the program, the market went down. Then we asked it again. The computer insisted on staying in cash. The market went down some more. We begged it to buy something. ‘There must be one stock somewhere that’s a buy,’ we said. You see, even computer people are victims of these old atavistic instincts from the pre-computer days. The computer just folded its arms. It wouldn’t buy anything. Then, just when we were worried that it never would buy anything, right at the bottom it stepped in and started buying. The market started going up, and the computer kept on buying. Pretty soon the computer was fully invested and the market was still going up.”
“What did it do then?” I asked.
“The market was still going up,” Irwin said, “and then one day it came and asked us for margin. It wanted to keep buying. So we gave it some margin. After the market went up some more it sold out a bit and came back to being fully invested. Right now it’s got buying power.”
“Irwin,” I said, “tell the truth now. If all these computers go on the air, as you say, does an individual investor have any chance?”
“There’s always luck,” Irwin said. “Luck—which is to say a serial run of random numbers—can happen any time. And the computer is out for really aggressive performance. An individual with a longer time horizon might be able to do passably. But on the whole, and over any given period of time, the computer will always win. It has to. The investing world will be divided feudally into fiefdoms, with the peasant investors all grouped around various computer castles.”
I was a little disturbed by some of the implications of the future as seen by Irwin the Professor, so I dropped by one of the investing institutions, not Irwin’s, which I knew had a computer-picked robot portfolio, and talked to an analyst I knew.
“All the robot portfolios do well at first,” the analyst said. “Then they begin to go a little bit off, as soon as people find out how they work.”
“Why do they go off? I would think they would get better.”
“Well, here we run a portfolio picked by the analysts against the robot portfolio, and of course the analysts get to use the computer for Fundamentals, screening and so on.”
“But if the computer uses screening too, why doesn’t it beat you?”
“Because the computer has to be updated all the time with fresh analysts’ opinions, new information, and so on.”
“So?”
“Well, we have twenty-one analysts working here, and the analysts supply the computer with fresh information. And every one of those twenty-one analysts knows that if that robot portfolio wins consistently, there will be twenty-one analysts out on the sidewalk.”
“You mean the analysts are subtly sabotaging the computer, just a little bit at a time?”
“You said it, not me. Sabotage is too strong a word. Just believe me, the robot is going to lose. Not by much. But when the final score is tallied, flesh and blood is going to beat that damn monster.”