Do people ever let computer programs figure out what to buy and what to sell on the stock market?
If not, have experiments been done testing the idea?
(I was idly speculating today about the viability of software that constantly scans the news for corporation names and evaluative keywords and buys and sells stock based on that. Of course, I’m not just asking about programs that scan news, that was just the crazy notion I was speculating about today.)
Guess that answers my question then.
On average, the best you can do with any intelligible stock-picking method is to match the indexes. Now, it’s conceivable that some humans can use methods which aren’t intelligible (though you’d never be able to know which humans those are, so it wouldn’t do much good), but any method programmed into a computer must necessarily be intelligible.
What do you mean by “intelligible” here?
According to Nate Silver, investing on the day after a fall is slightly to your advantage.
So after the Dow falls, for example, you’d buy up its exchange-traded fund, the DIA. I’m no expert, but I assume one would want to put a lot more work into their portfolio. And since the disparity seems to be dependent on how information moves, it probably won’t stay this way forever.
By “intelligible”, I mean that it’s possible to explain to someone else how to use the method. “I pick letters out of a scrabble bag and buy the stock whose ticker symbol is spelled out” is an intelligible (though probably stupid) method. “I buy the stocks that the voices in my head tell me to” is not.
And I should correct what I said before: It is possible to tell that a human is using an unintelligible method (like, say, the voices-in-the-head method), but if that unintelligible method is any better, on average, than following the indices, it’s impossible to know that.
Got it, I see what you’re saying.
I think an implication of what you’re saying is that there can be no such thing as having skill at stock trading. (Being divinely inspired about it is not being skillful at it. BTW shades of middle Plato here! If that made no sense, ignore the comment.) In other words, you can no more “skillfully” play the market than you can “skillfully” play the lottery. Do you agree? If so, is there research that bears out this claim? At first glance, I’d think that people with more information about the market can play more “skillfully” (so to speak) than people with less information, and I would think that gaining more information about the market is not always trivial.
There’s still the possibility that genuine skill at playing the market does exist; it’s just that if it does exist, then it can’t be recognized, because if it could be recognized that, say, Scrooge McDuck is good at playing the market, then a good, intelligible strategy would be “Buy the same stocks that Scrooge is buying”.
The problem is that trying to outperform the indexes is a zero-sum game: For every dollar more than average that someone makes, someone else has to lose a dollar relative to the average. But if someone had a good, intelligible strategy (or an unintelligible strategy that is recognizable as good), then soon everyone would be using it. But if the strategy is actually good, then that would mean that everyone is outperforming the indexes, which is impossible.
Hmm… I’m not sure this follows. Couldn’t it be that there is a good intelligible strategy that works only under certain circumstances? And couldn’t one of those circumstances be that not too many others are following the strategy? In such a case, if everybody starts using it, it’s no longer a good strategy, but as long as not too many people are using it, it’s a good strategy (and by hypothesis intelligible, though perhaps I’m missing something and the hypothesis is unrealistic).
At some point, the disparity in the returns might suggest that a given investor’s method is better than following an index. Anyway, you seem to be assuming that the efficient market hypothesis is true, which is controversial to say the least.
And at some point, the disparity in returns might suggest that a given lottery-ticket buyer’s method is better than just holding onto your dollar.
Are you arguing that an automated trading program that beats the indices is not possible because eventually everyone will have it and at that point it’s no longer better than average?
If so, that seems to be ignoring the time period during which it did outperform the average.
I don’t see how the answer can be anything but yes (that an automated program can theoretically outperform the average if everyone is not using that same program) if the system is non-random.
Doesn’t this argument prove too much? It seems to show that there can’t be a good intelligible strategy for any zero-sum game.
Search on algorythmic trading, which is the generally used term for the automated trading you’re talking about. DE Shaw was one of the first ones to uses these systems. They are commonplace these days and all of the big investment banks use them to one degree or another. All the major banks are trying to replace a lot of the traders with algo trading.
Chronos, you are right that it is a zero sum game. That does not invalidate that the algo trading can be done. Most algo trading is “true” arbitrage trading that simply captures mispricings that are in the market. For example, a long call and short put at the same strike price and maturity are equivalent to a short futures contract. If the call, the put or the futures is trading enough out of line to cover the transaction cost plus some profit, then an algo program can do the execution.
By the way, doesn’t potential growth or shrinkage of the stock market imply (or anyway, make it possible that) it’s not a zero-sum game?
Suppose there’s a game involving competition for “resources,” where one new resource point is generated every minute and the two players compete for that point. Suppose winning at this game consists in having a higher score than you did at some point in the past. (This is supposed to be analogous to the notion of profit.) Then over stretches of time, isn’t it possible for both players to win? (Or, since the game never ends, maybe I should say it’s possible for both players to “be winning” though I’m not sure what game theory has to say about never-ending games.) Wouldn’t that be a non zero-sum game?
Just to add another perspective arguing that the market isn’t a zero-sum game, what if an intelligible, good strategy is not imitated by others? I’ve seen quotes like this:
That assumes that every investor is rational, and willing to put aside short-term speculation, there magic 8 ball or d20 for gains on the long-term horizon.
Is there any analysis of the market at all that can be made without the approximation that all investors are rational?
But of course. Behavioral finance.
Yes. A pattern matching algorithm does not need to know anything about the underlying processes, although will also be limited to patterns that repeat and couold ce
artainly get the “wrong” answer.
Edit: Actually, as I re-consider, it would assume people are not random, so if rational is synonomous with non-random then you are correct.