Humans vs computers, chess vs trading

Discussion in 'Trading' started by Cutten, Nov 23, 2003.

  1. Cutten


    I was looking at the Kasparov vs Fritz chess game, which was a draw. After reading some background, it seems pretty clear that in a few years the best computer programs will be able to regularly beat the human world chess champion. I then wondered why, when computers are so good at chess, how come they are no way near as good at trading? After all, trading is incredibly simple - there are only a handful of variables (price, time, volume, open interest), and you can only be long, flat, or short.

    Now I am not very knowledgeable about computers, but the only computer programs that seem to be able to make profits similar to humans are arb programs. Profitable directional programs seem not to exist. Also, the vast majority of system trading seems to be done by humans, using the computers only as aids to system-design, rather than writing a program which then *creates* systems of its own accords. A chess program, once written, can play world class chess. Why can't a trading program, once written, scan the markets itself and then trade profitably?

    I assume that investment banks and other market players have spent literally billions of dollars trying to write a computer program that can devise profitable trading systems. Given the massively higher resources ploughed into trading computers vs chess computers, why haven't they achieved anywhere near the same success?
  2. i suppose there's lots of technical jargon out there that can make my remarks look pretty simplistic, but the bottom line is that the machine has no soul. That's the dimension that a machine will never have.

    It may dazzle with brilliance and/or baffle with bs, but it <i>ain't never gonna have no soul.</i>

    If they find a way to bridge that gap (i.e. wire the perfect memory and lightning processor to a human) then, well, I hope it's all over long before that.
  3. Cutten


    In chess, computers are still quite weak at "positional" analysis i.e. knowing just from looking at a position, what is the correct long-term strategy. They are very good at brute force calculation many moves ahead.

    Perhaps trading is dominated by "positional" analysis, and brute force is only useful in niche areas e.g. arbitrage, certain options calculations etc. Still, that begs the question as to why positional analysis can't be programmed well.
  4. the machine can not acquire the sixth sense.
  5. Ultimately I believe that humans are the ones to decide when to trade that system and to mettle with the parameters. With that discretion in mind, a system is not allowed to run its proper course.

    Quotes from Ed Seykota in Market Wizards

    In the early 1970s, Seykota was hired by a major brokerage firm. He conceived and developed the first commercial computerized trading system for managing clients' money in the futures markets. His system proved quite profitable, but interference and second-guessing by management significantly impeded its performance. (p.151)

    I do not publicize my track record other than my "model account," which is an actual customer account that started with $5000 in 1972 and has made over $15 million [up til 1988 when the interview was conducted]. Theoretically, the total return would have been many multiples larger had there been no withdrawals. (p.157-158)

  6. That statement is false Cutten. There are not only a "handful of variables". A markets participants can manifest there beliefs in an unlimeted number of ways at any given moment. With chess the game is static for the most part, where as the market is organic and constantly self-adapting...

    PEACE and good-specul8tion...
  7. Pabst


    It's telling that a computer still can't consistently beat a Master, even in a game with a finite number of moves. Price structure as expressed in double auction markets has infinite structural possibilities. While I believe there is some degree of "market logic", I do not think one can build a consistently profitable system that excels in all price environments. While a Chess computer can plot out billions of future moves in a few minutes, it doesn't really think, it merely processes pre-determined statistical probabilities very quickly. Since the notion that one price causes another is tenuous and dependant on multi variables, trading is akin to two chess games being played simultaniously with the Queens randomly rotating from board to board.
  8. Why would you think computers would be able to beat the world chess champion in a few years. They have been saying that for decades, and from time to time a computer program comes out that can beat the world champion, but only until he figures out how to exploit the program's weakness. Just like the market figures out how to exploit any systems that you might consider trading. Only that the market is not just one mind, but hundreds of very gifted minds who choose to make gazillions rather than play chess.
  9. Yeah! Thats what I meant to say :D
  10. chessman

    chessman Guest

    As others pointed out some of your assumptions are not correct, there are directional computer based traders and capital management firms that have consistently done quite well.

    Here are a few

    btw in chess as Boby Fisher pointed out when he was 15 that there are no new chess moves everything is know.
    #10     Nov 23, 2003