----- Original Message ----- From: "David Brin" <[EMAIL PROTECTED]> To: "Killer Bs Discussion" <brin-l@mccmedia.com> Sent: Tuesday, November 01, 2005 3:53 PM Subject: Re: Brin: predictions
> > > Holy incestuous Feedback Batman! > > > > It'll never work. Heisenberg had it right. > > > SO you are saying that nobody is right more often than > other people? > > Or that some people do not deserve more credibility > than others, for being right more often? Hm. Unique. I think that the problem with such a technique is properly identifying the statistics associated with being right n times out of m. Let me give two examples that frame the problem. 1) Let us consider a single elimination tournament based on coin flipping. Person 1 flips a coin. If person 2 matches the flip, they win; otherwise person 1 wins. If there 2^16 people entered in the tournament, there will be one person who won 16 times in a row. This person, with a perfect score, is no more likely to win the next coin flip game than anyone else. A real life situation that almost matches this is the performance of various investment funds. There are funds that boast of being better than the average of all other funds for 9 years out of the last 10. Yet, when funds like these are examined, they tend to have no more chance of being above the average for the 11th to 15th year than any of the other funds. It's akin to there being a 50-50 chance that a given physician was in the bottom half of his/her class at medical school. Past performance is, when considering funds in this manner, rarely an indicator of future performance. There are exceptions to this rule, of course, a very good case for Buffet can be made: that he really does know more about investing than most fund managers. His performance is longer term, and more statistically significant. 2) Richard Feynman is the classic example that I have for the other end of the spectrum. As with many areas of physics, Dr. Feynman made a rough guess predicting the rapidity distribution of pions in high energy interactions. Feynman scaling was found to be right, to the order of magnitude that he predicted this. As you know, but many on the list do not, he made many intuitive predictions, such as this, after he was a high profile physicist. If you look at the predictions made by physicists with a high profile, you will see his stand out. The percentage of times he "guessed right" was near legendary when I went to school. There was a rule of thumb that a Feynman guess was worth more than careful computation by nearly anyone else. My gut feeling is that an open prediction registry will be closer to the funds manager case than the Feynman case. I think a critical factor is limiting the number of potential candidates for being right. There are not that many physicists who have already done stand-out work. Thus, we do not have enough predictions to have a track record like Feynman emerge from mere chance. In economics, the clear example is Greenspan. List member and fairly well known economist, Brad DeLong, has noted how often Greenspan has guessed right when he himself has guessed wrong. He talked about cloning Greenspan because he has an uncanny ability to "guess right". Since there is only one chairman of the Fed. Reserve Board at any one time, this track record cannot be explained by statistics alone. There has to be some skill that Greenspan has that Brad does not. Brad acknowledges the skill, but cannot put his finger on exactly what it is. The common denominator that I see between Greenspan and Feynman is that they demonstrated, apart from their intuitive ability, above average skill in the field before many their guesses went on record. This culling significantly lowers the possibility that such a track record is due to chance alone. This allows those with unique intuition to rise above the background of lucky guessers. The final factor that I wish to consider is what I'd call the "astrology" factor. If you look at general predictions of astrology for most people, there will almost always be a manner in which the predictions were right. The reason for this is not due to a real power of the planets over our lives, of course, but the real power of people to match vague predictions to what actually happened and thus see the predictions validated. In the case of Feynman and Greenspan, there were hard numbers that their predictions can be matched against. I believe that aspects of science fiction novels that are later matched to reality can suffer from the same "20-20 hindsight" problem. The parts of the prediction that match are focused upon, and the differences are minimized. Thus, the predictions seem more accurate than they really were. So, while I am intrigued by the idea of a predictions registry, I think that there are two factors that must be considered before such a registry could really be useful. 1) A means of lowering the probability that good looking track records will emerge from chance alone. 2) A means of setting firm criteria for determining sucess (which of course ties into 1). Dan M. _______________________________________________ http://www.mccmedia.com/mailman/listinfo/brin-l