GIGO, game theory, and Nash dominant strategies
Firstly, a little historical note: Despite the PR boost that a box-office hit movie can create, John Nash did not create game theory; in fact he's hardly even a leading light of the field, despite the Nobel Prize for his equlibrium observation. In fact, rumors within the [small and close-knit] scholarly field are that the Sylvia Nasar biography and subsequent Russell Crowe-starring biopic were part of a deliberate PR strategy by Nash's friends to rehab his then rather battered image to coincide with a back-room push for the prize. In fact, the canonical text of game theory is John von Neumann and Oskar Morgenstern's classic 1944 Theory of Games and Economic Behavior, which like most truly revolutionary texts, was a complete bust when it was first published, but in fact laid out not just the seeds, but an entire prescription for a new field. After von Neumann and Morgenstern, the second key text of game theory has to be Kenneth Arrow's impossibility theorem, one of "the three great negative results of the 20th century" according to my friend Raja Sengupta. (Heisenberg's uncertainty being the second; can you guess the 3rd?)
Secondly, a direct response to the guest game theorizing of Ahpah on Yelnick (is that you lurking behind that screen name, PM?)
1) The vast statistical advantage of Bush in the coin-flip scenarios (~ 2/3 - 1/3 vs. Kerry) is entirely dictated by Bush's 19 electoral vote lead in "solid" states, 222 - 203. Thus Bush only needs an additional 48 electoral votes to reach the magic 270, whereas Kerry needs 67.
So what, you say?
Well, let us turn to a historical example for elucidation. As discussed at great length in the excellent book Rationalizing Capitalist Democracy, which happens to have been written by my ex-wife and copy-edited by me, game theory was invented by von Neumann and Morgenstern and almost immediately applied by them and a host of other RAND-affiliated researchers to the problem of nuclear strategy. The most notable policy success of this group, any of whom could have aced the GLAT (or, more likely, written it - von Neumann invented computer science as we know it) , was to convince first the proto-Kennedy administration, and later much of the governing establishment of the United States, of the existence of a "missile gap" between the well-armed Soviet Union and the weakly-thewed United States.
The only problem with this policy 'success' was that while the math was impeccable, the data was sadly lacking. All intelligence estimates regarding Soviet capabilities were wildly overstated, and in fact all that the build-to-catch-up missile strategy of the United States accomplished was a serious escalation of the Cold War. The recently departed Paul Nitze, author of the seminal NSC 68 and a founding member of this game-theoretic cadre, perhaps realized the folly of the strategy late in his life, when he significantly departed from form with his famous "walk in the woods" at an arms control summit under Ronald Reagan, where he attempted to set aside tried-and-dead-end strategies and cut a sensible deal.
But I digress.
The real point here is that ahpeh's analyis is based entirely on Rasmussen polling data, which spots Bush a formidable 19 vote electoral lead before the calculation even begins. This year, we have empirical evidence that reputable polls are returning both externally (to each other) and internally (to themselves over short time periods) inconsistent data. We further have logical evidence that the polls are systematically mis- or under-counting huge groups of voters, including cell-phone-only, caller-ID screening, not-answering, living-at-home, etc. Finally, there is systematic bias in any "likely voter" poll calculation, given today's (or rather, Tuesday's) highly "unlikely" voter climate.
Garbage In = Garbage Out, and the precision of the calculation mechanism only improves the seductive fineness of the trash. Based on my HUMINT from the field, a long-forgotten and implicitly discredited tactic in governmental (and, to our detriment, CIA) circles, I will boldly state that someone is going to win big on Tuesday, and that someone is almost certainly John Kerry.
- Ethan
Secondly, a direct response to the guest game theorizing of Ahpah on Yelnick (is that you lurking behind that screen name, PM?)
1) The vast statistical advantage of Bush in the coin-flip scenarios (~ 2/3 - 1/3 vs. Kerry) is entirely dictated by Bush's 19 electoral vote lead in "solid" states, 222 - 203. Thus Bush only needs an additional 48 electoral votes to reach the magic 270, whereas Kerry needs 67.
So what, you say?
Well, let us turn to a historical example for elucidation. As discussed at great length in the excellent book Rationalizing Capitalist Democracy, which happens to have been written by my ex-wife and copy-edited by me, game theory was invented by von Neumann and Morgenstern and almost immediately applied by them and a host of other RAND-affiliated researchers to the problem of nuclear strategy. The most notable policy success of this group, any of whom could have aced the GLAT (or, more likely, written it - von Neumann invented computer science as we know it) , was to convince first the proto-Kennedy administration, and later much of the governing establishment of the United States, of the existence of a "missile gap" between the well-armed Soviet Union and the weakly-thewed United States.
The only problem with this policy 'success' was that while the math was impeccable, the data was sadly lacking. All intelligence estimates regarding Soviet capabilities were wildly overstated, and in fact all that the build-to-catch-up missile strategy of the United States accomplished was a serious escalation of the Cold War. The recently departed Paul Nitze, author of the seminal NSC 68 and a founding member of this game-theoretic cadre, perhaps realized the folly of the strategy late in his life, when he significantly departed from form with his famous "walk in the woods" at an arms control summit under Ronald Reagan, where he attempted to set aside tried-and-dead-end strategies and cut a sensible deal.
But I digress.
The real point here is that ahpeh's analyis is based entirely on Rasmussen polling data, which spots Bush a formidable 19 vote electoral lead before the calculation even begins. This year, we have empirical evidence that reputable polls are returning both externally (to each other) and internally (to themselves over short time periods) inconsistent data. We further have logical evidence that the polls are systematically mis- or under-counting huge groups of voters, including cell-phone-only, caller-ID screening, not-answering, living-at-home, etc. Finally, there is systematic bias in any "likely voter" poll calculation, given today's (or rather, Tuesday's) highly "unlikely" voter climate.
Garbage In = Garbage Out, and the precision of the calculation mechanism only improves the seductive fineness of the trash. Based on my HUMINT from the field, a long-forgotten and implicitly discredited tactic in governmental (and, to our detriment, CIA) circles, I will boldly state that someone is going to win big on Tuesday, and that someone is almost certainly John Kerry.
- Ethan
posted by Ethan at 6:01 PM
0 Comments:
Post a Comment
<< Home