Monday, May 21, 2012

Art of Strategy

Version 0.6

Comparative Errata for The Art of Strategy by Avinash Dixit and Barry Nalebuff

This quasi-review has a kind of funny story behind it. A couple of things I read in the the book struck me oddly, so I wanted to find the errata and wound up emailing the authors in search of it. One of the authors responded with a link to a page of "Corrections" that I'd been unable to find, presumably because I was searching around the idea of "errata". Therefore I now claim the first erratum is that there "Corrections" page should include the search target "errata"--but my claim must be wrong, since my first test search now returns the authors' page in question... No, they didn't add "errata" in the meantime. I checked the HTML source just to be sure. I just overlooked it?

Anyway, in the meantime, I had mentioned the book to a few researchers in related fields, and one of them said that he hadn't read it, but the reviews made it sound like the author's earlier book on the same subject, Thinking Strategically, might be more interesting. He asked me for my opinions about the book, which helped lead to this review and which also caused me to start considering the book in terms of the question "Is this new version a classic text that people will read and reference for years?" It covers newer examples, but I was trying to decide if it was a better book. Of course I can't answer that question without access to the earlier book, but my tentative answer would be that "No, it doesn't seem like a great classic, but it's a sound introduction." I have read a number of related books, and none of them really seem to stick in my mind like The Selfish Gene by Professor Dawkins or Professor Hofstadter's Godel, Escher, Bach, just to pick a few examples that come to mind.

However, I've decided to go through my own notes as I read the book and consider them while going through the authors' own errata.

Let me begin by saying that my initial concern involved the unnumbered "quick trip to the gym" on page 50. Professor Nalebuff responded to my question by saying that the second paragraph should be regarded as the answer, though I think the wording is ambiguous. I did feel that it felt like a trick question, basically reversing the direction without any real change, but at a minimum, it created confusion as to the format of the actual trips to the gym... Does that count as an error?

Anyway, I have to say that I feel that their first erratum for page 17 does appear to indicate a minor error on the webpage. The text of the paperback version matches the webpage, but there is no asterisk on the webpage, which is supposed to indicate a mistake from the original hardbound version that was corrected in the paperback. Well, the paperback has been corrected, but if there is no asterisk, then something must be wrong. No asterisk has to mean that both the hardbound and paperback printings are wrong and will disagree with the webpage. The other possible case is that both versions were correct, in which case there's no erratum to report on the webpage. (The last theoretical case is impossible because I have the paperback here, and it is correct according to the webpage.) Anyway, this was weird enough to get me to reread the correction and the book several times to make sure they were the same... If there's a difference, then I missed it several times, though that would mean the website could be regarded as okay...

As regards the actual body the footnote in this erratum, I still find it extremely dubious. It was supposed to support the textual contention that "[The incumbents'] advantage in fundraising is what gives them job security," but it does not. The famous quote is that correlation is not causation, and it is possible that the money is directed to the probable winner simply because donating to a loser is pointless. In other words the money is following the safe winner rather than making the winner safe, and I contend that the actual cause is not the money, but the collusion of both parties in the gerrymandering that makes incumbents safe. In the winner-take-all American elections, the two parties may disagree on every other issue, but both parties still agree that they want to be reelected and that negotiated gerrymandering is the mechanism that insures their reelections.

What actually happens in the real world is that the weaker party sometimes gets gerrymandered into fewer districts, and then there will be a sincerely contested election, but only in the primary election of the weaker party. In fact, it is in the strategic interest of the stronger party to destroy the districts of the strongest politicians of the weaker party, because that forces those stronger opponents into primary contests against entrenched incumbents, hopefully eliminating their strongest adversaries. However as far as the actual elections are concerned, the effectiveness of modern computer-enhanced gerrymandering is such that the outcomes are almost certain--and of course the smart donors know better than to waste their donations on sure losers.

The authors' next acknowledged error is on page 94, but I doubted the accuracy of their footnote on page 24. Buying stocks only provides capital to the firm when the shares are purchased directly from the company, and that is not true for most share purchases. They say that share purchases provide investment resources to the firm, but actually it mostly works the other direction these days. Firms use cash that could be used for investment to repurchase some of their own shares and often then give those shares away as performance incentives. Maybe this benefits the companies, but you have to wonder since the terms of the rewards are too often determined by the same people who get rewarded...

I felt that page 43 had a kind of self-contradiction. At one point they say that "greater freedom of action can never hurt", but a few lines later they say "...can quickly get too complicated to draw or use." In other words, too much freedom of action becomes so difficult to analyze that the choice becomes overwhelming, which seems kind of hurtful to me. Is it sufficient to say that "paralysis by analysis" can be quite troublesome, or do I need to make a more elaborate argument about the time spent in researching the options?

Why do the authors seem to get into so much trouble with footnotes? On page 59, the footnote seems to be confused about how chess programs work, since those programs do exactly the same thing that they attribute solely to good human players.

The erratum on page 94 is wrong in the paperback and has no asterisk on the webpage, which supports the interpretation I mentioned that both printed versions are wrong, but the erratum on page 95 is printed correctly in the paperback and has no asterisk on the website.

Their errata on pages 95 and 150 are correct in the paperback and properly have asterisks on the webpage, so I hope these cases don't need to be checked... However, on pages 138 and 163 they were back to corrected errors without the asterisks... I think it's time to quit worrying about either side of the asterisk here and just suggest that the authors go ahead and make one list of errors for the hardbound version and a hopefully much smaller list for the paperback. At least that's how it seems to be working out so far. However, that makes me realize there is another possibility. Perhaps my paperback is a later printing that has corrections for errors that appeared in both the hardbound and earlier paperback versions?

Focusing on my own questions, on page 136, they conclude with an observation that the caller has the number, but ignore the fact that most modern phones also show the caller's number to the callee. I think the norm these days is for both sides to know, so their resolution is a kind of error.

On page 212 they say "his defeated bid for reelection", which should be "defeat in his bid for reelection" or "his unsuccessful bid for reelection". Okay, so it's a minor one, as is the mistake on page 340, which should be "one in San Francisco".

On page 465 in the Notes, they said that Rumsfeld's use of English was good, and want to defend his analysis, though he only considered three of the four cases. They say his analysis is not "foolish", but I cannot regard it as complete or profound without consideration of the "unknown knowns". Some of our greatest mistakes really are described by "I knew better."