Baseball isn’t really an exciting game, except for a few select moments. What it is, especially late in the season, is a suspenseful game. To an invested viewer, the wait between pitches becomes excruciating, and each pitch represents an individual battle. If you’re ahead, every ball feels like a tiptoe closer to the ledge. If you’re behind, every out represents your finite number of chances slipping away. Moneyball uses your standard cinematic tricks to pump up the suspense behind its in-game moments (slow motion! backstories! cherrypicking the most exciting moments!), but in doing so gives the average viewer an experience closer to that which a die-hard fan has than any scene from a real game ever could.

- Eliminate mortgage interest deductions.
- Eliminate corporate healthcare deduction.

In principle, I agree with both of these ideas. You’re right that these deductions introduce market distortions; they are also regressive, especially the mortgage interest deduction. However, I feel that an instant end to either would cause some extremely unpleasant short-term disruptions to the housing market and health care markets, and I’d be careful categorizing these as bugfixes. If I had made an investment in a house based on current tax law, I’d be pretty upset if that law suddenly changed. That goes double for health care; I really really don’t want to shop for insurance as an individual. The incidence of these deductions means that they have a large base of support among middle-class and wealthy Americans, making their repeal a giant political undertaking.

The best path forward I see is to cap each of these deductions. This will undercut the regressive nature of the deductions, and allow for a natural phase-in mechanism, as inflation pushes the real value of the cap lower. Note that Obama has proposed this for mortgage interest, and a similar scheme is currently law as part of the Affordable Care Act.

- Abolish payroll taxes and shift that all into income tax.
- Reduce social security by making it means tested.

From a social justice angle angle, I agree that these are good changes. I’m not sure that I would actually implement them though; one of the most important components of these programs is that they’re baked into the American social contract at this point, which may explain their remarkable resiliency. I feel that program’s political sustainability is as important as its fiscal stability, and I think that having Social Security apply broadly to all Americans is a good way to ensure it has the support of the vast majority of Americans.

- Reduce social security by raising the threshold age.

Disagree on this point. Studies have shown that longevity is correlated with income, and as such this would be a deeply regressive cut.

- Reduce military spending.

Totally agree on this point, though I think that free riding by American allies is an important reason for the “as much as N other countries combined” stats. As such, hopefully cuts would ensure a more multilateral foreign policy where a number of other countries that generally support American foreign policy share more of the costs.

How you want it to be

The problem is, our current situation is out of equilibrium. The financial crisis caused two major shifts. On the supply side, there was a large amount of deleveraging (shifting from debt-backed consumption to savings). This shifted the supply of loanable funds outwards, decreasing the equilibrium rate of return. In addition, the evaporation of a large amount of demand means that firms now perceive a lower rate of return on new capital expenditures; why invest in new factories and equipment if nobody wants to buy the additional output you produce? As a result, the demand curve based on rate of return contracted, also decreasing the equilibrium point.

Bad News Bears

As Paul Krugman and others have argued, this puts the equilibrium interest rate below zero, but interest rates don’t go below zero, so we have a gulf between the amount of usable savings and the amount of demanded money. As a result, we can pump a lot more demand in the form of deficit spending into the economy before interest rates start to rise and private investment starts to be crowded out.

(Whoa, are we returning to form with short, on-topic posts?)

IBM has accomplished something impressive by building a computer system that can answer Jeopardy! questions with decent confidence, particularly given the wordplay and nuance for which the show’s writers are known. But if the computer emerges victorious in this week’s exhibition match, does that mean it is “better at Jeopardy”?

Arguably, it does. But not for any really exciting reason.

Over the last week or so, I’ve rambled on a bit at people who will listen (and many who won’t) about the fact that Jeopardy is really two different games.

1. A contest of knowledge. This is, of course, what the show is known for. At its core, Jeopardy is about awarding points to players who know stuff.
2. A contest of reflexes. Viewers often underestimate the importance of speed in any Jeopardy-like game. Speed of recall is important, of  course, but reflexes prove even more important in close games.

(Well, it’s really three different games: Final Jeopardy is almost purely a game theory puzzle, as the optimal wagers are typically independent of a player’s confidence in the category. And it’s actually four different games, if you count an almost fanatical devotion to the Pope. …I’ll come in again.)

It’s useful to understand some rules. Players may not buzz in until the host finishes reading the question. (This wasn’t always true, but it does tend to make for better television.) When the question answer is over, somebody off-stage arms the buzzers. This activates a set of lights surrounding the game board, signaling to the players that it is legal to ring in.

Watson is required to physically actuate a button. But that’s still blatantly unfair; it’s undeniable that a computer could easily win a reflex contest against humans. That is why I have a problem with the portrayal of the Watson match as pitting man versus machine in a battle of knowledge and wits. The game outcome purportedly measures Watson’s question-answering (or answer-questioning?) skill, but that characterization is founded on the idea that Jeopardy is purely a game of knowledge.

It’s impossible to know whether Watson is winning the knowledge contest, because it is winning the reflex contest hands-down.

All three players are really, really good at the knowledge contest. Brad and Ken have demonstrated that throughout their nearly 100 combined previous appearances on the show. As for Watson: on the 30 Double Jeopardy clues aired on Tuesday’s show, Watson was confident in the correct answer for 25, had the wrong answer for 1 (“Picasso”), and opted not to buzz in for 4 (on one, the Daily Double, its best guess was correct; on the other three, its best guess was wrong). That’s damned impressive, and a fine achievement. The Jeopardy round was less good, but still impressive: Watson had the right answer for 19 clues, a wrong answer for 5 clues, and abstained from 6 (1 of which, “the 1930s”, would have been correct).

But when all three players know an answer, the game becomes a buzzer race. The popular portrayal of the Jeopardy IBM Challenge gives far too much credit for Watson’s past victories to the question-answering capabilities that IBM has developed.

The last two nights have made it very clear that Watson is absolutely dominant on the buzzer. Of the 29 clues in the Jeopardy round (Daily Double excluded), Watson successfully rang in first on 17 of 24 attempts. Of the 28 Double Jeopardy clues, Watson successfully rang in first on 22 of 25 attempts. In total, Watson rang in first on over two-thirds of all opportunities (39 of the 57 non-DD clues read). During Ken’s 75 regular-season games, he was similarly dominant versus mere humans, ringing in first 61.45% of the time. A significant fraction of those wins presumably were uncontested—coming when neither opponent was attempting to buzz in—a much less common scenario when facing Jeopardy’s greatest champions, making Watson’s .684 average all the more absurd.

Incidentally, how can you tell that the humans are trying and failing to buzz in? As the camera switches back to show the players at their podiums, you can usually see Brad twitch a bit as Watson’s podium lights up. That’s how he buzzes: he’s found that contracting every vaguely-related muscle is faster than trying to move just his thumb or finger. Some players have been known to hold their arms crossed, with the signaling device against their body, so that any arm movement activates it. Ken’s movement is much less pronounced–the best indicator that he was trying to ring in is that he looks frustrated a split-second after he fails. (This happened frequently when Brad pounded him in the 2005 Ultimate Tournament of Champions.)

It’s looking increasingly likely that Watson will win in a blowout thanks to its buzzer dominance and the resulting Daily Double opportunities. But news stories that scream “Computer defeats humans on Jeopardy!” give too much credit to the computer by ignoring the aspects of the game that are heavily rigged in its favor.