Monday, July 18, 2005

Computer Science Has Been Very Very Good To Me

Ever notice how computer science departments in the US are like baseball teams. They try to hire the best players so they can be better than other departments (try to be in the "top ten" for instance). Already strong departments with lots of resources continue to hire the best people and become even stronger. MIT, despite being close to Boston, is like the New York Yankees of computer science.

One can push an analogy too far and I've already crossed that line but let's keep going.

  • Baseball players are initially tied to a certain team though after a certain number of years they can become a free agent or prevent their team from trading them. Professors can become free agents after any year and after seven years, if they are still with the department, get a no-fire clause.
  • Baseball teams have minor leagues to train young players. We have graduate students.
  • Baseball has had a strong commissioner who mediates disputes and can make changes for the good of the game. We could use someone like that.
  • Baseball sells naming rights of its stadiums. Universities sell naming rights of their buildings.
  • Baseball has a hall of fame honoring the very best. We have the Turing award. But like baseball we could also have a physical location with memorabilia like the original draft of Cook's paper or the chalk Manindra Agrawal used to prove Primes in P with his students.
  • Baseball teams trade players. Imagine David Karger and Madhu Sudan for Umesh Vazirani, Luca Trevisan and a grad student to be named later.

22 comments:

  1. Aha, but then we should also have virtual baseball: compose a team of researchers and see which team has the most FOCS, STOC, SODA, CCC,... publications at the end of the season.

    - Wim

    ReplyDelete
  2. Plus we could have agents who would negotiate the best salary possible. Imagine all MIT professors as IMG clients...

    ReplyDelete
  3. The latest scores:

    STOC 2005; # of papers by group

    MIT - 13
    Berkeley - 9
    Princeton - 8

    (Note: The group at MIT is 3 times bigger than the groups at Berkeley and Princeton.)

    ReplyDelete
  4. The group at MIT is 3 times bigger than the groups at Berkeley and Princeton.

    Excuses, excuses. MIT beat 'cha :-)

    Other notables:

    Microsoft Research - 8 papers
    Weizmann - 6 papers (+2 multi institution)
    Tel Aviv - 6 papers
    Technion - 5
    Georgia Tech - 3
    CMU - 3
    Chicago - 3 (+1 TTI)
    UWash - 1
    Stanford - 1
    Waterloo - 1
    UT Austin - 1
    MPI - 1

    ReplyDelete
  5. .


    I hate baseball. How come this scam-pseudo-sport with 20 seconds of action per hour got everyone hooked??

    .

    ReplyDelete
  6. I don't understand why Stanford ranks so high in "top theory schools" lists. We all know that in the US, the "big four" are MIT, Berkeley, Princeton, and CMU.

    ReplyDelete
  7. s/baseball/basketball. or football for that matter.

    ReplyDelete
  8. What about Cornell?

    ReplyDelete
  9. So, I gotta ask. Who are the Cubs?

    ReplyDelete
  10. Baseball is not a sport; it's a pastime (literal translation: a way of passing time). It's not meant to be watched on television, and it's not meant to be watched without beer and friends.

    Basketball? Basketball is an exciting, artistic, and beautiful sport; it's powerful, graceful, almost majestic. I find the mere comparison to baseball bordering on offensive.

    ReplyDelete
  11. I suppose that it's about the only metric available at a first glance, but is it really so good that the _number_ of more-or-less unrefereed conference papers is the standard measure of accomplishment?

    ReplyDelete
  12. The title of the post is the punchline in the analogy between TCS and baseball. "Basaball has been very very good to me" is from a Saturday night live skit mocking an immigrant baseball player. Yes, they import immigrants to play baseball and...

    ReplyDelete
  13. One difference between sports and science is that in sports the referees are not associated with the teams!!!

    In other words, when you are so proud that MIT has 13 papers in STOC, maybe you want to think about this again...

    ReplyDelete
  14. The numbers talk a lot about the universities.

    princeton: small group, very talented.
    berkeley: reasonable group, under the aegies of GODS.
    MIT: BIG group, numbers game, get any person with potential. main line: get papers.
    Stanford, CMU: too diverse interests to get substantial work done in an area of CS, one in which there is less immediate monetary benefit, with the present day's state of TCS funding.

    Thanks for this post Lance!

    ReplyDelete

  15. I suppose that it's about the only metric available at a first glance, but is it really so good that the _number_ of more-or-less unrefereed conference papers is the standard measure of accomplishment?


    I have encountered several theoretical computer scientists (including Lance himself, on his CV) referring to "unrefereed conference papers". Can someone please clarify this terminology? Aren't conference papers refereed papers, because they are refereed (for what it's worth) by the program committee members?

    ReplyDelete
  16. continuing the previous thought

    Maybe a better metric (for the baseball analogy) would be to look at the accepted/rejected ratio, then whoever has the best "win/loss" record is the winner!

    ReplyDelete
  17. In baseball, the home plate defintion is not physically possible: http://mathworld.wolfram.com/HomePlate.html

    So what's the equivalent in computer science?

    ReplyDelete
  18. -- Basketball? Basketball is an exciting, artistic, and beautiful sport; it's powerful, graceful, almost majestic.

    Quoting someone else: soccer is all foreplay no climax, basketball is all climax no foreplay. The first time someone wins with a buzzard beater basket one gets excited, the nth time it happens during the same playoff season one starts suspecting how real the whole excitement is...

    -- is it really so good that the _number_ of more-or-less unrefereed conference papers is the standard measure of accomplishment?

    Relax, it's a joke, starting from Lance's title.

    -- One difference between sports and science is that in sports the referees are not associated with the teams!!!

    Now, now, that sounds like sour grapes. Biases aren't that much of a problem in the current system, which is not to say there aren't any. Certainly blind refereeing---while still imperfect---would be an improvement over the current system.

    ReplyDelete
  19. It is pretty amusing to imagine certain unnamed persons giving the "Computer science has been very very good to me..." speech.

    However, the mismatch between the weight that the community puts on conference proceedings and the amount of effort a typical sub-referee puts into the process is a serious issue.

    ReplyDelete
  20. -- However, the mismatch between the weight that the community puts on conference proceedings and the amount of effort a typical sub-referee puts into the process is a serious issue.

    This is true for pretty much all of computer science... I'm still hoping that as the field matures we'll move back to a journal based system and away from this 60-papers-per-PC-member situation.

    ReplyDelete
  21. According to another performance metric -- the number of best paper awards at STOC/FOCS--
    it seems that the world's two top theory groups are Princeton and Weizmann Institute (Israel), and by a wide margin.

    ReplyDelete
  22. Well, let's just hope that the Red Sox become the MIT of baseball.

    ReplyDelete