Mathematician Predicts Yankees To Dominate

anthemaniac writes "Computerized projections in sports are nothing new, but Bruce Bukiet of the New Jersey Institute of Technology has developed a model that seems to work pretty well. He projects how many games a Major League Baseball team will win by factoring in how each hitter ought to do against each pitcher in every game. His crystal ball says the Yankees will win 110 games this year, a pretty safe bet, many might agree. But he also projects all the divisional winners. He claims to be right more than wrong in five of the past six years."

I never understand these things...

[krbvroc1]

Isn't here some rule or law about 'fitting a curve' to past data? Yet, the sports predictions, and many of the 'stock market systems' are all about
finding some seemingly obvious pattern in past data. While you might come up with a 'back tested' model that matches really well,
it doesn't mean squat for the future.

Re:Claims to be right more than wrong, heh?

[BridgeBum]

Bruce is actually a die hard Mets fan. I helped work on this project with him back in my undergrad days 15 years ago or so. I doubt any of my code is still be used though. :-)

Re:I never understand these things...

[BridgeBum]

His models have evolved over the years, but he tries to simulate actual games using both individual statistics (players batting averages, etc.) as well as team trends (how well does a player do against a specific pitcher). He uses a large Markov chain to predict state transitions (Runner on first, no outs - how often does it go to two outs? That sort of thing.) Very interesting project, it was a lot of fun to work on. (I was an undergrad working with Bruce 15 years ago, when he was first starting this project. He's kept it going for years.)

Big Whup...

[Anonymous Coward]

Bill James came up with simple quantifiable statistics that could very accurately predict the success rate for a baseball team back in the '70s. The Oakland A's had a lot of success using those methods to put teams out of the field that would win between 95-100 games per year while spending as little as possible. It worked remarkably well and a book (Moneyball, by Michael Lewis) was written about it.

In short, this is old and well covered news, unless this guy has come up with a simulation that is significantly more accurate (doubtful).

Baseball is easy to predict

[obdulio1950]

Nobody could predict this one: http://www.planetworldcup.com/CUPS/1950/wc50index. html [planetworldcup.com] and the "Macacos" still cry about this......

From one of his students

[kenb215]

Wow, I never expected somebody that I knew to get on Slashdot. Bruce Bukiet is my Calculus II professor at NJIT.

He mentioned this before a few times, including today after that article made it to the most popular spot on Yahoo! [yahoo.com] News. This is more of a hobby for him than an official project.

From what he has said in the past about the model, it tends to overestimate the Yankees, among other reasons, because they often buy good players at the end of their prime. Thus the players won't play as well as they had in the past. He hasn't used it to make any bets. For the model, coming within a game or two of the actual results is considered a good prediction.

As some people above said, the model isn't intended to be extremely accurate, and is frequently off by a significant amount. The interviews he does are more to get people interested in math, and to see how it has real use, rather than to try and show off. He used to go into more details in the past, but doesn't now because they tend to confuse the interviewer, and don't make it into the final article.

Some pages of his own about the project are:
http://m.njit.edu/~bukiet/baseball/baseball.html [njit.edu]
http://www.egrandslam.com/ [egrandslam.com]

Re:Huh?

[AstrumPreliator]

...or a really, really bad statician.

Or a really good statistician. Remember, when you ask a statistician to crunch some numbers for you he'll reply back with "and what would you like the numbers to say?". They'll make it fit any curve you throw at them.