Projections For MLB Season

A Spin On The Fave Pastime

I actually like this part of the spring training where people present us with their projections. People are constantly predicting things through their intuition but computer technology has also given rise to the multiple simulated seasons method where people run the teams through the season many times to get an answer.

The most interesting projections come out of the multiple projections systems that get put through the Diamond Mind baseball game 1000times, done at the Replacement Level Yankee Weblog, courtesy of 'SG'. The various projection systems all emphasise different aspects of a player, league contexts, ballparks and so on, but once they go into the blender of 1000 seasons it's remarkable how similar the outcomes are.

The average of the 6000 simulated seasons has the AL as:

EAST:

• NYY: 95.9-66.1


• Bos: 94.3 - 67.7


• Tam: 90.1 - 71.9


• Tor: 75.6 - 86.4


• Bal: 74.5 - 87.5

CENTRAL:

• Cle: 85.5 - 76.5


• Det: 81.4 - 80.6


• Min: 79.5 - 82.5


• KC: 74.6 - 87.4


• ChA: 74.1 - 87.9

WEST:

• LAA: 85.4 - 76.6


• Oak: 81.1 -80.9


• Sea: 77.8 - 84.2


• Tex: 72.1 - 89.9

All of which seems like a just about what your intuition might tell you with a few caveats that SG comments upon on the entry when you follow the link above. Of course, the projection systems have all come in under on Chicago for a number of years, and injuries and such tends to affect how these things go.

Of the 6 projection systems, CHONE likes Boston more over the Yanks, but apart from that the 6 projections line up roughly as above.

What got my attention today was this simulation here on Yahoo.

2009 American League Preview

SIMULATION BASED FORECASTING & METHODOLOGY
AccuScore uses past player performance statistics to describe how players perform under different environmental, match-up, and game situation conditions. Using projected starting lineups AccuScore simulates each game of the 2009 Season one at-bat at a time. By repeating the simulation 10,000 to 20,000 times per game AccuScore calculates the precise probability each team has of winning each game, winning their division, making the playoffs, advancing in the playoffs and winning the World Series. Here's AccuScore.com's analysis of the 2009 American League.

Wow. 20,000 sims per game! And the AL looks like this according to this test:

EAST:

• Bos: 95 - 67


• Tam: 93 - 69


• NYY: 91 - 71


• Tor: 80 - 82


• Bal: 68 - 94

CENTRAL:

• Min: 86 - 76


• Cle: 83 - 79


• ChA: 82 - 80


• Det: 79 - 83


• KC: 75 - 87

WEST:

• LAA: 88 - 74


• Oak: 81 - 81


• SEA: 72 - 90


• Tex: 69 -93

As you can see, that's a pretty different outcome in the 1-2-3 slots for the AL East and the AL Central is quite different too. Accuscore thinks the extended absence of A-Rod is enough to pare back the Yankees while still having a 25.4% chance of winning the division. They're also high on Minnestoa's pitching, while they're a little sour on Cleaveland's rotation. There's little love for a bounceback season for Detroit. All of which are interesting takes on the possible outcome of the season.

Hence I thought I'd note this all down so that when the season is done, I can come back and have a look and see whose projection system did what, and to see if Accuscore's projection does any better than the sabermetricians.

UPDATE:

Here's another one.

Professor Bruce Bukiet, of the mathematical sciences department at the New Jersey Institute of Technology, developed a formula in 2000 to predict game results. As the season gets under way Sunday, his formula sees the defending World Champion Philadelphia Phillies, the New York Mets (which Bukiet admits is his favorite team) and the Atlanta Brave will each win 88 games and tie for the NL East title.

Bukiet's formula says the Chicago Cubs Los Angeles Dodgers and Los Angeles Angels of Anaheim should easily win their divisions.

The New York Yankees, predicted to win 99 games, were a two-game favorite over the Boston Red Sox in the AL East and the Cleveland Indians were installed as the team to beat in the AL Central.

"These results are merely a guide as to how teams ought to perform. There are many unknowns, especially trades, injuries and how rookies will perform," Bukiet said. "Over the years, the predictions have been about as good as those of the so-called experts. It demonstrates how useful math can be in understanding so many aspects of the world around us."

A 3-way tie in the NL East sounds interesting.