<
>

Diamond Mind: Simulating the Series

10/22/2005

Editor's Note: Diamond Mind is devoted to providing realistic, strategy-oriented baseball games for use at home and on the Internet. ESPN.com asked Diamond Mind to simulate the World Series, and here, Tom Tippett describes how it was done. For more information, go to Diamond Mind's Web site: www.diamond-mind.com.


The players were rated to perform based on their 2005 statistics, including left/right splits. When the simulations were run, the final rosters for the World Series had not yet been announced, so we used the 25-man rosters on MLB.com.

All games were simulated with the computer manager handling both teams according to the manager profiles we set up.

For each team, the manager profile included four saved lineups, for left- and right-handed opposing pitchers and for DH and non-DH games. Each of those starting lineups was paired with a depth chart that governs how bench players are to be used, including defensive replacements and platoons.

Both teams were set up with a four-man starting rotation based on recent usage. For Chicago, we went with Contreras-Buehrle-Garland-Garcia. For Houston, it was Clemens-Pettitte-Oswalt-Backe. In both cases, it allowed all starters to work on normal rest. Relievers were assigned to the roles in which they've been used -- closer, setup man, long relief, or mopup duty.

Pitcher fatigue isn't a big issue for this series. Chicago has had enough time off so that all of its pitchers are at full strength. Houston played only one game in the four days leading up to the Series, so we entered the pitch counts for their Game 6 win. But those pitch counts won't have much of an effect because Oswalt wasn't slated to pitch until next Tuesday and the Game 6 relievers only threw a few pitches before getting two days off.

Because there is no way to predict who might or might not get hurt in the course of short series, the DMB injury system was turned off.

The DMB weather system was turned on, so if you check the box scores, you'll see typical October weather for the two cities.

Running the simulations
We simulated the series 100 times, with Houston winning 55. The outcomes were distributed as follows, with a progression from Houston dominance at the top to Chicago dominance at the bottom:

Houston in 4 -- 3%
Houston in 5 -- 19%
Houston in 6 -- 16%
Houston in 7 -- 17%
Chicago in 7 -- 12%
Chicago in 6 -- 16%
Chicago in 5 -- 13%
Chicago in 4 -- 3%

Although a Houston victory in five games was the single most common outcome, we don't believe that it is representative of the matchup. If we had run the simulations 1,000 times instead of 100 times, we're sure that Houston-in-5 would not have been in the top spot.

Houston in seven fits the data much better -- it reflects the fact that Houston won only 55 percent of the time overall, it's the second most common outcome, and it's in the center of the range of the most likely outcomes.

As a result, we chose to provide you with box scores from one series that Houston won in seven games. As it turned out, among all 100 simulations, Houston won Game 1 62 percent of the time and Chicago won Game 2 60 percent of the time, so we chose the first (of two) series that began with this pattern and ended with Houston winning in seven.

Here's one comment about the computer manager's decisions: You'll see that Willie Harris and Geoff Blum were used as pinch-hitters for Chicago in Game 1. That might seem odd given that neither is an especially good hitter. But the computer manager saw how dominant Lidge is against right-handers and thought it would get an edge by sending a lefty to the plate. As it turned out, it worked, as both were able to reach base.