The 2007 postseason has produced one of the more intriguing World Series matchups of recent years: the hot-hitting Boston Red Sox, fresh off yet another dramatic ALCS comeback, against the Colorado Rockies, winners of 21 of their last 22, but hardly "fresh" after a lengthy layoff following their NLCS sweep.
We previously employed our Diamond Mind simulation software to project the results of the four Division Series and two League Championship Series. We projected the Red Sox and Yankees as ALDS winners, and the Rockies and Cubs to win in the NLDS, but with one caveat -- winning the first game of a best-of-five series gave a team such an advantage that it shifted the odds in favor of the underdogs if they drew first blood, particularly given the pitching matchups. For example, the Yankees and the Cubs dropped their openers and ended up getting eliminated.
Our LCS projections for ESPN.com were even more on target. Not only did they indicate that the Rockies and Red Sox would win, but that the Rockies would do so fairly easily, while the Red Sox probably would need seven games to defeat the Indians.
Before revealing our World Series simulation results, here is a brief recap of our methodology:
• Before the postseason got under way, we updated our projections and ratings for each player based on his 2007 regular-season performance.
• Before running our simulations for each series, based on the information available at the time and using our best judgment, we set the teams' starting rotations, batting orders, bullpen and bench roles.
• Then we played each series 1,000 times.
As we noted in our previous articles for ESPN.com, luck plays an even greater role in a short series than it does during the regular season, which itself can be rather unpredictable. With that in mind, the results of our simulations project the Red Sox to reprise their 2004 championship and defeat the Rockies. In fact, Boston won over 70 percent of our 1,000 series simulations, the largest winning margin of any of our 2007 postseason projections:
Over our 1,000 series simulations, the Red Sox averaged 4.81 runs per game to just 4.09 for the Rockies, which is equivalent to 779 and 663 runs, respectively, over a 162-game season. A 779-663 run differential equates to a record of 94-68, a .580 winning percentage, using Bill James' Pythagorean formula for projecting team wins based on runs scored. Sure enough, Boston's aggregate record against Colorado over1,000 simulated series was 3,341-2,417 (.580).
There are, however, a number of features presented by this series capable of confounding any attempt to project or predict the outcome:
The Designated Hitter
In Colorado, there won't be a DH, which will present Red Sox manager Terry Francona with a dilemma: send Kevin Youkilis (who banged out 14 hits in the LCS against Cleveland) to the bench and play David Ortiz (with his bad knee and weak defense) at first base? Or use Ortiz and Youkilis in a platoon? If Ortiz plays first base, could Youkilis be shifted to third, weakening another position defensively and benching Mike Lowell (and his team-leading 120 RBIs during the regular season)? We chose to start Ortiz at first base in all three games in Colorado, but then tried a few other combinations as well. However, it made little difference to the outcome. For example, when we started Youkilis instead of Ortiz, Boston actually increased its series edge from 703-297 to 707-293.
The Rockies have won a remarkable 21 of their last 22 games, but will not have played for nine days when the first pitch is thrown on Wednesday. Our simulation does not factor in any effect for that unprecedented layoff, nor is the projected performance of players or teams in the simulation influenced by the fact that they may be in the midst of a streak or slump.
Our Diamond Mind simulation software does include weather effects, so seasonably chilly weather in Boston and Denver did affect the results of our series simulations. Nevertheless, unseasonably cold (or warm) weather, including gusty wind and snow, could affect the series in a number of ways, such as causing postponements that might allow Josh Beckett to start three series games on full rest.
As good as Colorado's pitching has been in its Cinderella run to the World Series, our simulations suggest that the Red Sox will turn giant-killer Josh Fogg (6.43 ERA in our simulations) and flame-throwing youngster Ubaldo Jimenez (5.19) back into pumpkins. While Ortiz and Manny Ramirez both continued their postseason assault against the Rockies in our simulations, it was Lowell's .304 batting average that led the way for Boston, with substantial contributions from Dustin Pedroia (.292) and Jacoby Ellsbury (.286).
In our ALCS simulations, we projected that one key to a Red Sox victory might be that their pitching would control Travis Hafner -- he averaged .227 in our simulations, and he hit just .148 in the series. In our World Series simulations, Matt Holliday (.292) and Todd Helton (.286) were productive offensively for the Rockies, but Boston's pitching managed to hold Brad Hawpe (.250), Troy Tulowitzki (.238) and Kazuo Matsui (.280 OBP) in check.
A short series often will produce standout performances, sometimes from unlikely sources (like David Eckstein in last year's World Series). Imagine if a series actually were replayed 1,000 times, which is more postseason series than have been played throughout the entire history of Major League Baseball: the odds of something extraordinary occurring would increase significantly. Here are some of the more spectacular performances from our 1,000 series simulations:
• No pitcher on either team threw two shutouts in even a single series out of 1,000. Interestingly, though, the pitcher to come closest to the feat wasn't Josh Beckett, but veteran Curt Schilling, who in one of our simulations had two complete-game victories in which he allowed just one run total.
• Ortiz belted five homers in four different series simulations.
• Mike Lowell had one 18- and two 16-hit series.
• Willy Taveras, perhaps as unlikely a potential series MVP as Eckstein, had two scintillating series, tallying 19 hits in one and 16 hits in another.
If the Rockies were to defeat the Red Sox, it may be Holliday, however, who carries them to victory. Over our 1,000 series simulations, Holliday had one series with 17 hits and three with 16; he hit five home runs in five different series; and he registered one 15-, one 14- and two 13-RBI series.
Winning Game 1 of a short series is such an advantage that it frequently will shift the odds in favor of an underdog. In our simulations, Beckett and the Red Sox bested Jeff Francis and the Rockies in Game 1 59.5 percent of the time. Simulating the series 1,000 times after a Game 1 Boston win put the Red Sox in a commanding position, with an 81.4-percent likelihood of winning and a 15-percent chance of a sweep:
On the other hand, if Francis and the Rockies best Beckett and the Red Sox in Game 1, the odds get much better for the Rockies, with the Series essentially becoming a toss-up. Still, Boston enters the series with such a substantial projected edge that the Red Sox remain very slight favorites, winning the series 50.9 percent of the time even if they drop the opener, compared to the Rockies winning it 49.1 percent of the time:
If the Rockies are to overcome the odds and win the World Series in their first attempt, the key might be to overcome the postseason dominance of Josh Beckett and the rust of a eight-day layoff to win Game 1. The bottom line, however, is that Imagine Sports projects another championship for Red Sox Nation -- their second in four years -- with their victory most likely coming at Fenway Park in six or seven games but with a decent chance that Boston could take it in four or five games.
Hold the presses!
After we'd run and analyzed the results of our simulations, the Red Sox and Rockies each announced changes to their playoff rosters and pitching rotations: Boston removed Tim Wakefield from its World Series roster, with Jon Lester the likely Game 4 starter, and Colorado activated Aaron Cook, who is slated to start Game 4 in place of Franklin Morales.
We ran our simulations again to reflect these changes, and the advantage was even more decidedly in favor of the Red Sox. With Game 4 being a matchup of Lester against Cook instead of Wakefield against Morales, Boston's edge over 1,000 series simulations increased from 703-297 to 720-280 and raised the likelihood of Boston sweeping or winning in five games from 28.7 percent to 32.2 percent. Furthermore, the Series changes from a virtual toss-up if the Rockies win Game 1 (509-491, Boston) to a more significant Red Sox advantage (542-458) in that scenario.
The last-minute changes to the rosters and rotations appear to have increased Boston's advantage, possibly because of the Rockies' dropping Morales in favor of Cook. Ironically, Morales -- now in the Rockies' bullpen -- threw the only no-hitter in our original simulation runs.
As a result, we still see the Red Sox winning in six games, but there is now a more substantial possibility that Boston will take it in four or five.