How will the free-agent spending splurge this past winter play out in 2007? Will Alfonso Soriano, Carlos Lee, Gary Matthews Jr., Gil Meche, Juan Pierre and others justify their big contracts, or will they prove to be multiyear financial millstones for their teams? When and where (if at all) will Roger Clemens pitch in 2007? Will the Cinderella team of 2006, the Tigers, go the way of the White Sox, who took the slipper in 2005, and the Red Sox, who ended the Curse in 2004, and fall short of repeating last year's success?
As final roster decisions are made and Opening Day approaches, the best-laid plans of major league teams are subjected to the scrutiny of commentators, analysts, fantasy addicts and everyday fans, who offer up a varying mix of sabermetrics, wishful thinking and fatalism with their predictions for the coming season.
This is our 10th season at Diamond Mind trying our hand at the preseason predictions game. Our approach has produced some reasonably prescient (if, for some teams, sobering) forecasts. For example, in 2006 our system correctly identified five of the six division winners, and we were only an NL West tiebreaker away from a clean sweep. We surveyed the relative success of prognosticators across the nation in 2006 Predictions -- Keeping Score.)
Before revealing our final standings for the 2007 season, here's an overview of how we produced them.
We began by projecting the 2007 performance of over 1,800 players contending for major league jobs. To do this, we took their major and minor league stats for the past three seasons, adjusted for factors such as the level of competition (majors, Japan, AAA, AA, etc) and offense in a league, park effects, and whether the DH rule was in use. Then, giving greater weight to more recent seasons, performances at higher levels, and seasons with more playing time, and adjusting for age, we projected their performance into the league and park where they will be competing in the coming year.
We didn't merely project the aggregate "headline" stats for each player, but their left/right splits as well. We also assigned ratings for skills such as bunting, baserunning, defensive range and throwing.
After all players have been rated, we set up a manager profile for each team, consisting of a starting rotation, bullpen assignments, projected lineups against right- and left-handed starters, and a positional depth chart. Once these profiles were in place for every team, we played out the season using our Diamond Mind Baseball simulation game. The computer manager, guided by our manager profile, makes decisions about starting pitchers, lineups, substitutions, and pitch-by-pitch tactics. Because luck can play a major role in any single season for players and teams (both in real-life and our simulations), we ran the season 200 times and averaged the results.
Factors and Non-Factors
To provide you with a bit more insight into the process, factors that we do and don't take into account in our projections include the following:
• We take past performance in the major and minor leagues, including the Japanese leagues, into account, but not performance in college or high school, independent leagues, winter leagues, spring training, or other foreign leagues.
• We go beyond aggregate projected hitting and pitching stats, taking left/right splits and defense, running, bunting and other skills into account. If a team is imbalanced in some material way, that will show up in its results.
• We take injuries into account in two ways when we project player performance: by discounting past performance that may have been adversely affected by a player attempting to play through injury, and by taking playing time away from a player we know is beginning the season with an injury. However, we don't attempt to project the likelihood of a player getting injured over the course of a season. Every player has the same chance of injury in our season simulations.
• We don't attempt to emulate the tactics of specific managers. The computer manager manages each team according to the strengths and weaknesses of its roster. Nor do we attempt to rate managerial ability. We just haven't seen the evidence over the years to indicate that particular managers' teams consistently over- (or under-) perform their projected results.
• We don't factor preseason hype (so-and-so has added a new pitch, is in the best shape of his career, has a more focused and determined attitude, etc) into player projections.
• Strength of schedule necessarily comes into play in our methodology, so teams with relatively weaker divisional opponents or interleague schedules have an advantage. Player matchups come into play also, so, for example, a right-handed starter with extreme left-right splits pitching in a division loaded with left-handed hitting is going to have a relatively tougher time of it.
It is important not to take our team and individual player projections too literally. Many of the most noteworthy events of a baseball season -- the breakout performances and fantastic flops by individual players, the teams for which everything goes right or everything goes wrong, the crippling injuries -- are things that might occur in individual seasons that we've simulated, but are unlikely to appear in the averaged results for 100 seasons. There is a large element of luck involved in baseball, and any given season, real or simulated, will produce a larger spread of runs and wins than are reflected in our projected standings. That's because averaging the results of 100 or 200 simulated seasons will tend to smooth out the catching-lightning-in-a-bottle features of any real baseball season.
Postseason Qualifiers and Other General Observations
The trend toward increasing parity that we noted in our Projected Standings for the 2006 Season looks set to continue for 2007. In our 2006 simulations, for the first time just three of the 30 major league teams failed to reach the postseason in at least one simulated season. For 2007 that number has dropped to two, with the Tampa Bay Devil Rays and Kansas City Royals the only teams to be shut out.
In 2004 10 teams (five in the AL and five in the NL) managed 90 or more wins. In 2005 that number dropped to seven (five in the AL and two in the NL). Our 2006 projected standings had just four teams reaching 90 wins (projected/actual wins): the New York Yankees (93/96), Minnesota Twins (90/96), Oakland A's (96/93) and St. Louis Cardinals (95/83), although six actually achieved it, the other three being the Tigers (79/95), White Sox (86/90) and New York Mets (87/97). For 2007 the number of teams projected to win at least 90 has dropped even further to just three: the Cleveland Indians and Los Angeles Angels (91 each) and the Yankees (96).
Last year we projected the runs scored for all National League teams to fall within the fairly narrow range of 705 to 818 runs. The actual spread was a bit wider, from a low of 691 for the Pittsburgh Pirates to a high of 865 for the Philadelphia Phillies (not dissimilar to 2007, with the Phillies projected to score a league-high 852 runs and the Pirates bringing up the rear with just 708). More noteworthy for 2007, however, is the fact that no team in the National League is projected to win more than 88 games (the San Diego Padres, with the next best total just 85 by the Phillies and Cardinals) or fewer than the 72 (the Pirates).
This doesn't mean there won't be a 90-game winner in the NL this year. The real season will be played only once, and it's quite possible that two or three teams will find a way to reach that threshold. However, our simulation results suggest that no NL team has put together a roster strong enough to make 90-plus wins a sure thing or even a high probability.
As far as races to qualify for the postseason, the 2006 season generally followed our projections in the American League, with both the wild card, and the only real divisional race, coming out of the Central. For 2007 the only close divisional race in our AL projections again is the Central, and the Tigers again eke out the wild card by the barest of margins.
The 2007 National League divisional races look to be closer, with most teams at least on the fringe of postseason contention (division or wild card) deep into the season. However, the lack of dominant teams in the NL means that any team that manages to put together a big season may waltz home in their division, as the Mets did in 2006, and should two teams manage the feat in a single division, even the wild-card race could turn into a runaway.