It seems that one thing most baseball observers understand about the post-season is that having two dominant starters is the key to winning the World Series. Two dominant starters, because of the extra days of rest, can start almost half of there team’s games and carry their teammates to championship. We know this because this is what Randy Johnson and Curt Schilling did with the Arizona Diamondbacks in 2001. Spending too much time trying to figure out what happened in other recent post-seasons is, apparently, not worth our time.
The 2001 World Series was extraordinary; and Johnson and Schilling were fantastic throughout that post-season, but it seems like extrapolating from 2001 is rarely questioned. Instead, around this time of year, journalists start seeking to identify who will be this year’s Johnson and Schilling.
I sought to examine this question more closely by looking at which playoff team in each year since 1995, the first year when the current system was used, had the best top two starting pitchers during the regular season, and how successful that team was in the post-season. I did this by using ERA+ as a quick and easy, although not perfect measurement. For each team I took their best starting pitcher’s ERA+ and added it to two times their second best starting pitcher’s ERA+. I did this to weigh the data in favor of having two top line pitchers. For players who split their season between two teams, I looked at ERA+ for the whole season. I did not consider players who pitched well during the regular season but were injured for the post-season. The methodology here is clearly somewhat simple, and other measures might lead to other tandems, but it worked for my goal of seeing how having two great starters in the regular season projected success in the post-season. The data is in the table below.
During this 14 year period, other than the 2001 Diamondbacks, only one other team in the table, the 1995 Atlanta Braves, led by Greg Maddux and Tom Glavine, won the World Series. That strongly suggests that having the top two starting pitchers has no real impact on winning the World Series because two out of 14 is what would be expected if the distribution was random. The team with the top two starting pitchers was eliminated in the first round four times, eliminated in the second round six times and lost the World Series twice.
It is possible to argue that the Johnson/Schilling argument is not that the team with the best two starting pitchers will win, but that if there is a team with two dominant starting pitchers in the playoffs, than they will win. It is not reasonable to expect that the 2006 Dodgers, led by a resurgent Greg Maddux and Derek Lowe, the best one two starting pitching duo in the post-season that year, would have success comparable to Johnson and Schilling who pitched far better throughout 2001 than Maddux and Lowe did in 2006.
If we only consider the seasons where the ERA+ score for the top two starting pitchers exceeded 500, we can look at the question of dominance more closely. There were six years where this was the case. Interestingly, Schilling (22-6 2.98 293Ks) Johnson (21-6, 2.49 372) and Schilling were sixth on this list with regards to ERA+. The 1995 Braves led by Greg Maddux (19-2, 1.63 181Ks) and Tom Glavine (16-7, 3.08, 127) won the World Series, but the 1998 Braves, also led by Maddux (18-9, 2.22, 204) and Glavine (10-6, 2.47, 157), the 1999 Red Sox, which had the most dominant one two starting pitching of any playoff team in this period with Pedro Martinez (23-4, 2.07, 313) and Bret Saberhagen (10-6, 2.95, 81), the 2005 Houston Astros with Roger Clemens (13-8, 1.87, 185) and Andy Pettitte (17-9, 239, 171) and the 2008 Chicago Cubs led by Rich Harden (10-2, 2.07, 181) and Ryan Dempster (17-6, 2.96, 187) did not.
It could be argued that Johnson and Schilling were not just dominant starters, but a certain kind of dominant starter, big pitchers who throw a lot of inning and strike out a lot of batters. This would be a flawed argument because Maddux and Glavine, better pitchers who don’t quite fit this description, carried there team to the championship in 1995, while Roger Clemens and Andy Pettitte, who fit this description better, did not. Moreover, it is so specific as to be meaningless and not replicable.
A more likely critique of this approach is that championships are not won by teams with two pitchers who dominate during the regular season, although Johnson and Schilling did that in 2001, but by teams with two dominant pitchers in the post-season. This is probably true, but both tautological and meaningless. It is very likely that once a World Series winner is determined, it will be possible to point to two starting pitchers who pitched very well in the post-season, but there is little predictive value in this approach.
For some reason, almost eight years after what was admittedly dramatic and exciting World Series, we continue to extrapolate far too much from those 2001 Arizona Diamondbacks. The evidence shows that since 1995, having the two most dominant starters during the regular season has not been a particularly strong indicator of success in the post-season. Good starting pitching, especially at the top of the rotation is important in the post-season, but it is only one of many important strategic components.
The New York Yankees, have been the most dominant team in the post-season since 1995, but have rarely had truly great top of the rotation starters, except, ironically for 1997 when they got knocked out in the first round. Their formula, instead, had been based largely on a lineup that hits from top to bottom and a lights out closer who can enter the game earlier than the 9th inning. Similarly, other teams such as the 2008 Phillies or the Red Sox in 2004 and 2007 relied on great offensive attacks. These seem to be just as legitimate models for post-season success as that of the 2001 Diamondbacks. For some reason, almost eight years after what was admittedly a dramatic and exciting World Series, we continue to read far too much into those 2001 Arizona Diamondbacks and continue to ignore any data that weakens the dominant starter argument.