ESPN has fan voting for the best at each position in baseball in the last twenty years. Finalists for right-handed pitcher were Greg Maddux, Roger Clemens and Nolan Ryan. I intuitively pick Greg Maddux because of his dominance in 1994-95. Bill James considers Clemens to be greater because, even though Maddux was superior inning-for-inning in his prime (his ERA compared to leauge average was historically great, as was that of Pedro Martinez during his prime) Clemens pitched more innings.
When the model is comparison to league average, Maddux's ERA advantage looks impressive. But when our model is comparison with a replacement-level pitcher, Maddux's advantage over Clemens is mitigated. On this model, Clemens' higher innings total also looks more impressive, because he is taking innings away from a less capable pitcher- the replacement-level pitcher, rather than the league average.
But is he? By starting, both pitchers relieve their teams of the need to start their top long-reliever or minor-league starter, which in turn allows their top long-reliever to take some unimportant innings away from their worst long reliever, or a minor leaguer. By getting through five innings, they also prevent their teams from needing to bring long-relief in.
After that, however, the pitchers they are replacing get better and better. How often is the soft underbelly of your bullpen going to see action during a Maddux or Clemens start, unless you have a huge lead? I think Maddux had one low-quality start during the entire strike-shortened 1994 season. If Maddux goes 6.2 innings and giving his closer, setup man and top middle-reliever or specialist a lead, whereas Clemens goes 7.1 before turning it over to his setup man and closer, we now need to compare the two men's innings not with a replacement-level player, but with pitchers who on the average team give up far fewer than an average number of earned runs per nine innings; in other words, pitchers who will accentuate Maddux's ERA advantage and mitigate Clemens's innings pitched advantage.
On the other other hand there is a mitigating factor to this analysis. Because Maddux has required .2 innings more of top-middle-reliever/specialist - setup man - closer, those pitchers are less available to future starters, who then ultimately require more from lower-quality pitchers. The reason this does not mitigate all the way, however, is because managers usually make pitching decisions based on the logic of a particular game rather than a concern for bullpen wear-and-tear. You bring your best relievers into a close game with a lead, because you never know when you'll have another slim lead; indeed, sometimes you bring them in just to get work when they have not pitched in a long time. A long reliever (the closest thing to a replacement-level player) by definition is mostly going to come in when a starter gets shelled, which is entirely a situational decision and not a function of whether the ace of the staff got you into the ninth four days ago.
I'll give Bill James that argument, but their is more to the tale. For over all innings pitched, Maddux has almost 100 more (5008.1 to 4916.2). Maddux also pitched 1 less season than Clemens. He could argue Clemen's partial season in 07 (99 inn.) shouldn't count, but then neither should 86 (31 inn) count for Greg. Maddux also has over 30 more starts than Clemens as well. So over a season Greg would give you more innings and more starts. So that in itself should mitigate James's argument to begin with.
Posted by: Ben Upchurch | November 19, 2009 at 01:18 PM