How good is a #1, #2, #3, #4, #5 starting pitcher? The NL remix!

This was inspired by a piece over on Royals Review, which made an effort to establish typical values for ERA, etc. returned by starters #1 through #5. It got me to do something similar, albeit with a slightly different approach - and also for the National League, as for obvious reasons, Royals Reviews did it for the AL. They also only included the figures for the top five starters on each time, by number of starts, which excluded a number, generally quite significant, of games. For example, in Arizona, we had 21 games by starters outside the top five (Webb, Haren, Johnson, Davis, Owings). I wanted something that would also take those into account.

He's my methodology. 2,592 starts were made in the National League this year, by 165 different pitchers. Their ERAs as starters ranged from three pitchers who each made one start and allowed no earned runs, all the way down to the 36.00 returned by Adam Pettyjohn of Cincinnati, who allowed eight earned runs in the two innings he lasted during his sole appearance. Here's the list of all 165, ordered by ERA. Initially, my thought was to divide those 2,592 starts into five bands of 518. However, I don't feel that Mark Mulder's 0.1 IP outing was worth the same weight as each one of CC Sabathia's seven complete games.

So, instead, I used the 14,911 innings pitched by starters, and broke that down into five bands of 2,982 innings. Starting at the top of the ERA list, I worked my way down the list, summing up the innings thrown by the starters, until I got the end of that band.

Mulder     0.2  -> Total 0.2
Guzman     2.0           2.2
Valdez     2.0           4.2
Sabathia 130.2         135.1
Harden    71.0         206.1
Gallardo  24.0         230.1
  ...      ...          ...
Volquez  194.1        2615.1
Webb     226.2        2842.0
Haren    216.0        3058.0

So, the 2,982nd best inning was thrown by Dan Haren, whose ERA was 3.33. That or better, puts you in the top 20% of starters. The process was then repeated: the next slice covers innings 2,983-5,965, and goes down to the Padres' Chris Young. The third slice brings us to Yusmeiro Petit, and the fourth to Barry Zito. Everyone below that falls into the #5 category, or the mop-up bucket. Now, obviously, this is based purely on ERA, with all the imperfections it contains [park factors and defense are excluded]. It's also a sliding scale: nothing makes a pitcher with a 3.34 ERA - thus a "#2" - significantly worse than one with a 3.32, who falls into the top group. However, as a guide as to how pitchers should be considered "aces", it's a reasonable starting point. Here's the full breakdown for the NL:

It's no surprise that the last group includes a lot more pitchers than the others, as it includes all the replacement level arms, used quickly then discarded. The top category also contains a number of names who threw too few innings to be generally considered "aces" - only five of the top twenty pitchers by ERA as a starter, threw more than a hundred innings [Sabathia, Santana, Lincecum, Peavy and Dempster]. For Arizona, Webb is a #1; Haren is on the #1-#2 cusp; Johnson is a #2; Davis a #3; and only Owings is lower, being a #5. That does back up the general view that our rotation was a strength.

The final column is the expected number of wins and losses for a pitcher in each group, based on 33 starts. This is based off the average records of each group. For example, the Group #1 pitchers had a cumulative record of 225-112 in 479 starts, which breaks down to fifteen or sixteen wins and eight losses, numbers with which you would be pretty happy, from any pitcher over a season. Note in particular the gap between a  #1 and a #2 in wins, is almost as much as that between a #2 and a #4. And if your #3 is better than .500, you're ahead of the curve there.

However, I also see that the gap at the bottom of the chart is equally as large as the one at the top. This means that a team will get just as much benefit - about three wins - by upgrading one slot in their rotation from a #5 pitcher to a #4 one, as by going from a #2 to a #1. The former approach will, needless to say, be a good deal more inexpensive - the likes of Tim Redding (10-11, 4.95 ERA for Washington) are a lot cheaper than someone you might actually have heard of. Equally, improving a team's defense might be more cost-effective than better pitching, with the added benefit that a rising tide of good glovework floats all boats, helping out every day, rather than every fifth game.

[A quick aside. While looking at those stats in my spreadsheet, I could also see which teams sent most and least starters to the mound. Arizona used eight in total, which is well below average - only Philadelphia had fewer, with seven, getting 145 games from their top five starters. The highest number went to San Diego, who tried fourteen different starters, just ahead of Pittsburgh's thirteen. I don't think it's too much of a coincidence that these teams had two of the three worst records in the league.]

For amusement, I also ran the same analysis over in the American League, using the 13,287.2 innings pitched by 145 different starters there. The methodology was the same, and the chart below shows the results, with the addition of an additional column - that shows the results obtained by Royals Review from their analysis

It's interesting to note how close the results are for my method - based on innings and including all 145 pitchers - and the RR approach, which took only 70 pitchers (the top five, measured by games started, for each AL team). It looks like the arms RR eliminated don't affect the final figures too much, because they a) throw only a small fraction of the innings, and b) are not all that much worse. The 70 most active pitchers made 14 or more starts, threw 85% of all AL innings and had an ERA of 4.30. The remaining 75 pitchers, threw 15% of the innings, with an ERA of 5.42. That's worse, but not as bad as I'd expect - I would have predicted something well into the #5 group rather than a poor #4.