Friday, October 12, 2007

NL Rotations by FIP for 2007

EDIT: Upon further review, I don't like the way I originally came about these numbers. For some reason, I had simply averaged each team's FIP in each spot for the league numbers. A smarter way to do it would be to treat the entire league as one giant rotation and determine the top 20% of starts (generally around 518 for the NL and 454 for the AL) for spot #1, etc. It makes no sense to penalize the league average rotation because the top 15 pitchers are clustered on, say, 8 teams. This tends to lower the value for #1 starters and raise the value for #5 starters while leaving the middle guys generally unchanged. I've changed the numbers in the tables to reflect the new method. The "average" rotations at the end are the same as before. Sorry for the mistake.

There are a number of different things to get out of the way before the meat and potatoes of this post. Most importantly, the idea for this fun little exercise came from the article "How Good Is Your #4 Starter?" by Jeff Sackmann and the follow-up "More Fun With Rotation Numbers." He used ERA to calculate his numbers and I will have another post soon utilizing ERA. For this post, however, I am using FIP, or Fielding Independent Pitching. FIP attempts to measure a pitcher's worth through the outcomes for which he is directly responsible during a game: home runs, hit batsmen, bases on balls, and strikeouts. In this way, defense largely is taken out of the equation. Thus, FIP is not affected in the same way ERA would be for a pitcher if he had eight David Ortiz's on the field with him.

The formula I use is (HR*13+(BB+HBP)*3-K*2)/IP + 3.2 = FIP. In actuality, the constant should be slightly larger than 3.2, depending on the league, but since I don't know exactly how much larger, I used 3.2 for simplicity's sake. Since it affects everyone in the NL, no one is given an unfair advantage.

In order to determine the numbers for each team's rotation spots, I figured the ideal rotation would give a team 33 starts by their #1 and #2 pitchers and 32 starts by the other three (San Diego and Colorado both had an extra start by their #3 starter for my calculation). Using this idea, I took the weighted average of the team's top 33 starts by FIP to determine the FIP for rotation spot #1, and so on for the other four spots.

Example, to determine the #1 spot for the Brewers, I would see that Yovani Gallardo had 17 starts with a 3.49 FIP, Manny Parra had 2 starts with 3.87 FIP and Ben Sheets had 24 starts with a 4.07 FIP. Sheets' starts are split between the #1 and #2 spot and you get
  • #1 FIP = (17*3.49+2*3.87+14*4.07)/33 = 3.76
Therefore the Brewers got a 3.76 FIP from their "#1 spot."

The SFIP column stands for "Starter FIP" for each team, i.e., the FIP put up only by starting pitchers in games they started. The NL averages were computed by simply taking the average of each team's FIP in each spot. The NL FIP was computed by applying the FIP formula to the raw numbers of HR, HBP, BB, K and IP throughout the league. The STDEV column is the standard deviation of the team's five rotation spots. The smaller the number, the closer together the five spots are and the more "even" a team's rotation is. This is fallible, in the sense that a team with a great ace will appear to have an uneven rotation even if the #4 and #5 starters really aren't that bad.

Finally, as Jeff said in his follow-up article:
These calculations don't hold the key for any breakthrough new approach to roster construction, but they do illustrate some of the ways in which good (or lucky) teams are different from bad ones.
Now that the explanation is out of the way, on to the table!

TeamSFIP#1#2#3#4#5STDEV
San Diego Padres3.822.803.383.544.375.671.11
Los Angeles Dodgers4.223.593.904.024.375.550.76
San Francisco Giants4.283.553.774.434.815.030.64
Milwaukee Brewers4.363.764.134.374.515.080.49
Chicago Cubs4.544.064.264.524.695.320.48
Pittsburgh Pirates4.543.914.184.524.765.780.72
Cincinnati Reds4.543.674.454.534.556.290.96
Arizona Diamondbacks4.583.194.164.745.346.021.09
Atlanta Braves4.593.183.425.175.406.701.47
New York Mets4.593.804.294.724.825.540.65
Colorado Rockies4.673.964.324.645.075.540.62
Houston Astros4.793.574.215.345.475.610.90
Philadelphia Phillies4.863.764.584.985.206.110.86
St.Louis Cardinals4.913.704.764.885.136.420.97
Florida Marlins5.124.075.055.195.336.170.75
Washington Nationals5.414.455.155.525.786.430.74
NL4.603.484.174.665.04
6.040.96

I guess the old adage "pitching wins championships" didn't hold especially true in the NL this year as Chicago and Arizona were the only teams to be much above league average in FIP from their starters and they were still very close to the mean. Another thing I noticed that kind of surprised me was how mediocre Florida's starters were. Granted a team losing 90 games generally won't have very good starters in the first place, but I was shocked to find out they only got 37 starts all season from a pitcher winding up with an ERA below 5.00 as a starter (27 from Sergio Mitre, 6 from Anibal Sanchez and 4 from Ricky Nolasco). Furthermore, they only got 42 starts from a pitcher with an FIP under 5.00. That's perversely impressive.

Regardless, simply seeing the numbers might not strike your fancy. Let's look at which starters came closest to each rotation spot's average, in order to give some context.

First, here are the starters that were "aces" (3.69 or lower FIP) in 20 or more starts:
  • Jake Peavy, 34 starts, 2.80 FIP
  • John Smoltz, 32, 3.17
  • Brandon Webb, 34, 3.20
  • Chris Young, 30, 3.39
  • Tim Hudson, 34, 3.42
  • Greg Maddux, 34, 3.54
  • Roy Oswalt, 32, 3.55
  • Brad Penny, 33, 3.59
  • Tim Lincecum, 24, 3.59
  • Aaron Harang, 34, 3.67
No real surprises, though it must be unsettling for batters in the NL West to see that rookie Tim Lincecum shows up as an ace already (note: Yovani Gallardo made 17 starts with a 3.49 FIP, so he was good as well).

Let's see who fits the mold as a #2 starter (~4.25 FIP):
  • Chris Capuano, 25 starts, 4.16 FIP
  • Tom Gorzelanny, 32, 4.20
  • Rich Hill, 32, 4.28
  • Oliver Perez, 28, 4.30
It's tough for a Brewers fan to see that Capuano was actually better than an average #2 starter by FIP, but there it is. His struggles could very possibly be a combination of bad luck and poor defense behind him inflating his numbers. Regardless, the other names on the list are not very surprising as all three pitchers had good seasons.

Our #3 starters (~4.69 FIP):
  • Paul Maholm, 29 starts, 4.56 FIP (everyone between Maholm and Davis had less than 20 starts)
  • Doug Davis, 33, 4.68 FIP
  • Micah Owings, 27, 4.78
  • Braden Looper, 30, 4.79
Doug Davis nailed the middle-of-the-rotation niche and it makes sense. He eats innings and nibbles too much on the corners to be good, but he gets the job done most nights.

Alright, we're headed to the back half of the rotation. The #4 starters (~4.98 FIP):
  • Jason Marquis, 33 starts, 4.94 FIP
  • Jamie Moyer, 32, 5.00
  • Anthony Reyes, 20, 5.02
  • Dontrelle Willis, 35, 5.09
Jason Marquis has always been kind of middling. Moyer gave you about what you could expect from a 44-year-old soft-tosser. Reyes is young and struggling, while Willis continued sliding further from the dominance he showed in 2005. Is he sick of playing in Florida or is his delivery no longer fooling hitters the way it did?

Finally, we've reached the land of the damned--er, I mean the #5 starters (~5.83 FIP):
  • Livan Hernandez, 33 starts, 5.73 FIP
  • Byung-Hyun Kim, 22, 5.76
  • Adam Eaton, 30, 5.93
  • Mike Bacsik, 20, 6.38
Unsurprisingly, not many pitchers this bad made more than 20 starts. Other than Bacsik, only Joel Hanrahan (11) and Jo-Jo Reyes (10) reached double-digits for starts with an FIP over 6.00. Tim Stauffer had the worst FIP, putting up a 12.85 in 7 2/3 IP in his two starts for the Padres. Five HR, 6 BB, 6 K and an HBP will do that to a pitcher. Wilfredo Ledezma and J.A. Happ also scored above 10.

To recap, the National League average rotation would look like this:
  • Aaron Harang
  • Rich Hill
  • Doug Davis
  • Jamie Moyer
  • Byung-Hyun Kim/Adam Eaton
The Mets' rotation was pretty close to league average across the board, as well.

I'll be putting up a similar post using ERA soon. I think this sort of number-crunching is interesting, if not especially useful.

No comments: