More Fun with FIP

I had intended for this to be an off-day diary for everyone to digest but I didn't get it done in time. It'll have to function as baseball brain food to get you through the day on Friday.

Since my last diary on the topic was well-received, I thought I'd take another look at FIP in the NL. This time, I don't want to be limited by only starting pitchers. I'd rather look at team pitching overall. With that in mind, I calculated the FIP for every pitcher and every team in the NL, using the flat 3.2 constant (see the diary link above for a refresher on FIP if needed). I separated the numbers into two groups: starter FIP (SFIP) and bullpen FIP (BFIP). Then I calculated each starter's and each team's average innings per start and converted that number into outs (i.e., 5.74 turns into 17 outs). Assuming each team must record 27 outs in the course of a game (not true, but close enough), I then took the weighted average of SFIP for outs recorded by the starters and BFIP for outs not recorded by the starters and generated the team FIP (TFIP) for an average game for each starter and each team.

I think that is useful because if you have a terrible bullpen, a guy who can give you 18 outs on average becomes that much more valuable than a guy who gives you 15. Conversely, if you have a dominant bullpen, you can live with a guy only recording 15 sub par outs rather than 18.

Finally, I wanted to have something to compare starters to starters (another diary in the future, perhaps). TFIP is too reliant on the individual's team's bullpen, something he has no control over. Thus, I created a neutral FIP (NFIP) which is TFIP with the bullpen's FIP replaced by the average NL bullpen FIP. This allows comparisons between starters to also include how many outs on average that starter gets per game. One example of what this can do involves Tim Hudson and John Smoltz. Hudson has an FIP of 3.14 against Smoltz's 3.08, but since Hudson records 20 outs in an average start against Smoltz's 19, Hudson actually has a slightly better NFIP. The problem with this is that when you start looking at guys who are almost equally bad (i.e., Rick Vanden Hurk and Byung-Hyun Kim's Marlins numbers), NFIP gives the edge to the guy who doesn't last as long in the game. Obviously, in real life, you'd rather not have to burn your bullpen out every fifth game, even if it's smart to only take five (or less) innings from the bad starter you run out there. I don't know what to do to fix/change that; suggestions?

Anyway, here's the team results, sorted by TFIP, the FIP expected in an average game with that team's starters and bullpen:

San Diego Padres 3.69 3.78 3.72 3.88 1/2/1/1
Los Angeles Dodgers 3.92 3.52 3.77 4.03 2/1/2/2
San Francisco Giants 4.24 4.17 4.21 4.23 3/9/3/3
Milwaukee Brewers 4.42 3.91 4.25 4.35 4/3/4/4
New York Mets 4.61 4.02 4.41 4.48 9/4/5/9
Atlanta Braves 4.58 4.14 4.41 4.44 8/8/6/7
Chicago Cubs 4.66 4.03 4.45 4.51 10/5/7/10
Colorado Rockies 4.72 4.08 4.51 4.55 11/6/8/11
Arizona Diamondbacks 4.56 4.43 4.52 4.45 7/13/9/8
Pittsburgh Pirates 4.47 4.73 4.56 4.39 6/14/10/6
St. Louis Cardinals 4.82 4.19 4.59 4.60 13/10/11/13
Florida Marlins 4.86 4.21 4.60 4.60 15/11/12/15
Cincinnati Reds 4.47 4.91 4.61 4.38 5/15/13/5
Houston Astros 4.75 4.34 4.62 4.57 12/12/14/12
Washington Nationals 5.35 4.13 4.85 4.89 16/7/15/16
Philadelphia Phillies 4.84 4.93 4.87 4.63 14/16/16/14
National League 4.56 4.21 4.44 4.44 --/--/--/--

What's interesting to me is how the Phillies, Reds, Pirates, and even the Diamondbacks collectively pull down the average NL bullpen FIP.  The Mets' bullpen has helped them out a lot, as they have received only the ninth best starts yet still place fifth in the league overall. As maligned as their rotation has been, the Nationals have been hiding a pretty good bullpen down in the nation's capital and Chad Cordero's actually a drag on them (Saul Rivera and Jon Rauch are the workhorses).

Since I just sort of mentioned individual numbers, here's the Brewers' individual bullpen FIP's:

  1. Francisco Cordero, 1.94
  2. Yovani Gallardo, 2.13
  3. Brian Shouse, 3.07
  4. Seth McClung, 3.20
  5. Manny Parra, 3.27
  6. Derrick Turnbow, 3.27
  7. Matt Wise, 4.04
  8. Dave Bush, 4.53
  9. Chris Spurling, 4.61
  10. Elmer Dessens, 4.80
  11. Carlos Villanueva, 4.86
  12. Scott Linebrink, 5.11 (he had 5.69 with the Padres fueled mainly by home runs)
  13. Greg Aquino, 5.29
  14. Jose Capellan, 5.53
  15. Claudio Vargas, 6.20
  16. Grant Balfour, 11.45
Obviously some of those guys have small sample size issues, but there it is. That also reminds me of one final note about the team FIP table. It's impossible for a starting pitcher to make a relief appearance in a game he starts but I didn't take, for example, Dave Bush's relief appearances out of the bullpen FIP when I calculated his TFIP. Technically, I probably should have, but that would take an awful lot of time. Oh well.

I'm not sure this is useful or if it echoes something already out there but I thought it was worth posting. This weekend I might do another average rotation table like in the last diary but this time using NFIP unless the real stats gurus here rip that stat to shreds. :)