Monday, February 18, 2008

Optimizing the Starting Rotation

There is going to be a lot of talk this spring on how the Starting Rotation should be ordered. Besides, after Justin Verlander, the Tigers currently have four players who could be anywhere from great to bad this season. Jeremy Bonderman was lights out for much of the first half of 2007, and was on the rise until the elbow problems rendered him ineffective last season. Dontrelle Willis has been among the game's best in his career, but has declined significantly the past two seasons. Nate Robertson and Kenny Rogers also pose serious question marks. The likely rotation will likely have Verlander pitching on opening day, with Rogers pitching between Verlander and Bonderman

A lot of the rotation talk will focus on the order of the rotation, and who deserves to go in which spot. The best way to tackle the rotation is to look at the match-ups of the opponents. Below are the OPS vs. LHP - OPS vs. RHP for each American League team in 2007.

Toronto 0.133
Tampa Bay 0.065
Oakland 0.049
Seattle 0.043
Cleveland 0.037
Detroit 0.034
Kansas City 0.028
Chicago 0.012
Texas 0.008
Baltimore 0.006
Los Angeles Angels -0.002
Boston -0.008
Minnesota -0.014
New York -0.055

If we assume the tendencies stay the same (an assumption that may not hold for overhauled lineups like Minnesota, Chicago, and Oakland), this table shows that you want your right-handed pitchers facing teams like Toronto and Tampa Bay, while saving the lefties for Boston, Minnesota, and New York.

Looking at the Tigers first 20 opponents, the starters would face the following teams:
1
Kansas City
Chicago
Chicago
Cleveland
2
Kansas City
Boston
Chicago
Toronto
3
Kansas City
Boston
Minnesota
Toronto
4
Chicago
Boston
Minnesota
Toronto
5
Chicago
Chicago
Cleveland
Toronto

Next I averaged the OPS differences for each opponent, putting Kenny Rogers between Jeremy Bonderman and Justin Verlander, and assigning remaining highest split to Dontrelle Willis, who has a bigger split for his career than Nate Robertson. The result was:

Kenny Rogers
Justin Verlander
Nate Robertson
Dontrelle Willis
Jeremy Bonderman

Can you imagine the outcry if Justin Verlander was not the opening day starter? Indeed, this seems to make the most sense. The opening day starter will appear to miss the right-handed heavy Blue Jays. It would be best if that pitcher is left handed. Surely, there must be a way to justify the decision to make Justin Verlander the Opening Day starting pitcher.

There is.

The above analysis ignores an all too common phenomenon that occurs in April: postponements. When looking through parks where the Tigers play outdoor games, I found roughly 10% of games April 20 or sooner are rained out. I made some quick assumptions to make the calculations a bit easier:

  • Only the last game of a series getting rained out affects the rotation. I am assuming that games earlier in the series will be made up the next day.
  • Only one game can be lost per series. This one is a bit precarious, especially after the blizzard in Cleveland last year. This cuts down scenarios significantly leaving only six games to worry about.
  • The Tigers five starters will stay healthy and pitch their turn throughout the course of the first 20 games.
  • Only one or two of these games will be postponed. Scenarios of three or more were deemed unlikely, and their probability was close to zero.
Since there were six games that could affect the rotation, each at 10% likelihood of raining out (or snowing out or freezing out), the odds of pitchers facing their no-rain-out opponent is 40%. Here is the weighted average OPS difference of each rotation slot:

1. 0.160
2. 0.170
3. 0.158
4. 0.159
5. 0.140

According to this, the right-handed pitchers should go 1-2, with the lefties going 3-5. 1,3, and 4 are interchangeable. I am going to move one of the right-handers from 1 to 4 to split them up, and put Rogers in between. Bonderman has a higher platoon split than Verlander, so he will go in the #2 spot. Dontrelle Willis's platoon split will put him into the #5 spot, leaving the rotation as:

1. Nate Robertson
2. Jeremy Bonderman
3. Kenny Rogers
4. Justin Verlander
5. Dontrelle Willis

I don't think that rotation would be too popular, even if this study shows it to be the most effective. There were so many assumptions and shortcuts made here, that this rotation may not even be the best. Jim Leyland will still have to consider home/away splits, and other factors that come up. Still, while it is a nice reward for your ace, it is not always the best strategy to pitch your best pitcher on opening day.

No comments: