07 May 2010

Shutdowns and Meltdowns

In search of a count stat for relievers that moves beyond the arbitrary nature of saves, blown saves and holds, David Appleman at FanGraphs devised two stats called Shutdowns (SD) and Meltdowns (MD), based on Win Probability Added. A reliever who posts a WPA of 0.06 or higher receives a shutdown, whereas a reliever who posts a WPA of 0.06 or lower receives a meltdown.

I like it because of its simplicity and its consistency in explaining what relievers are supposed to do, that is shut down the game and prevent meltdowns (if there's a lead), or provide the offense with a chance to win the game (if there isn't). All the baseball stat nerds are writing a piece on this, so I thought I would do some of my own analysis. Here's my version of Erik Manning's "Fun with Shutdowns and Meltdowns" (linked in the first paragraph; data from FanGraphs, of course).

Most Net Shutdowns (SD - MD) since 2008

Name Games WPA WPA/LI SD MD Net SD Net SD% SV  BS
Mariano Rivera 196 12.69 6.01 105 18 87 44.39% 117 5
Francisco Rodriguez 208 6.56 2.72 110 24 86 41.35% 131 20
Joe Nathan 192 9.69 4.93 96 16 80 41.67% 115 14
Francisco Cordero 208 3.06 1.21 103 26 77 37.02% 114 19
Carlos Marmol 233 9.21 4.93 103 29 74 31.76% 27 8


Fewest Net Shutdowns

Name Games WPA WPA/LI SD MD Net SD Net SD% SV  BS
Jason Grilli 159 -1.05 1.4 27 31 -4 -2.52% 2 2
Luis Ayala 163 -5.26 -1.41 33 37 -4 -2.45% 10 11
Cla Meredith 214 -3.38 -1.16 39 42 -3 -1.40% 1 14
Brian Bass 95 -1.2 -1.57 18 20 -2 -2.11% 1 1
Brian Stokes 149 -1.75 -0.54 23 23 0 0.00% 1 5

Best Net Shutdown Rates (SD - MD / Games)

Name Games WPA WPA/LI SD MD Net SD Net SD% SV  BS
Mariano Rivera 196 12.69 6.01 105 18 87 44.39% 117 5
Joe Nathan 192 9.69 4.93 96 16 80 41.67% 115 14
Francisco Rodriguez 208 6.56 2.72 110 24 86 41.35% 131 20
Joakim Soria 170 11.1 4.42 84 16 68 40.00% 91 11
Brian Wilson 166 4.12 1.34 88 25 63 37.95% 91 15

Worst Net Shutdown Rates

Name Games WPA WPA/LI SD MD Net SD Net SD% SV  BS
Jason Grilli 159 -1.05 1.4 27 31 -4 -2.52% 2 2
Luis Ayala 163 -5.26 -1.41 33 37 -4 -2.45% 10 11
Brian Bass 95 -1.2 -1.57 18 20 -2 -2.11% 1 1
Cla Meredith 214 -3.38 -1.16 39 42 -3 -1.40% 1 14
Brian Stokes 149 -1.75 -0.54 23 23 0 0.00% 1 5

Some interesting facts pop up when we look at these new stats. First off, I think it's safe to say that GMs and fantasy owners should stay away from Grilli, Ayala, Bass, Meredith and Stokes. Second, SD pretty much confirms the general consensus about who the greatest relievers in the game happen to be, with Mo Rivera leading a pack that also includes K-Rod, Joakim Soria and Joe Nathan. Carlos Marmol's appearance on this list seems to make the best case for the new stat. We all know he's a very good reliever, but having few chances to save he got very little credit for his efforts. Fret no more, Carlos.

Tango at Inside the Book paraphrases Bill James when he points out that the best new stats are part confirmation, part surprise. Tango shoots for 80% confirmation as a confirmation threshold. Personally I think that seems high, but perhaps I'm just a risk-taker. What kind of confirmation are we looking at with SD and MD?

When I regress Shutdown Rate (SD - MD / Games) as a dependent variable against Save Percentage (SV / SV Opportunities), I see an R^2 of 0.5292, which works out to a correlation (R) of about 73%. Raw SV and SD correlate at about 76%. I'd say that's pretty close to Tango's 80%. Graphs illustrating the relationship between the old and new stats are posted below.

(Photo Credit: CNNSI)

2 comments:

Brian L Cartwright said...

Remember to includes Holds as a success in the Save Rate, as the definition of Blown Saves includes Blown Holds. For example, Cla Meredith with 1 SV and 14 BS looks pretty horrible, because he's not being given credit for his 15 Hld in 2008-09. 16 SV+Hld and 14 BS is still well below the MLB average rate of 85.3%

J-Doug said...

@Brian: Good points. I didn't see holds as a stat in the FanGraphs data I used, but I suppose I could add that retroactively.

Post a Comment

Please Be Civil.