Bullpen shutdowns and meltdowns

Demoting David
A roster move on the horizon?

In David Appelman’s never-ending quest to take over the (baseball statistical) world, FanGraphs introduced two new metrics yesterday called Shutdowns and Meltdowns. It’s a simple idea involving relievers and win probability. A Shutdown (SD) is when a reliever accumulates at least 0.06 WPA in an outing, while a Meltdown (MD) is when he accumulates no more than -0.06 WPA. In English, they refer to when a guy comes out of the bullpen and increases or decreases his team’s chances of winning by at least six percent. Got it? Simple enough.

Last season, the Yankees’ bullpen recorded 141 SD, the third most in the game behind only the Mariners (159) and Dodgers (155). The chart to the right has the breakdown by pitcher; the missing ten belong to guys like Edwar Ramirez, Jose Veras, and Brett Tomko, who weren’t worth the time to include. I went with the core guys only. On the flip side, they had the sixth fewest MD with 62, and their +79 SD-MD differential was second only to Seattle, who had a staggering +90 mark. I don’t think it’s possible to win 103 games with a shaky bullpen, so these numbers really aren’t much of a surprise.

Fast-forward to this season, and we see something a little different. Yankee relievers have racked up just 17 SD (tied for 21st) but only eight MD (tied for the fewest), good for a +9 SD-MD differential that’s tied for the 12th best in the game. The chart down and to the left has the full breakdown. This might make it sound like the Yankee bullpen hasn’t been getting the job done, and while that’s true to a certain extent, it doesn’t tell the whole story.

In the 28 games they’ve played, the starting pitcher has completed at least six innings 17 times (60.7%) and at least seven innings a dozen times (42.9%). That’s a ton. Add in the fact that ten of their 19 wins came by at least four runs, and you’re talking about very few really high leverage spots where a reliever could generate as much a 0.06 WPA without throwing multiple innings. Just as an example, let’s look at Wednesday’s game.

Sergio Mitre had just surrendered a two-run homer to Ty Wigginton in the 8th inning to bring the Orioles to within three runs. Damaso Marte was summoned to face the lefty Nick Markakis with one out, and at the time the Yankees had a 94.6% chance of winning the game. Five pitches later, Markakis had struck out, increasing the Yanks’ chances of victory to 96.2%. Marte gets credit for that 1.6% increase, or 0.016 WPA. It was the late innings and technically a save situation, but Marte could only muster about a quarter of a SD. That’s what’s going on this year with Yankee relievers, the opportunity for Shutdown appearances just haven’t been there. Frankly, I like it this way.

I don’t believe SD and MD have any analytical value because they lack context, meaning how the guy actually went about increasing or decreasing his team’s chances of winning. If you’re an end results kind of person that doesn’t worry too much about the process, then this is for you. For me, SD and MD just serve as nice reference stats, giving us an idea of who’s generally gotten the job done and who hasn’t.

Demoting David
A roster move on the horizon?
  • Jose

    I wish the Fangraphs article mentioned if this comment by tangotiger had been corrected:

    “There’s an extra added issue that Sean and David park adjust the WPA numbers. That would need to get resolved, or neutralized, as you can’t start saying it’s a shutdown in one park and not another. Otherwise, someone is going to come in and talk about the identity of the teams next. ”

    Also the .06 isn’t exactly scaled to saves+holds. That would be somewhere around .0665 Still it is an interesting concept stat.

  • Rey22

    D Rob made for his 4 and 4 last year with his insane playoff magic tricks.

    • Jammy Jammers

      Marte, too.

    • idon’tknow

      playoff magic tricks

  • http://twitter.com/Stephen_R Stephen R – formerly tafka (sic)

    Is there any talk of scaling this to SPs?

  • nothalo

    I haven’t actually crunched any of these numbers, but the comments section on “the book blog” notes that the wpa will increase .055 when a home team closes the 9th inning with a 3-run lead. How do you compensate this statistic to reflect [1] runs being scored by the relievers’ team (this will increase the wpa, which doesn’t reflect the relievers’ performance) -OR- [2] unearned runs, which again do not reflect the relievers’ performance. The degree of variability will likely be higher when applying this statistic to middle relievers.

    However, it would be interesting to try and calculate RATE of increase of WPA per save opportunity. This may further differentiate the performance of closers..

  • http://baseballalamode.blogspot.com Crunchy Frog

    Ranking the Yankees relievers so far this year from best to worse:

    1. Al Aceves 2 wins – 0 losses – 1.000 Winning %
    2. Chan Ho Park 1-1 .500
    3. Joba 0-1 .000
    4. David Robertson 0-2

    Anybody’s guess/Unrankable (can’t divide by zero)
    Mariano Rivera
    Damaso Marte
    Boone Logan
    Sergio Mitre
    Mark Melancon

    • Once and Future Lurker

      You’re assuming getting the W is a reflection on the pitcher’s ability, and it really isn’t. Mo basically never comes in with a chance to get a win, and if he does it’s because he blew the save first. Does that make him worse than Park? It certainly doesn’t put him in the same category as Mitre.

      • http://baseballalamode.blogspot.com Crunchy Frog

        Hey, over here. Can you give me a hand with this, I caught a big one…