Thursday, November 08, 2012

IBA Ballot: Cy Young

2012 was not really a banner year for starting pitchers. The RAR totals turned in by the top pitchers in each league were not quite at the typical levels for recent seasons. To illustrate, here are the AL and NL leading RAR totals for 2008-2012:

AL: 80, 91, 76, 84, 72
NL: 75, 70, 82, 73, 67

In the American League, though, I think the Cy Young choice in 2012 is easier than it was in 2011, when Justin Verlander’s 24-5 record propelled him to a unanimous Cy Young and a MVP award. And once again it is Verlander on top of the ballot. He led the AL in innings, eRA, RAA, and RAR, and was third in RRA and second in dRA.

The second place finisher in RAR was David Price, who led the league in RRA. But he pitched 27 fewer innings than Verlander and his RRA was only .16 runs lower, giving him 67 RAR to Verlander’s 72. If you use a higher baseline, it’s naturally closer--Verlander's RAA lead was 39-38. Verlander and Price were very close in both eRA and dRA, but in each category Verlander had a narrow lead. Verlander also had a narrow edge in strikeout rate (9.5 to 9.1) and walk rate (2.4 to 2.6). Price is close enough that I wouldn’t dismiss a case made on his behalf (that Price led the AL in QS% 81-76 over Verlander would not be the way to start that case)--but I wouldn’t make it myself.

After Verlander and Price, five pitchers fall between 55-61 RAR and make a good group to consider for the final three spots on the ballot. Chris Sale led this pack with 61, and his 2.86 RRA was second only to Price. However, he only pitched 192 innings and his peripherals were not as good as his RRA. The pitcher right behind him, Matt Harrison, had 59 RAR but a much bigger problem with peripherals--his 3.99 eRA, 4.43 dRA, and 5.7 KG don’t stand up in this company. Felix Hernandez, Jered Weaver, and Jake Peavy had just one RAR between each step (57, 56, 55); Hernandez logged the most innings (232 to 189 and 219), Weaver’s 3.03 RRA was .42 runs lower than Sale’s and .44 lower than Hernandez’s, and while only .14 runs of eRA separated the three, Hernandez led the league with a 3.10 dRA while Peavy (4.00) and Weaver (4.11) had more pedestrian marks. All had essentially the same walk rate (2.1-2.3), but Hernandez’s 8.9 KG led Peavy (8.3) and Weaver (7.3). Given how close these three were, Herandez’s peripherals put him over the top in my book, but not convincingly enough to move him ahead of Sale. So my ballot was:

1. Justin Verlander, DET
2. David Price, TB
3. Chris Sale, CHA
4. Felix Hernandez, SEA
5. Jered Weaver, LAA

The NL Cy Young race is a great example of why it’s silly to pretend that award winners from year-to-year are created equal. The third-place finisher from one season could very well have contributed as much to his team as the winner from the next season. Clayton Kershaw is my choice for NL Cy Young in 2012, but his season was no better than his 2011:



I didn’t adjust these figures for league scoring context, but the NL average R/G was 4.13 in 2011 and 4.22 in 2012, so it would make little difference.

In 2011, Kershaw won the real award, but was edged out on my ballot by Roy Halladay. This year, RA Dickey was second to Kershaw with 60 RAR, and I would love to find an excuse to elevate the knuckleballer to #1 on my ballot. But I can find no such reason. Dickey’s eRA was .42 runs worse than Kershaw’s, his dRA .35 runs worse.

Much closer is the second-place choice between Dickey and Johnny Cueto. Dickey’s edge comes from his sixteen additional innings as Cueto had a better RRA and they were very close in eRA (3.43 to 3.49, Dickey) and dRA (3.69 to 3.72, Cueto, with the caveat that knuckleballers and DIPS don’t play nice), so Cueto’s injury was the only thing stopping him from earning the second spot on my ballot.

The final two spots on my ballot went to Matt Cain and Gio Gonzalez, with Gonzalez getting the edge over Cain thanks to his significantly better eRA (2.91 to 3.40) and dRA (3.25 to 3.86):

1. Clayton Kershaw, LA
2. RA Dickey, NYN
3. Johnny Cueto, CIN
4. Gio Gonzalez, WAS
5. Matt Cain, SF

A quick word on relievers, since Fernando Rodney and Craig Kimbrel will garner a fair amount of Cy Young votes (although we now know that neither finished in his league’s top three). First, both had outstanding seasons and ranked as the top relievers in their league in RAR. It’s quite common for a non-closer to lead the league in RAR thanks to pitching more innings, but Rodney’s .73 RRA in 75 innings was good for 35 RAR (Kelvin Herrera was second with 25), and Kimbrel’s .84 RRA in 63 innings for 27 RAR led NL relievers (excluding teammate Kris Medlen, who with less than 15 starts was classified as a reliever in my stats; Aroldis Chapman was next with 25).

However, the bottom pitcher on my AL and NL ballots had 56 and 53 RAR respectively. That means that I would have to give Rodney a 1.6 leverage credit and Kimbrel a 2.0 leverage credit to pull them even. However, I would not simply apply their leverage indices directly to adjust their RAR. This really should be a longer, post-length explanation of my position, but to try to sum it up in two paragraphs:

WPA and associated metrics (like LI) are based on a real-time perspective on value. A real-time perspective on value is quite valuable, as it is what the participants in the game actually face at any given time. But since estimating value is by its nature a backwards-looking exercise, I don’t see any reason why we should feel constrained to measuring real-time value. After the game is over, it is clear that each run was equally important (absent the effect that it had on future strategic decisions, any emotional/psychological effect, and the like), regardless of how it appeared at the time. A run-scoring play in the bottom of the ninth of a 1-0 game will almost certainly have a higher WPA than the same play in the bottom of the first. But both were equally essential to the game’s outcome, and looking back after the fact, I don’t feel bound in the slightest to give more credit to the former.

So I believe that elite relievers do in fact deserve credit beyond what their non-contextual stats would indicate, as they log a higher percentage of their work in high-leverage games (that is, close games) than starters. But their leverage indices from an ex post facto game perspective would necessarily be less than their leverage indices from a real-time perspective--and the former are barely high enough to push them into the Cy Young conversation.

2 comments:

  1. Are you using a FIP-based RAR model? If so, you could be underestimating Dickey, since we know that knuckleballers create low BABIPs. But regardless, I felt similarly about the pitchers: they were a close group this year, and it was really tough to separate a leader from the pack. So here were my ballots:

    AL CY:
    Verlander
    Price
    Hernandez
    Scherzer
    Sale

    NL CY:
    Dickey
    Lee
    Kershaw
    Strasburg
    Kimbrel

    I did a lot of mixed-method sort of stuff, with no clear method as to how to combine it. Probably the most important measure I used, though, was (K-BB)/IP. So I was VERY tempted to put Cliff Lee first on my ballot in the NL. The guy had OUTSTANDING peripherals. He struck out 207 while walking only 28, and while 26 HR allowed is a lot, it's okay in light of the strikeouts, walks, and fact that he pitched his home games in Philly. And that's also why Scherzer ends up so high on my ballot, but left off yours completely. And I gave Kimbrel a 5th-place vote, just because I still am not sure how to judge relievers who pitch such limited numbers of innings through no fault of their own, against starters. And Kimbrel's rate stats were good enough for him to earn a vote in my book.

    ReplyDelete
  2. No, my RAR is based on actual runs allowed with adjustment for bullpen support.

    You seem more likely than me to throw out homers and/or consider the effect of a park on particular events (I only consider runs-based park factors). That probably explains our differences re: Lee and Scherzer. Scherzer's K and W rates were great, but he gave up a HR in 3% of PAs which was right at the league average. From a pure DIPS perspective, he's almost dead even with Sale, as they each have a 3.58 dRA and a difference of just 12 plate appearances.

    ReplyDelete

Comments are moderated, so there will be a lag between your post and it actually appearing. I reserve the right to reject any comment for any reason.