As the 2013 regular season begins to wind down, the opinions and analysis regarding who should win a Cy Young Award for the American and National League begins to wind (and heat) up. Earlier this week at Beyond the Box Score, we had a couple of posts stumping for Clayton Kershaw and Matt Harvey, respectively, for the NL hardware; even the fine folks at Bless You Boys got in on the action, touting the spectacular season of Max Scherzer, with or without the impressive win total and percentage, as proof of his deserving the AL Cy Young.
The races for the award in each league look to go down to the wire; even in light of the unfortunate news that Harvey will be shut down for the remainder of the season due to an ulnar collateral ligament tear, the NL race still looks to be close, with the performances of Adam Wainwright and even Jose Fernandez making it difficult for Kershaw win by a landslide. The AL award will also be interesting, with perennial Cy Young candidate Felix Hernandez and dark horses like Anibal Sanchez and Yu Darvish still making waves worthy of award consideration, outside of Scherzer.
With the sizable list of candidates comes the sizable list of statistics used to quantify and qualify a given candidate's worthiness of the award. While more traditional stats like wins or earned run average still continue to paint a picture of who is deserving (fair or not), there are plenty of people who use stats more sabermetrically derived, such as FIP, WPA, and the always contentious WAR, to balance out and enhance the more traditional stats.
While Cy Young Award voting has evolved quite nicely in the last several years to include some of these advanced metrics, many fall short of quantifying one of the endless number of requirements for winning the award -- consistency. Older methods of quantifying consistency, in particular the quality start, fall short of properly evaluating true consistency as well as exceptional performance. However, with the help of sabermetrics and some recent history, we can take a stab at measuring consistency in performance and see how it could possibly affect a pitcher's hopes for winning the Cy Young.
As we have previously seen at BtB, the statistic RE24 is a nice way to quantify a player's contribution to the run environment with base-out states considered and in many ways, is a better evaluator of a player's offensive performance than things like RBI. RE24 is also a nice tool to measure and evaluate pitching performances as well, in particular, reliever performances. Since it takes situational performances into consideration, its contributions aren't limited to just relievers or hitters and can be applied to starting pitchers, with it potentially being useful in measuring the quality of performance in the form of putting the team in a position to win. Not only can RE24 evaluate the overall worth of a pitcher's outing, we can also add them up and use it to measure consistency -- great pitchers will have significantly more outings where they provide positive RE24 than negative RE24, thus providing a consistent level of excellence.
Now, to history for some assistance. Looking at the last ten years, has there been a difference in the number of quality outings (as determined by RE24) between Cy Young Award winners and the runner-ups? Can that be used to help sort the wheat from the chaff, so to speak, with respect to this year's Cy Young races?
To answer this, I first went over to the Baseball Writers' Association of America's webpage to grab the names of the winner and top two finalists for the CY Young Award in both the AL and NL from 2003 to 2012. From there, I took those names and went to Baseball Reference and looked at each player's game log for the year of interest, looking at their RE24 stats for each start, and counting up the number of positive and negative performances for that year. At this point, I still included relievers in my analysis -- more on that in a bit.
With a spreadsheet full of RE24 numbers and Cy Young winners and also-rans, I then applied this formula to their data to get a grasp of how consistent each pitcher's performance was in the Cy Young hopeful year:
# Outings - (-RE24 outings) / GS
Pretty simple -- take the number of outings, subtract the number of outings with a negative RE24 and that gives you the number of 'quality outings'. Take that number and divide it by total outings and you get a percentage of quality outings. Initially, I looked to keep the handful of relievers that have been considered, or, in the case of Eric Gagne in 2003, have won the Cy Young, in the data set. Taking this into consideration and also the fact that not all starters started the same amount of games over a season, I tried to standardize things by looking at the sum of RE24 per innings pitched. However, when comparing the starters to relievers, there was a significant difference between the two, so I felt it was best to not consider these five reliever performances. While the data set loses some sample size, it removes unaccountable variability between the starter and reliever performances with regards to RE24. Moving forward and looking at only starters, I then continued to standardize the data, using the average number of starts of all starters in this ten year sample, which came out to 33 starts. So that leaves us with this formula for RE24-determined quality games (labeled QG_RE_33), now with outings labeled as games started (GS):
GS - (-RE24 outings) / GS* 33
Applying the formula to the former Cy Young winners and hopefuls gives us the following two tables of averages, the first broken down by 1-2-3 voting results, and then a simpler, win-no win ranking:
Running a t-test on the win-no win sample and an analysis of variance for the 1-2-3 sample, we get a statistically significant difference between the winner and the runner-ups, regardless of the method of handling the runner-ups. Broadly, this means that Cy Young winners have one more quality games than runner-ups, thereby providing more consistent performances.
So with that assist from history, we have a reasonable stance on the notion that RE24 can help guide our decision as to which starting pitchers stand a chance in the Cy Young race for 2013. Let's look at the top 10 pitchers in contention for each league and see how they all stack up in the QG_RE_33 stat. Here, I am using the top-10 performers in overall RE24 (again, throwing out relievers; sorry, Craig Kimbrel!) to select my cohort:
For these tables, I have included the number of quality games pitched (as of August 27th) and labeled them QG_RE, to go along with the standardized/extrapolated 33 game totals.
So what do the tables show? Overall, nothing much that hasn't already been trumpeted -- Scherzer, with the injuries sustained by Clay Buchholz notwithstanding, is the toast of the AL, with Kershaw following suit in the NL. What does come to light with this alternative way of looking at consistency in performance are the unheralded performances of a handful of pitchers, particularly, Jose Fernandez, Chris Sale, James Shields, and Francisco Liriano, who are all performing admirably per RE24, but don't get the acclaim or accolades of a Kershaw or Scherzer. Much of this lies in their respective teams' performances (or lack thereof), but also to their own excellence in performance. To use an example, while Liriano's number of quality games per 33 starts is within shouting distance of Clayton Kershaw, his best game performance per RE24 -- 3.30 -- is nowhere near Kershaw's best performance -- 4.65. Going back to Buchholz, this analysis also speaks to the sensational start to the season he had before being beset by injury -- the fact that he remains in the top-10 in AL RE24 after only 12 starts speaks volumes as to how well he was pitching. Staying on the injury slant, this analysis also shows that even after his season ending prematurely, Matt Harvey should still be considered a strong Cy Young candidate.
While it doesn't necessarily tell the whole story, the basic premise posed does serve to better quantify the concept of consistency when Cy Young voting is discussed. While the RE24 stat is powerful, the manipulation and interpretation posed here does come with its own caveats, the largest one being the simple counting of positive and negative RE24 values. Here, all positive outings are considered equal -- theoretically, a perfect game is counted the same as a 7.1 inning, 6 hit, 4 earned run outing, which is worth 0.003 RE24. In spite of the potentially vast differences in RE24-deemed 'good' performances, with the statistical tests performed and their resulting significances, we find that the positives derived from this method of evaluation outweigh the drawbacks and show not only the importance of excellence, but consistency in that excellence when looking to award each league's best with the Cy Young Award.
All stats courtesy of Baseball Reference.
More from Beyond the Box Score:
- The Most Undervalued Player Type in Baseball
- Team WAR, Pythagorean Wins and Actual Wins
- The Chicago Cubs and their Starlin Castro Dilemma
- The Pirates and Mets make a good August deal
- The leading NL Cy Young Award contender is Kershaw, obviously