(Look, Ma, No Spread Sheets!)

            The Cy Young contests this year, to judge from my mailbag, seemed to occasion more than the usual amount of discussion.   “I'm probably not the first person who's asked this question since Greinke's award was announced,” asked Rob T. on November 17—actually he was the first person to ask it—“but do you think it's a slight "victory" for yourself and every other sabermetrician? I honestly thought CC was going to get it. But maybe at least the writers are starting to get it...”

            “Let's hope.,” I responded.   “By now, I've declared victory more often than Casanova. . ..”.   Which was a lousy answer; I was in smart-ass mode rather than thoughtful mode.   The thoughtful answer is that we should resist invitations to personalize the discussion, even when the invitations come in attractive envelopes.  It’s not about “my side” of the argument and “your side”; it’s about what’s true.   We’ll close in on the truth more quickly if we resist projecting personalities into the discussion.

            Jumping out of sequence, here’s another one:

Bryan Burwell wrote in today's St. Louis Post Dispatch that "What was my greatest fear in the past is now upon us. Armed with their "advanced metrics" and clutching their spread sheets, the new-age baseball voters have officially taken over the sport both in the front offices and behind the scenes."
Asked by: The Pope of Chili Town

Really?   Your greatest fear?   Because you know, honestly, I think I could find bigger things for you to be afraid of.   Spiders, serial murderers, terrorists, global warming, venereal disease. . .pick your poison; I’m sure you can do better than this.

            Again, this is personalizing the debate, but I don’t think Mr. Burwell has mixed feelings about it.   Burwell, writing for a St. Louis audience, is trying to smear sabermetrics by saying, in essence, that we were responsible for taking the award away from St. Louis pitchers.   Setting aside the position that it may be better not to personalize the debate, is that even what happened?   Isn’t what happened here more like two St. Louis pitchers split the vote and allowed the San Francisco pitcher to win it?

            Maybe, but there was something else going on that I didn’t know at the time.   Will Carroll voted.    According to Carroll (November 16, Baseball Prospectus), “Later on in awards season, I'll touch on the other award I had a vote for, and explain my selections and criticize the picks if you want, but in a year where the BBWAA opened up its voting to members they knew might think a little different, I have to feel like it's a step forward. Now about those Gold Gloves...”

            Ooooooooooh.    Oh, I had missed that entirely; I didn’t realize the BBWAA had granted us the vote.   Sabermetric suffrage has arrived.   Susan B. Anthony and me, we’re tight.   Now I understand what Burwell was raving about.  

            Matthew Namee weighed in, more with a comment than a question. 

More a comment than a question: Some people are obviously upset that two writers left Carpenter off of their NL Cy Young ballots, thus allegedly depriving him of the award. He only threw 192.2 innings, though. Excluding strike years, the fewest innings pitched by a Cy Young-winning starter was Sutcliffe in '84 (150.1 IP, but before that he was in the AL, and his total was 244.2). Nobody else won with fewer than 213 innings (Pedro in '99). In other words, had Carpenter won, he would have had, by far, the fewest innings pitched of any Cy Young starter, by a wide margin.

            I’m not sure that I get the logic for excluding Sutcliffe there.   Sutcliffe won the National League Award while pitching only 150 innings in the National League.  Doesn’t that pretty well cover Carpenter winning the award while pitching only 193 innings?   Rcberlo also commented on that:

What I've read about the NL Cy Young vote seems to indicate that the decision between Lincecum and Carpenter hinged mainly on the value of Lincecum's 30 "extra" innings. Does it make sense to estimate that value by looking at who occupied Carpenter's slot in the rotation when he was gone and, in effect, add 30 innings worth of that pitcher's record to Carpenter's stats?
Asked by: rcberlo

            Which doesn’t quite make sense, to me, because isn’t that saying that if the guy who filled in for Carpenter when Carpenter was out was good, then Carpenter deserves the award, but if the guy who filled in for him was lousy, then Carpenter shouldn’t get the award?

            I’m not sure I buy the theory that Lincecum’s extra innings are the deciding factor here.   I think I might stick with the explanation that two St. Louis players split the vote and allowed the San Francisco guy to win—exactly like the American League MVP Award in 1954, when two Cleveland Indians split the vote (Larry Doby and Bobby Avila), and allowed a Yankee to win, or 1965, when two Dodgers split the vote and allowed Willie Mays to win, or the Cy Young vote in 1970, when three Baltimore Orioles split twelve first-place votes and allowed a Minnesota Twin to win with six.   Et cetera.

            Ventboys, on November 19:

To amplify on the Cy Young voting, I found it encouraging that CC finished 4th, behind Felix and Verlander. Up here in the northwest we are celebrating Felix' 2nd place finish like he won. We weren't sure that the writers had heard of him. Wilbon and Kornheiser couldn't even name the teams that the two Rookies of the Year played on.

            I was even getting questions from non-subscribers:

A question from Cy Morong, who isn't a subscriber: On page 448 of the new Handbook, you have both Greinke and Hernandez with 26 win shares. How did they come out equal when most other evaluations give Greinke a substantial advantage? For instance, Fangraphs has Greinke with 9.4 wins above replacement, when Hernandez has only 6.9.
Asked by: Cy's messenger
Answered: November 19, 2009

            OK, setting aside the reader comment and getting to the issue itself.  ..I have long been a huge fan of Zack Greinke.   When he first came to the Royals in 2004 he was twenty years old, but he was a lot of fun to watch.   At that time he was one of those pitchers, like Sabathia and El Duque and Randy Johnson and Lincecum and Tiant and Fernando Valenzuela, who was instantly recognizable as something different, something outside the ordinary.   He took an epic detour between then and 2008, as has been well documented, but when I returned to the Kansas City area in late summer 2008 I saw him pitch three or four times, and I was just blown away.   He’s a very different pitcher now than he was in 2004.   He has morphed from a guy who makes pitching look interesting to a guy who makes it look easy.   I’ve never seen anybody have so many easy innings.   I told everybody who would listen a year ago that Greinke was as good a pitcher as anybody in baseball.

            Greinke, in important ways, saved the season for Royals fans.   He enabled Royals fans to go to the game every fifth day and forget that the team stunk on the other four days.    Four days in five, being a Royals fan might be painful, but when Greinke took the mound, they were more than the equal of whoever it was they were facing.   We got drunk every weekend on ZG power.

            But having said that, I still don’t see that the contest between Greinke and Felix Hernandez is a mismatch, and I was surprised that Greinke dominated the voting the way that he did.   Yes, Greinke was sensational, but so was Felix.    It seemed to me that both Greinke and Felix were well above the median standard of a modern Cy Young pitcher, and I really wouldn’t have been upset if Hernandez had won.

            The first two things I look at are the pitcher’s Season Score, and his Win Shares.   These are the Season Scores for the top American League pitchers this year:

            Had to go to eleven to get Jon Lester in there.   And here are the top Season Scores in the National League:

            The second thing I looked at was Win Shares.   These are the 2009 Win Shares for the top Cy Young contenders:

            Season Scores are just a method of making simple comparisons of what the value appears to be, rather than what it really is, and Win Shares are not a way of comparing two seasons, but rather, a system of very carefully comparing groups of players or groups of seasons such as careers.   What these systems do is not resolve the Cy Young or MVP debate, but rather, make it apparent who the leading contenders are.

Very often these systems make it apparent that there is only one serious contender for an award—the Cy Young Award in the American League in 2008, for example, when Cliff Lee outdistanced the field by 57 points in the Season Score.    The Season Score predicts the Cy Young voting a high percentage of the time, but it doesn’t prove anything, and isn’t intended to.

2009 is not one of those years when these two systems will end the debate.  What these systems do in 2009 is make it apparent who the serious candidates are—Greinke and Hernandez in the American League, Wainwright, Carpenter and Lincecum in the National.   Which, of course, we knew anyway; we’re spinning our wheels here. 

            Sometime this summer I introduced the Strike Zone won-lost record, which is a way of stating a pitcher’s strikeouts and walks as if they were wins and losses.   These are the top strike zone winning percentages of 2009. ..”KS” stands for “Strike Zone Wins” and “KL” for “Strike Zone Losses:”

            Greinke’s strike zone won-lost record is 18-8, a little better than his actual record of 16-8, and Hernandez’ strike zone won-lost record is 16-10, which is significantly worse than his actual record of 19-5. 

            Of course, again, a great pitcher doesn’t have to have a great strikeout to walk ratio; it merely often happens that he does.   95% of Cy Young Award winners have strike zone winning percentages over .500-but there are other elements to the game.  I am always amazed at how many pitchers have about the same strike zone won-lost record as actual won-lost record.   These are the National League leaders:

            I think the data for Cliff Lee and John Smoltz there includes the data from both leagues.    At this point in the analysis I understand something about the Cardinals that I didn’t previously understand.   Throughout history there have been several teams that, working in a pitcher’s park, were able to make good finesse-type pitchers appear to be as dominant as power pitchers.  The Orioles of the 1960s/1970s are the classic team of this type, but there have been others:  the Oakland A’s of the 1970s and 1980s, the Dodgers of the 1970s and 1980s, the Mets with Koosman and Seaver, the Reds of 1939/1940, the Milwaukee Braves of the late 1950s, and others.   If you work in a pitcher’s park and you don’t walk anybody and you have a good offense behind you, you’re likely to have a winning record and a low ERA.   You’re going to look a lot better than you really are.   The situation favors a finesse pitcher. 

            The Cardinals, I now understand, have become one of those teams that makes low strikeout/low walk pitchers look better than they really are.    Wainwright, Carpenter, Pineiro—these guys are all low strikeout/low walk pitchers.   It’s not that they’re not good; it’s just that they’re not as good as their records look.   I didn’t get that a half-hour ago.  

Stripping those charts down to the relevant contenders:

            This still is not how we find the deserving Cy Young winners; it is merely another thing that is worth looking at.   Probably 70% of a pitcher’s value is centered in the strikeout and walk categories.   Looking at that 70%, Greinke and Lincecum are ahead.

            Starting now along the path toward an actual answer, let us begin with Runs Saved.   Greinke pitched 229.1 innings with an ERA 2.29 runs better than the league average (2.16 versus 4.45).  That makes him 58 runs better than an average pitcher.   Felix pitched 238.2 innings with an ERA 1.96 runs better than the league average.    That makes him 52 runs better than an average pitcher.   Advantage, Greinke.

            This, however, is without making park adjustments.   Kauffman Stadium was re-modeled between 2007 and 2009, and is, in essence, a new park.    The new park played in 2009 as the best hitter’s park in the American League.  The Royals scored and allowed 805 runs in Kauffman Stadium, 723 on the road.   Adjusting for that, Greinke is not 58 runs better than an average pitcher, but 64 runs better.

            Puget Sound Park in Seattle had a park effect of .95—a pitcher’s park.  Adjusting for that, Hernandez is not 52 runs better than average, but 49 runs better.   Greinke is now ahead, 64 to 49. 

            OK, Greinke is well ahead now, and I’m happy that Greinke won, and nobody seems to be too unhappy about it, so let’s set that one aside for now and concentrate on the National League race. In the National League without park effects, Lincecum in 42 runs better than the league average, Carpenter 41 runs better, and Wainwright 40 runs better.

            Departing from this point, one can construct a sabermetric argument that Carpenter is the best pitcher in the league.   Lincecum leads Carpenter by a razor-thin margin, but this is without considering:

1)  Un-earned runs,

2)  Pitcher’s hitting, and

3)  Efficient use of the runs.

            Carpenter gave up 48 earned runs, but gave up only one un-earned run.   Lincecum gave up 61 earned runs, plus seven un-earned runs.  Much analytical thinking proceeds from the assumption that only earned runs are relevant, but clearly this is untrue.    The first un-earned run that Lincecum gave up came about when Juan Pierre was on first for the Dodgers, and Lincecum tried to pick him off.   Lincecum threw the ball away—E-1—putting Pierre on third base.

            I won’t argue about it being an “earned” as opposed to “un-earned” run—but it is clearly and absolutely a run for which Lincecum is responsible.    Even setting that aside, it seems to me more reasonable to hold a pitcher 50% responsible for un-earned runs, rather than to say that he is totally absolved of responsibility for any run that is tainted by fielding malfeasance.   So let’s hold Lincecum totally responsible for the Juan Pierre run and hold the pitchers 50% responsible for other un-earned runs.    Carpenter goes from 1 run behind to 2 and a half ahead:

            Next, let’s consider the pitcher’s own contributions to the offense.   Runs created formulas don’t work well when the totals get close to zero; what works better there is just to assume that every four bases is a run and every seven outs are a negative run.   Lincecum, with a .386 OPS, had 11 total bases and 6 walks, contributed 17 bases with 57 outs, not counting sacrifice hits and flies, which are sort of “justified outs”.    Carpenter, with a .482 OPS, had 17 total bases, 2 walks and a hit by pitch, 20 bases contributed with 53 outs.   Figuring 4 bases is a run, that’s ¾ of a run for Carpenter plus 4/7 of a run for having fewer outs; Carpenter gains another 1.3 runs:

            Carpenter is now ahead by almost four runs. 

But this now gets Wainwright back into the discussion; Wainwright was the best hitter of the three.   Wainwright hit 5 doubles, a triple and 2 homers.  His total bases plus hits are 30, with 73 outs.    Going back a step to include him, he also allowed 7 un-earned runs:

            But we haven’t yet dealt with “efficient use of the runs”.    Stating the traditionalists argument in its best form, Carpenter had, among these three pitchers, both the best winning percentage (.810) and the best ERA (2.24).   Carpenter was 17-4, 2 and a half games better than

Lincecum.  This has usually been enough, in the history of Cy Young voting, to sway the decision.   If you have the best record and the best ERA, you win; end of discussion.   One can look at strikeouts and walks, but, as we showed earlier, Lincecum doesn’t really have better strikeout and walk data than Carpenter; he merely has more strikeouts.   Lincecum’s strike zone winning percentage is .656; Carpenter’s is .653.  It’s not a meaningful difference.

            The bottom line is winning.   Look at it this way:  the Cardinals scored 125 runs in Carpenter’s 28 starts—4.45 runs per start—and Carpenter was 17-4.   The Giants scored 146 runs in Lincecum’s 32 starts—4.56 per start--yet Lincecum was “just” 15-7.  The Cardinals scored 125 runs in Carpenter’s starts, and won 18 games.   That’s one win for each 6.94 runs.   The Giants scored 146 runs in Lincecum’s starts, and won 19 games.   That’s one win for each 7.68 runs.  Carpenter simply made more efficient use of his runs. 

            At this point, however, the pro-Carpenter argument begins to collapse.    There are four issues here we haven’t fully examined:

1)  Park Effects,

2)  Durability,

3)  Quality of Competition,

4)  Systematic as opposed to testimonial study of “Run efficiency”

            I carefully dodged the park effects data before I departed on the pro-Carpenter argument, because if I hadn’t, there’s no pro-Carpenter argument.   Busch Stadium is a pitcher’s park, with a 2009 Park Effect of .919.   Willie Mays Park in San Francisco is a hitter’s park, with a 2009 Park Effect of 1.052.   If you adjust for that, Lincecum isn’t 42 runs better than average, he’s 45 runs better than average, and Carpenter isn’t 41 runs better than average, he’s 38 runs better than average.   The difference between them isn’t one run, it’s 7 runs.

            It’s seven runs versus an average pitcher, but value does not consist of being better than average; it consists of being better than replacement level.   No one can say exactly what the replacement level is, but let’s say the replacement level is a pitcher 35% worse than the league average (an ERA about 5.66 in the National League, 6.01 in the American League, without park adjustments.)

            If you compare the pitcher to a replacement level pitcher, Lincecum’s advantage over Carpenter isn’t 45-38, it’s 83-68.   The advantage for Lincecum grows because, pitching 30 additional innings, he is taking a innings away from a bad pitcher, which is the real situation that baseball teams face.   When your quality pitchers go out, the guys you have to put on the mound are usually pretty awful.  You don’t win too many of those games.

            Then let’s look at the issue of quality of competition.   Chris Carpenter made 28 starts in 2009.   Eighteen of those starts were against teams with losing records.   Carpenter started only 3 times against teams with 90 or more wins.   Lincecum started 11 times against teams with 90 or more wins.   That’s looks like huge difference, although, as we’ll see in a moment, it really isn’t.  Carpenter had only three starts all year in which it was a fair fight.   Lincecum had 11.    Let’s chart it:

            The aggregate won-lost record of the 28 teams that Carpenter faced was 2193-2341, a .484 winning percentage.   The aggregate won-lost record of Lincecum’s opponents was 2603-2579, a .502 percentage. 

            Let’s see. . .an average National League team this year allowed 4.49 runs per game.   To cause an 18-point difference (.018) in winning percentage would require almost exactly one-sixth of a run per game.   Assuming 24 games—216 innings--that’s 4 runs.   To offset his disadvantage in facing tougher competition, we need to credit Lincecum with an additional four runs.

            Four runs isn’t a huge difference, but. . .it’s a difference.  Lincecum was ahead 83-68; he’s now ahead about 87-68.    You can give Carpenter a run for being a better hitter, three or four runs for the un-earned runs. . .it doesn’t matter as much if Lincecum starts out 19 runs ahead. 

            Carpenter’s defense collapses completely when we take a more systematic look at the issue of Run Efficiency.   The Cardinals with Carpenter got one Win for each 6.94 runs versus one Win for each 7.68 with Lincecum, true, but you know what that is?  That’s just the Park Effect again—the Park Effect, combined with the fact that Carpenter was a little bit more effective per inning pitched.    When you have fewer runs, each run has more impact in the win column, independent of the issue of run/win efficiency.  A better way to look at the issue of run/win efficiency is to look at the runs scored, runs allowed and won-lost record in the pitcher’s starts, which can be found in our system under Statistics/Team Profiles/Performance by Starting Pitcher.

            In Carpenter’s starts this year the Cardinals scored 125 runs, allowed 68 runs, and finished 18-10.    If you score 125 runs and 68 runs, what should your winning percentage be?  Sabermetrics 101; it’s .772.   (In Sabermetrics 413 you’ll learn that it’s actually not .772, but more like .750, but let’s keep it simple; Burwell might be listening at the door.)   At 18-10, their actual winning percentage in these games was .643. 

            Far from being efficient in his use of runs, Chris Carpenter was actually one of the most in-efficient pitchers in the major leagues, by this method.   He was short by 3.6 wins, making him the sixth most inefficient pitcher in the majors. 

            This doesn’t help Lincecum a lot; Lincecum was short by 2.6 wins as well.   The Giants winning percentage in Lincecum’s starts should have been .676 but was actually .592.   This doesn’t help Lincecum a lot, but he doesn’t need the help.  At this point in the analysis it is Carpenter who is grasping at straws, and this particular straw isn’t going to hold him up.

            That isn’t a perfect analysis, either; the run/win efficiency method that I just outlined has three serious problems:

            1)  That it treats runs scored after the pitcher has left the game the same as runs scored while he was in the game,

            2)  That it treats offensive support of 11-1-1 in three games the same as offensive support of 4-5-4, when in reality they are very different,

            3)  That is uses the Pythagorean approach in a very low-run context, where the Pythagorean method becomes less accurate.

            One never gets perfect answers.   I have been doing post-award analysis like this for 30 years, actually more than 30 years.   The analysis that we can do now is much, much more sophisticated and more accurate than the analysis that we could do 30 years ago, but we are no closer to perfect methods now than we were then.   We have introduced many more factors into the analysis, and much more available data.   These new benefits give us better answers—but they also expose flaws in the old reasoning, requiring us to think more deeply about some of the issues.   There is no end to it.

            There is no end to the argument, but I am now satisfied that, at the end of the argument, Lincecum is still going to be ahead.   Tim Lincecum won the award; Tim Lincecum deserved the award.   I am satisfied that that is true.

            Let’s go back now to Greinke versus Felix, and look at some of the same issues.   Hitting. . .we can skip that; American league hitters don’t hit very often.   Greinke had a double and Felix didn’t, but we’ll skip it.

            Greinke allowed 9 un-earned runs, Felix allowed 15.   Holding the pitcher 50% responsible for those, that gives Greinke an additional 3-run edge, which puts him ahead by 18 runs, a probably impossible advantage for Felix to overcome.

            To get to durability, let’s back up and look at runs vs. replacement pitcher, rather than runs versus average.   Greinke was 106 runs better than a replacement-level pitcher; Hernandez was 89 runs better—but remember, each run has less impact as more runs are scored.   Greinke was pitching in a hitter’s park; Hernandez, a pitcher’s park.   If you normalize for that, the advantage isn’t 106-89, but 101-92.  It’s close.

            Then we look at run/win efficiency. . ..you remember, Chris Carpenter was minus 3.6, Lincecum minus 2.6.    Felix Hernandez was +2.3.    Greinke was minus 1.6.  Let me chart the data:

            Putting this in plain English:   The Mariners were 25-9 with Hernandez on the mound.   The Royals were 17-16 with Greinke—and the difference cannot be fully explained by run support.   There is something else going on.   

            There’s a problem with that method, as I said before; Greinke’s bullpen was awful, and we can’t hold him personally accountable for that.   On the other hand, if we did hold him personally responsible for all of it, a difference of 3.9 wins is a difference of roughly 39 runs.   Hernandez would vault far ahead in the overall analysis.

            Then we get to the issue of Quality of Competition. . .quality of the opposition, whatever you want to call it.   Hernandez started 22 times against teams with winning records.   Greinke started 15 times.  Let’s do the chart that we did before, with Lincecum and Carpenter:

            The 33 teams faced by Greinke had an aggregate record of 2599-2733, a .487 winning percentage.    The 34 teams faced by Hernandez had an aggregate record of 2840-2674, a .515 percentage.   Greinke had an advantage of .028 (winning percentage), based simply on facing weaker opposition.

            Stating that as runs. . .to offset an advantage of .028 winning percentage, at the level of offense in the American League in 2009, requires about .275 runs per game.   Given that they each pitched about 26 full games (234 innings), it’s a difference of about 7 runs.  

            Well. ..I’m not here to argue that Greinke didn’t deserve the award.   Greinke very much deserved the award, very richly deserved the award.   But whether he was actually better than Felix Hernandez. . .I don’ t know.    It comes down to the issue of how much weight is given to the differences in run/win efficiency, and our methods on that specific issue are crude and un-convincing.  

            But here’s what I would say.   In the National League, the vote was split three ways, it was a very close vote, and it’s been a controversial vote.   In the American League Greinke won easily, and this vote has been uncontroversial, and this vote has been celebrated by the analytical community as a victory for reason and logic.

            But actually it seems pretty clear to me, under the most careful analysis that I can do, that Lincecum was the best pitcher in the National League and deserved the award—whereas in the American League, under the most careful analysis that I can do, it is unclear to me whether Greinke or Hernandez is more deserving.  They are both very deserving, but I can’t say which is more deserving.   Ask me again in another 30 years, I’ll have a better answer for you.