It is my understanding that Carl Yastrzemski (a) was generally regarded as an outstanding defensive left fielder, but (b) does not have very good defensive statistics, very good range numbers, because he played in Fenway Park, where the left field area is small compared to other parks.   Let us accept for the sake of argument that these assumptions are true and move on; it doesn’t actually matter whether they are true or not, they are just here to illuminate the problem.  IF that is true, what does the analyst do about it? 

            Sabermetrics is not generally about "what if" questions.   Sabermetrics does not generally worry about what if questions.  It does not and cannot answer them with confidence.  Suppose that we start with Carl Furillo, Dodgers.   (I thought we were starting with Carl Yastrzemski?)  Suppose that we start with Carl Furillo.   Carl Furillo could have played Center Field for that great Dodgers team, rather than right field.  We know that he could have, because he did, for most of three years, until the Dodgers came up with Duke Snider.  Furillo was the Dodgers’ #1 center fielder in 1946, 1947 and 1948.  As much as we can judge, he was pretty good; the Dodgers were 84 games over .500 in those three seasons (96-60, 94-60, 84-70), and Furillo’s putout numbers were good.  He made 276 putouts in 764 innings in center in 1946, and 274 in 781 innings in 1948.   The league average was less than one putout for every three innings in center, so a guy making more than one putout every three innings for a championship team looks like a good center fielder.   In 1948, when Snider and Furillo split the center field job with others, Furillo made easily more putouts per 9 innings than Snider did. 

Furillo moved to right field, and that is what we know him as, that is what we evaluate him as, but what if he had stayed in center field?  Might he not then be a Hall of Famer?

It’s impossible to answer.  We can’t evaluate him as a center fielder.  We can’t evaluate him based on what-ifs.   It doesn’t make any difference—and further, if you DO go down that road, you find yourself wandering lost in a forest of unanswerable questions.  What if the Royals had come up with Willie Wilson before they came up with Amos Otis?  Otis would have had to play left.  How would we evaluate him then?  What if Hal McRae hadn’t destroyed his ankle in the minor leagues in 1968?  He would probably have been the center fielder on the Big Red Machine—the job that went first to Bobby Tolan, then to Cesar Geronimo after Tolan snapped his Achilles Tendon.    Wouldn’t McRae clearly be in the Hall of Fame if that had happened?   What if Tolan hadn’t gotten hurt?  How would we evaluate him then?  What if Denkinger had called Iorg out?  What if there had been replay?  What about Gil Hodges?  People who watched that team would say that Hodges could have played shortstop.  The Dodgers had Pee Wee at short and Jackie at second and needed a first baseman, so they put Hodges on first, but should this decision keep him out of the Hall of Fame? 

The number of what-if questions that pop up if one attempts to shoot them down is endless, and trying to answer them is not what sabermetrics is about.  In this form, people understand this and accept it.    The question of "What if Carl Yastrzemski had played in some other park?" is a what-if question, on a certain level no more legitimate of a subject for sabermetric evaluation than any other what-if speculation.  The confusion comes from the overlap of questions.  The overlapping question which confuses the issue is "How good of an outfielder was Carl Yastrzemski?" 

Sabermetrics does not and cannot directly answer questions like "How good. ..?"  "How good" is (a) comprehensive, (b) unlimited, and (c) a value judgment.  "How good" includes EVERYTHING.   What Sabermetrics does is to define the more limited questions which give us valuable information about the unlimited questions.   Sabermetrics creates a pathway toward understanding.  Value is in Wins; wins are created by runs, losses are created by outs.  The pathway toward evaluating Carl Yasztremski’s defense—or Carl Furillo’s, or Carl Crawford’s—is built by posing a series of limited and answerable questions, like this:

1)     How many runs did Carl save for his team?

2)     How many runs did he have the opportunity to save?

3)     What was the win value of each run saved? 

The question of "How many runs did Carl save for his team?" breaks down into about 5,000 other, smaller questions.  I have been working on all of those questions for the last four or five months.   After we have a consensus understanding of that one, then we can go on to the next one. . . .or you can go on to the next one; I’m an old man and won’t necessarily be around to participate in all of these discussions.   The question of how the park effected his ability to save runs can also be studied in hundreds of different ways.  You can take 50 years of Fenway Park box scores, and look at how many plays per game the left fielders made in Fenway and how many they made on the road, and how many plays per game opposing left fielders made in Fenway and how many in the Red Sox road games. 

A big thing is the doubles.  Fenway Park is a great park for doubles, because of the Green Monster and also because of the right field wall curving rapidly backward from the Pesky Pole.   There are a substantial number of balls which would be fly outs in another park which become doubles in Fenway because they hit the Monster before they hit the left fielder’s glove, and there are also a substantial number of balls that would be doubles or homers in another park which become singles in Fenway because they ricochet off the wall too quickly to allow the batter to get to second.  But there is also a range of in-between balls in Fenway, unlike plays in other parks, which can be doubles off the wall or singles with the runner thrown out at second, depending on how the left fielder plays them.   Yastrzemski’s reputation as a great defensive left fielder was mostly based on those either/or plays.   Documenting that ability, finding a way to measure it, is not a what-if exercise; it is a real-life problem.  It is also not easy.  It would be a lot of work.   But it is a legitimate sabermetric issue, rather than a what-if problem.  

You can think what you want to about it, but my understanding of this issue derives from discussions in the late 1970s.  We were exploring and documenting park effects, which had not previously been documented for the most part, and we were becoming aware for the first time of how differently some players would be evaluated if it were not for the parks that they played in.  We were trying to adjust players into a neutral context, which is a what-if exercise, and this was leading to unresolvable problems.  Eventually the discussion settled on the comparison of Elston Howard to Bill Dickey.   

Bill Dickey and Elston Howard, both Yankee catchers, both played their best years in the same park.  Yankee Stadium’s park effects relative the league were essentially the same in Howard’s era as they were in Dickey’s era.  Their careers are of roughly the same length, and Howard won an MVP Award, which Dickey never did. 

The park, however, was great for Dickey, and terrible for Howard.  Dickey hit 135 career homers in Yankee, 67 on the road.  Howard hit 54 in his home parks, 113 on the road.  There are three reasons Dickey is in the Hall of Fame and Howard is not, none of which have anything to do with ability.  The three reasons are:  (1) Yogi Berra blocked Howard during what would have been Howard’s best seasons, (2) the park helped Dickey and ruined Howard, and (3) the hitting numbers were much bigger in the 1930s than in the 1960s, which misleads people in their evaluation.   If you project both Dickey and Howard into a neutral environment and given the same opportunity, let us assert for the sake of argument that Howard is clearly greater than Dickey.  

But the problem is (a) it is the same park, and (b) in establishing the value of each player, you cannot adjust real wins and real losses out of existence.   If Dickey is able to take advantage of that park and Howard is not, there are real wins and real losses that result from that ability.   The foundation of sabermetrics is that individual production by players explains wins and losses.   If you blow up that bridge between performance and wins, you’re not doing sabermetrics, you’re just farting around with statistics.

What matters is not what Dickey or Howard "would have done" in another park or in a neutral park.  What matters is, In the time and place where he played, how many games did he win for his team?   How much opportunity space is there for him, and what did he do with it? 

The sabermetric pathway for a hitter is:

1)     How many runs did he create?

2)     How many outs did he make?

3)     How many wins resulted from those runs? 

That is as close as you are going to get to the question of "How good was he?"—and it involves no what-if questions.  And my answer to the question of "How many runs would Carl Yastrzemski have saved, playing in another park?" is "It doesn’t make any difference.  He did what he did.  He created the number of runs that he created; he saved the number that he saved.  That’s the best answer you can get that has value.   The what-if stuff is not creating knowledge; it is just farting around with statistics." 

If you start with the question "How good of a defensive outfielder was Carl Yastrzemski?", then you are forcing yourself to address a question that you cannot answer.   If you use the sabermetric pathway, you have an explanation for Yastrzemski’s defense:  Yastrzemski saved relatively fewer runs than other left fielders, but this was not because of his skills.  It was because the park limited his ability to save runs."  But if you start with the question "How good of a defensive outfielder was Yastrzemski?" then you have left the sabermetric pathway, and you will never reach your goal.   That’s what I believe.

You will note that I have left three doorways open here.   The three doorways are:

(1)  That there is still much work that could be done in documenting Yastrzemski’s defense, such as his ability to prevent doubles, and

(2)  That sabermetric methods may be useful in addressing what-if questions, even though what-if questions are not a part of sabermetrics.   What-if questions are not a part of sabermetrics because they do not and cannot produce knowledge, and sabermetrics is a process of building pathways to understanding based on knowledge.

(3)  What-if questions are vital to a baseball team in real time.     In looking at the future, not at the past, but in looking at the future, everything is a what-if question.  What if we move this starter to the bullpen?  What if we move our shortstop to second base?  What if we sign this man as a free agent; what will he hit in our park?  What if we bring this player out of the minor leagues?  

The future is a long list of what-if questions.  My point is that the past is not.  Carl Yastrzemski is retired.  Carl Furillo is dead; he’s been dead for 30 years.  Asking what-if questions about them, in my view, is no longer of any value.