Recently in the “Hey, Bill” section we had a discussion about how one could evaluate hitting coaches.   A simple approach to the problem is just to count how many hitters have good years under the coach’s guidance, and how many have poor years.   How do you do that?

            The process that I am going to outline here has some issues, but. . .it’s a start.   Let us suppose that ignore we players with less than 250 plate appearances on the season, and sort the regulars into

            a)  those who had very good years,

            b)  those who had very poor years,

            c)  those who had reasonably typical years. 

            I created a data file of all players who had 250 plate appearances in a season from 1950 to 1999, ignoring those before 1950 because most or all of them didn’t have hitting coaches, and those since 2000 because the full picture of their careers is not in yet, at least in many cases, which could effect how we evaluate the season.  There is no urgency about evaluating recent seasons; we’re just trying to develop a method here.

            Anyway, there were 11,287 players who had 250 plate appearances in a season from 1950 to 1999; doesn’t matter, but I always like to get those things in.  What is a hitter who has a “very good” year?   Let us say that the top 25% have very good years, and the bottom 25% had very poor years.

            Top 25%, by what?   How about OPS vs. career OPS?

            I subtracted each player’s career OPS from his season OPS.   These were the top ten and bottom ten seasons of the years 1950-1999:

             And no, Norm Cash did NOT say later that he was using a corked bat in 1961.   He said that somebody (presumably Billy Martin) showed him how to cork a bat ten years later, 1971, which extended his career by about three years. 

            This method, however, ignores season-to-season fluctuations in batting levels, which might cause batting coaches from 1968 to score as having bad seasons, and hitting coaches from 1998 to score as having good seasons.   Is that just a theoretical concern, or a real one?

            I scored all seasons 1950-1999 as 3 (outstanding), 2 (ordinary) or 1 (poor) by OPS advantage, without worrying about season-to-season fluctuations in batting levels.    But in 1950, the first year of the study (and a big hitting year), I wound up with 59 hitters having outstanding seasons, 12 hitters having poor seasons.   In 1952 (a pitcher’s year) the ratio was 13 to 40.   In 1962 the ratio was 70 to 18; in 1968 it was 13 to 93.     In 1992 it was 31 to 102; in 1999 it was 112 to 41. 

            Obviously, then, we can’t ignore this in sorting seasons as good, bad, and ugly.  To avoid this problem, I ranked the players in each season by their OPS advantage.   In 1999, for example, there were 307 players who had 250 plate appearances.   Terry Shumpert, who had a career OPS of .724, had an OPS that season of .997, or +.273.  He ranks as the number one out-performer of 1999.

            Shumpert ranks first, Troy O’Leary ranks 100^th (.838 vs. .780, +.058), Ricky Gutierrez ranks 200^th (.690 vs. .689, +.001), Doug Mientkiewicz ranks 300^th (.655 vs. .766, -.111), and Sean Berry ranks dead last, 309^th (.582 vs. .779, -.197).  (Mientkiewicz’s career OPS is now .765, not .766, but this study was done with data through 2007. Berry and Shumpert, oddly enough, were long-time teammates in the Royals minor league system, both reaching the majors with the Royals in 1990.)

            Anyway, we then convert the relative OPS performances into percentiles.   Using “BR” for batter’s rank and “GT” for group total, the formula is 1 - (BR-1)/(GT-1).   Shumpert ranks at 100.00, O’Leary at 67.9, Gutierrez at 35.4, Mientkiewicz at 2.9, and Berry at 0.0.  I then re-ranked the players, Top 25% in group “a”, next 50% in group “b”, bottom 25% in group “c”.   In this way, the season bias is taken out of it to a certain extent. 

            Not entirely, of course; I am sure you will all realize that a player whose career runs 1963-1968 (a pitcher’s era) would have a different “1968 bias” than a player whose career runs 1968-1973 (more hitting.)   That’s another problem; you can deal with that one when you can figure out how.  

            Anyway, a problem with this method at this point is that it expects Carl Yastrzemski to hit the same in 1983, when he was 43 years old, as he did in 1967, when he was 27. . .actually, it expects him to hit better in 1983 than he did in 1967, since there are many more hitters have good years in 1967.   Obviously, this would penalize whoever was the hitting coach for the Red Sox in 1983 (Hriniak, I guess), by charging him with failure to lift the 43-year-old Carl Yastrzemski to the levels expected of Carl Yastrzemski.             

            Is this a real problem, or a theoretical one?   I took the a-b-c groupings derived through the previous stage, and sorted them by age.   This established that it was a very real problem. . .apparently I failed to save some data there.   It’s a shame, because it was really nice data.  . . .here; I’ll re-create a little bit of it.  

            At age 20 there were 57 players in the study, of whom 9 were in the top 25% of their league in relative OPS, 19 were in the middle range, and 29 were in the bottom 25% of their league. 

            At age 25 there were 1034 players in the study, of whom 272 were in the top 25% of their league, 537 were in the middle range, and 225 were in the bottom 25% of their league.

            At age 27 there were 1143 players in the study, of whom 326 were in the top 25% of their league, 589 were in the middle range, and 228 were in the bottom 25% of their league.

            At age 30 there were 881 players in the study, of whom 231 were in the top 25% of their league (not in OPS, but in OPS relative to their career norm, league-adjusted), 463 were in the middle range, and 187 were in the bottom 25% of their league.

            At age 35 there were 278 players in the study, of whom 59 were in the top 25% of their league, 111 were in the middle range, and 108 were in the bottom 25% of their league.

            At age 40 there were 33 players in the study, of whom 7 were in the top 25% of their league in OPS relative to their career norm, 9 were in the middle range, and 17 were in the bottom 25% of their league.

            At age 42 there were only 7 players in the study, of whom 6 were in the bottom 25% of the league in OPS relative to career norm.   The other one was Carlton Fisk, who was still hitting well.

            Again, an obvious bias that we could not just ignore.  

            I then took the league rankings from the previous stage, and sorted them by age; that is, Terry Shumpert, who scored at 100.00 in 1999 in the previous stage, was entered for age 32 (his age in 1999) at 100.00.    At each age, the top 25% of the players were recorded as “a”, the next 50% at “b”, and the bottom 25% at “c”.  

            That was as far as I got.   Obviously I would have liked to take park effects out of it, too, but I don’t have the data to do that in this file.    The best I can do is to observe that we may have reason to hope that the distortions from that are not too serious, but certainly there are some problems there. 

            Anyway, at this point I have every season of 250 or more plate appearances from 1950 to 1999 coded as “a” (outstanding for this hitter), “b” (ordinary), or “c” (subpar for this hitter.)   Let me run through the process with a typical case and an atypical case, for purposes of illustration.

            Typical case, Ray Lankford, 1996.   Lankford in 1996 hit .275 with 21 homers, 86 RBI, .852 OPS, career high in runs scored of 100.   Lankford’s career OPS was .841, so Langford in 1996 was +.011.   

            1996 was a fairly good hitter’s year, however, so Lankford’s +.011 ranks 167^th among 284 regulars that season, for an OPS Percentile of 41.3.   

            Lankford in 1996 was 29 years old.   Among all 29-year-old regulars, Lankford’s 41.3 percentile rank comes in 632^nd among 1004 players.   That’s a percentile rank of 37.1.  

            That’s in the middle group.   Lankford exceeded his career OPS, but by only 11 points.  When we factor in that it was a hitter’s year and Lankford was in his prime, it’s actually not a particularly good year, but on the other hand, it’s nowhere near qualifying as a bad year.   It’s just an ordinary performance, isn’t really going to tell us anything about his hitting coach, who at this point I don’t know who was; I am doing this part of the study without identifying the hitting coaches, trying to avoid any bias there.

            Atypical case, Henry Aaron in 1976.   Aaron hit .229 with 10 homers, 35 RBI—obviously not good numbers for Henry Aaron.   His OPS that year was .687; his career OPS was .930, so he’s -.243.   It’s actually the 11^th-worst relative performance in the study, just missing the “poor seasons” group I showed you up top.

            1976, however, was not a hitter’s year.   Aaron ranks 249^th out of 250 players that season, ahead of Willie McCovey, for a percentile rank of 0.4.   Among the seven 42-year-old players in the study, this is a fairly typical performance; five of the seven 42-year-olds are in the bottom ten percent, and three are in the bottom one percent.  Aaron ranks sixth out of seven.

            I told you before that I ranked the hitters by age the same as I had by year, but that’s not entirely true.   The formula I introduced before was batter’s rank minus one, divided by group total, minus one.  Aaron, 6^th out of 7, would score at 16.7.

            However, that formula doesn’t work very well with small groups of hitters.   I had only two 18-year-old regulars in the study, Ed Kranepool and Robin Yount.   Kranepool was 148 points under his career OPS, Yount 153 points under.   By this formula they would rank 1 and 2, and whoever was first (Kranepool) would rank at 100.0, and whoever was second (Yount) would rank at 0.0, even though their relative performance was essentially the same.

            To prevent that from happening, I changed the formula for small groups (less than 100 players) from

            1 -  (BR – 1) / (GT -1) 

            To

            1 - (BR +2) / (GT +5)

            Basically, what we’re saying here is that in small groups, we’re going to regard most seasons as being typical.   To do this, we’re adding three imaginary players to the top of the group, and three to the bottom of the group.    Aaron is 6 out of 7.   This changes his percentile score from 16.7 to 33.3.   This changes Aaron from group c (poor season) to group b (normal season).

            This adjustment only re-classifies about fifteen players out of the 11,000-plus, but Aaron is one of the fifteen—Aaron, Kranepool, Yount, and a few other 19-year-olds and 40-plus year olds.   In plain English, our conclusion is “Yeah, it’s a lousy year, but given the standards of the season, given that Aaron was 42 years old, it’s reasonably near what you’d expect, and we’re not going to ding the hitting coach for that.”  

            One more example:  Tony Gwynn in 1997.   Gwynn in 1997 hit .372 with 17 homers, 119 RBI. . .a hell of a year.  His OPS was .970.   His career OPS was .848.   Gwynn was +.122. 

            1997 was kind of a hitter’s year relative to surrounding norms, but even so, Gwynn’s +.122 ranks 14^th among 285 regulars in 1997 (number one being Jeff Reed, who was +.225.)   Gwynn has a percentile rank of 95.4.  

            He was 37 years old.  Among all 37 year-old-hitters, Gwynn’s 95.4 ranks fourth, behind Henry Aaron, Andres Galarraga and Jim Eisenreich.  There are 132 37-year-olds in the study, and Gwynn ranks fourth among them, so that’s a percentile rank, for the age, of 97.7.   It’s an outstanding season.   Kudos to the hitting coach, whoever he was.   I’m sure Tony Gwynn himself had nothing to do with it. 

            Anyway, as I said before, at this point I have every season of 250 or more plate appearances from 1950 to 1999 coded as “a” (outstanding for this hitter), “b” (ordinary), or “c” (subpar for this hitter.)   I then changed those codes to “3” (outstanding), “2” (ordinary) and “1” (subpar), because numbers are easier to evaluate than letter codes.   These we will call Batter Performance Categories.  

            I then created a version of the file in which I had stripped out everything except the player’s name, the year, and the Batter Performance Category.   This is what is left for Mike Schmidt when we’re done:

             Some players—Aaron—have mostly “2s” with a few “3s” and a spotted “1”, and some players, like Willie Stargell, are mostly “1s” and “3s”.    It’s not a perfect system, but nobody would reasonably question that Mike Schmidt had a very disappointing year in 1973, a great year in ’74, a poor year in ’78, great years in ’80 through ’82, and a poor year in ’88.   People would argue about ’86, which was Schmidt’s last MVP season.   We score it at 70.4, meaning that it’s a good season, but not quite good enough.  

            (Another brief digression.  I have commented before on the concept of players who won MVP Awards while just having basically their normal year.  Here, then, is a system by which one could identify and list those players, if you took a notion.   Schmidt in ’86, Yogi Berra in ’51, ’54 and ’55.)

            I’m happy to share the Batter Performance Category file with you. . .I’d post it here, but it would run 50 pages even five abreast, and I know from experience that our programmers have a hard time getting files that large to display.  Maybe we can hook it up to a download button. 

            Our next task is to tie hitters to batting coaches.  Let’s start with Charlie Lau.  Lau was a bullpen coach with the Orioles in ’69. . .won’t count that. .. hitting coach with Oakland in 1970, with Kansas City 1971-1978 (he was fired a couple of times in mid-season, but we’ll count them anyway), was a coach with the Yankees 1979-1981 (we’ll count it) and was a coach with the White Sox 1982-83 (we’ll count it.)  

            The 1970 Oakland A’s had ten players with 250 plate appearances:

             By their career norms, we count Campaneris, Duncan and Rudi as having outstanding seasons, Green, Alou, Fernandez and Reggie as having poor years, and Mincher, Bando and Monday as having normal years.  So he’s off to a weak start here. . ..3-3-4.   (The A’s two catchers, Frank Fernandez and Dave Duncan, have nearly identical on-base percentages and slugging percentages.  However, this counts as a poor year for Fernandez, who had a short career but had a .399 on base percentage the previous year, but a strong year for Dave Duncan, who had a career OPS of .636.)

            By my count, Charlie Lau in his career worked with 142 players having seasons of 250 or more plate appearances.  Of those 142, 39 had excellent seasons, 72 had ordinary seasons, and 31 had disappointing seasons.   Here’s a year-by-year summary:

             The percentage is based on a simple formula, which you can decipher.   Of course, whether this system is fair and accurate is certainly a debatable point, and here’s an argument on the “not” side.   Lau became famous as a hitting coach in large part because of his closeness to George Brett.  Brett gave Lau much credit for his success.   When Lau was fired in early season, 1975, Brett ripped the manager in the newspaper for firing Lau.   The manager was fired shortly thereafter, and the new manager brought Lau back, only to fire him two or three years later.  

But by our study, Brett never had a good year for Lau.   His good years were 1979, 1980, 1983, 1985, 1988 and 1990—all of them after Lau was gone.   Under Lau he had poor years in 1974 and 1978, ordinary years in 1975, 1976 and 1977.  

If, in fact, Lau’s tutelage made George Brett the great hitter that he became, or even if it helped make Brett a great hitter, even at a low level, then Lau’s actual record is much better than it appears here.   In our system, Brett’s greatness as a hitter, after Lau was let go, makes Lau look bad.   You can argue it either way.  Brett was regarded as a great young hitter when he was with Lau, but he did have his best years after Lau was axed. 

            In my opinion, there is probably some truth on both sides of it.   Lau’s approach, more or less, was to make everybody a singles hitter.  Brett won a batting title under Lau in 1976, but he did it while hitting only 7 home runs.   Brett was very strong, and there was really no reason for him to be hitting 7 home runs a season.   I think Brett probably did learn things, working with Lau, that helped to put him in the Hall of Fame, but at the same time, I think he probably reached a point at which Lau’s approach was holding him back.  I think he probably knows that now, although he is not going to say so because this could be considered speaking ill of his old mentor.   This system counts George Brett as a vote against Charlie Lau, and that’s probably unfair. 

            Let’s do Walt Hriniak; Walt Hriniak is to Wade Boggs and George Brett is to Charlie Lau.     Hriniak was the first base coach for Montreal in ’74-’75 (won’t count it), the hitting coach for the Red Sox from ’76 through ’88 (including a couple of years as a dual-assignment coach), and a coach-without-portfolio for the White Sox from 1989 to 1995 (we’re counting it.) 

            Hriniak’s record is better than Lau’s.  There may be a park effects problem here, but I count 62 hitters having excellent seasons under Hriniak’s guidance, as opposed to 43 guys who struggle.   Year by year:

             Perhaps the most prominent batting coach now is Rudy Jaramillo of the Rangers, who once again this year scored a very large number of runs.   We can’t evaluate Jaramillio’s full career by this method, since it requires that we put each season into the perspective of the player’s career, and there is no way to do that with recent seasons.   But Jaramillo’s early years (1990-1993, when he was with Houston, and 1994-1999 with the Rangers) don’t actually score real well:

             Actually he scored very poorly up until 1995, very well in the late 1990s.   Jaramillo has a “George Brett problem” with Jeff Bagwell.    Bagwell had what were considered at the time very good years from 1991 to 1993, as a young player.   He jumped up to a much higher level, however, as soon as Jaramillo left—just as Brett did when Lau left--causing his early years to score as disappointing seasons.   Which may be unfair.

            I was going to do a few more of those, but I’ve run out of time, and in any case this is just a kind of sketch of how a system might work, as opposed to an actual effort to evaluate batting coaches.    I’ll upload the file, we’ll post it and you can download it if you want to play around with the data.  Thanks.