Suppose that we take two players, for whom we have no individual batting or fielding statistics.  Player A is Andy Alomar, and Player B is Bobby Buttscretcher.  Andy Alomar plays shortstop and hits third for the best team in the league.   Bobby Buttscretcher plays first base and hits 7^th for the weakest team in the league.  Who would you guess is the better player?   Obviously, Andy Alomar is the better player, but what are the odds?  Is there ANY chance that Bobby Buttscretcher is actually the better player? 

            Let me ask another theoretical question.   What if it was possible to compare players just as accurately without any reference to their individual batting and fielding statistics as by doing it WITH reference to their individual batting and fielding statistics?   Can you see the power of that idea?   It would be useful, particularly, in scouting.  In scouting amateur players, their batting and fielding statistics are of limited use.  I promise you:  on almost every college team in America in a more typical year, there is a first baseman who is hitting .350 or .380 and who is first or second on the team in home runs—and no major league team has ANY interest in signing him.   The difference between hitting .380 and .315 is not all that reliable or all that significant as the player moves to more difficult competition. 

            This thread developed—as actually many of my ideas have developed—when I was just messing around, not actually working, but creating Mythical Careers.   I was using my Game Log file to create Mythical Careers, and I thought. . . well, what if I create a player who always hits sixth and always plays third base.  What would he look like? Randomly chosen games from 1950 to 1967, but let’s assume that the player always bats sixth and always plays third base, and let’s assume that he has either a strikeout or a walk in 80% of his games. 

            From that, it occurred to me that one could create a "Batting Order Position Strength Indicator" for every player in my data based on where they hit.  Suppose that:

            If a player hits cleanup in the game, we score that at "10",

            If he bats third, we score that at "9",

            If he bats leadoff, we score that at "8",

            If he bats second, we score that at "7",

            If he bats fifth, we score that at "6",

            If he bats sixth, we score that at "5",

            If he bats seventh, we score that at "4",

            If he bats eighth, we score that at "3", and

            If he bats ninth, we score that at "1". 

            But these numbers apply ONLY if he is in the starting lineup.   If he enters the game late, then he scores at "2" if he has more than one plate appearance in the game, and "1" if he has one or fewer plate appearances in the game.  

            So he can score for a game at anywhere from 1 to 10, based on how crucial a batting order position he is assigned for the game.   In my data I have batting lines for 2,890 player/seasons, and there are 11 players who score at "100" for a season, meaning that they batted cleanup in every game of the season in which they played.  Those 11 players are Joe Medwick (1935, 1936 and 1937), Hank Greenberg (1936 and 1937), Ralph Kiner (1951), Hank Sauer (1952), Rocky Colavito (1965), Joe Torre (1971), Andre Thornton (1984) and Fred McGriff (1995).   Seven of the 11 finished in the top 5 in the league in MVP voting, the exceptions being Hank Greenberg, 1936 (injured, appeared in only 12 games), Ralph Kiner (10^th in the MVP voting), Andre Thornton (19^th) and Fred McGriff (20^th). 

            You might notice that those are all really good hitters.  Three of them won the MVP Award in that season, and seven of the eleven are in the Hall of Fame.  The lowest score in my data, for a player appearing in 100 or more games, is 10.58, by Brian Downing in 1975.  Downing appeared in 138 games that season, and batted 9^th in 135 of them, batting 8^th twice and 7^th once.  The second-lowest score, 11.55, was by Mark Belanger in 154 games in 1973.

            Noting, of course, that a lot of great hitters have hit 3^rd or 1^st, it is a safe and accurate generalization that the best hitters occupy the key spots in the batting order with great consistency.    The career data is actually more interesting.   In my game logs I have data for 168 players who played in 1,000 or more games, plus 28 more who had shorter careers.  Among those with at least 1,000 career games, the highest and lowest  "career batting order position strength indicators" are these 20:

            I would argue that while the Batting Order Positional Strength Indicator is not a PERFECT measurement—that is, Del Ennis is not a better hitter than Hank Greenberg—it is nonetheless a near-perfectly RELIABLE indicator.   That a weak hitter is used in key batting order positions throughout his career, or a strong hitter used in less crucial positions, never happens.  The largest limitation to that generalization is catchers.   Probably most of you know that, in the early history of baseball, once the tradition of the pitcher batting 9^th was established, the tradition actually was that the pitcher always batted 9^th and the catcher always batted 8^th—even if he was a good hitter.  While there were some limited exceptions to that, Mickey Cochrane really broke that practice in the early 1920s.  But what I did not realize until now, until doing this study, is that the practice of burying the catcher low in the batting order persisted, really, until the 1960s.   Ernie Lombardi, Roy Campanella and Bill Freehan have much lower batting order positional strength scores than you would anticipate that they would. 

OK, so that’s offense.  Now let’s do the same for defense.   Let us say that a player is credited with:

10 points if his defensive position for the game is Catcher,

9 points if it is Shortstop,

8 points if it is Center Field,

7 points if it is Second Base,

6 points if it is Third Base,

5 points if it is Right Field,

4 points if it is Left Field,

3 points if it is First Base,

2 points if it is Pinch Hitter, and

1 point if it is DH. 

DH is lower than Pinch Hitter, because pinch hitters are usually position players who are just not in the lineup.  If you’re in the lineup as a DH, consistently, then you’re not highly regarded as a defensive player. 

You can’t switch batting order positions in a game, but you can switch defensive positions, and this requires us to create rules for players who switch defensive positions in a game. . . not interesting, but we have to do it.  

(1)   A player who plays two defensive positions in a game is scored by the average of the values for the two positions, rounded down (so that if you play left field (4) and right field (5), it scores as 4.) 

(2)   A player who plays three or more defensive positions in a game is scored only by the first two.

(3)   If a position player pitches, we score that as "10", the same as catcher.  If, for example, you play 1b,p (first base and pitcher), that would be 6.  First base is three, pitcher is 10, average 6.5, rounded down is six. 

Let’s list the Top 20 and Bottom 20 (among my 168 players) by DEFENSIVE position score:

            These are career numbers.  Again, you can easily see that the players who had more defensive value, the players who played the key defensive positions, are listed highest, and those who were DHs, first basemen and left fielders have the lower scores.  They are arranged by position, of course, but even by position, the players scale by defensive quality.   There are eight catchers on the top-defenders list, but the list starts with Roy Campanella and Jerry Grote, who were great defensive catchers, and works down to Ernie Lombardi and Ted Simmons, who were catchers, but not great defensive catchers.   There are ten shortstops on the list, but Luis Aparicio is the FIRST of the shortstops.  

            Suppose, then, that we combined the offensive and defensive rankings, weighting them each at 50%:

            Again, those are career numbers.  As you may know, I am not a Thurman Munson fan, but we’re not done here.   There are a lot of really good players on the low-scoring list, because my file of 168 long-term regulars is mostly composed of good and great players.  But the "good list" above contains 11 MVP Awards and 13 Hall of Famers, while the "weak list" contains no MVP Awards and no Hall of Famers.     

            My next observation, trying to shape this into a more reliable rating system, was based on single-season numbers, which, for the sake of brevity, I won’t show you.   While the system did give very high scores to the Most Valuable players in many seasons, particularly when a catcher or shortstop won the MVP Award, it gave relatively low scores to first basemen and left fielders within my game logs who had MVP seasons, like Hank Greenberg, Dick Allen, Harmon Killebrew, Orlando Cepeda, Steve Garvey and Willie McCovey.  

            The reason for this is obvious.  Baseball is 50% offense, 50% defense, but the pitchers account for probably more than half of the defense (run prevention.)  If you assume that defense is 50% the responsibility of the pitcher, then 67% of position players’ value is in batting (50/75).   If you assume that defense is 60% the responsibility of the pitcher, then position players’ value is 71% in batting (50/70); if the pitcher is 70% of defense, then the position player’s value is 77% in batting (50/65). 

            I started adjusting the weight given to offensive and defensive position, trying to make it match better with MVP voting and WAR.   The ratio that seemed to work best was about 3-1, three parts batting to one part fielding. 

            But there is a third element in this, which is the performance of the team.  Bobby Bonds in 1981, his last season in the majors, played center field and batted first or third in two-thirds of his games, thus scoring at 78 offensively and 78 defensively, but he did this for a terrible team, a team which finished 38-65 in the strike-mangled season.   If you play center field and bat leadoff for a great team, that’s one thing; if you do it for a terrible team. . .well, somebody has to play center field, and somebody has to bat leadoff, but it doesn’t necessarily mean as much. 

            By trial and stupidity I stumbled onto a system in which I evaluated each season by:

            The Batting Order Positional Score, times 6, plus

            The Defensive Positional Score, times 2, plus

            A number 1 to 20 based on the winning percentage of the team, as follows.  If the team has a winning percentage of .680 or higher, 20 points; .660 to .67999, 19 points; .640 to .65999, 18 points; .620 to .63999, 17 points; .600 to.61999, 16 points. . . .etc., down to .320 to .33999, 2 points, and less than .320, 1 point. 

            Under that system, the highest-scoring seasons in my Game Log data are the following: 

            It’s not a perfect measure of the quality of those seasons or all of the seasons in my data, but I believe there are 6 MVP seasons in the top 13—with a system that makes no direct use of batting or fielding statistics.  You would be hard-pressed to beat that with a system which DID use batting and fielding statistics.   There are 1,740 seasons of 120 or more games in my Game Log Data.   Here is a chart that shows the number of MVPs in each range of Positional Strength Scores:

            21% of players who scored at 80 or higher in this system won the MVP Award, versus 1% for those who scored at 74 or lower; actually, just a little better than one-half of one percent.  For a career, Yogi Berra has the highest Positional Strength Score in my data—a cleanup hitter, a catcher, playing for formidable teams.   He might have the highest score ever, even when I get the Willie Mayses and Ted Williamses and such like included in the data, I don’t know.   These are the career Positional Strength Scores for all 196 players within my data, with the Hall of Famers highlighted in Gold:

            Of the top 40 players in Positional Strength Score, over half are in the Hall of Fame—21 of 40.   Of the second 40, 11 are in the Hall of Fame; of the next 40, 5 are in the Hall of Fame, and of the bottom 76, 1 is in the Hall of Fame.   For a system that doesn’t have ANY knowledge of who hit .320 with 500 homers and who hit .215 with 16 career homers, that’s a pretty good separation. 

            My main task now is to explain what could be done to make this system more valuable than it would be in its current form.   Before I get to that, though, I would like to point out a couple of very neat things about this system. 

1)      It tells you whether a player’s value is mostly in his offense or in his fielding.  Since the batting order position points are stated on the same scale as the fielding position points, a player whose value is mostly as a hitter will have a higher batting-order score than fielding-position score, and vice versa.   Few other systems would do that, because few other systems (if any) state offensive and defensive value on the same scale. 

This is a list of the 20 players whose value is most heavily dependent on batting, among those in my data:

Most of those players were DHs and very good DHs, some playing first base.   And here the opposite list, players whose value was in their fielding: 

Brian Downing appears near the top of both lists.  Of course, Downing had a very unusual career, starting out as a pudgy catcher who hit ninth, and transitioning gradually to a muscular slugger/DH.   I don’t know how many people now even remember him.   Roy Campanella appeared on the "defense only" list before his MVP years, which is a fluke and an error by the system.  The Dodgers having many good hitters, Campanella batted low in the order for several years before moving into the middle of the order, where he won three MVP Awards.  And this is a list of player/seasons in which the value was evenly distributed:

2)      Another thing which is neat about this system is that its center is locked in place over time, which eliminates the need to make a lot of adjustments that would otherwise be needed.   1927 Yankees or 2004 Tampa Bay Devil Rays; they both have a leadoff hitter, a cleanup hitter, an eighth place hitter, a catcher, a shortstop, and first baseman.   Of course, the DH screws with us a little bit. 

OK, turning our attention now to the issue of things which could be done in the future to make this system work better than it does. 

One of the things which defines sabermetrics, one of the things which distinguishes sabermetrics from other sportswriting, is that we revisit issues again, and again, and again.  Most of the article that I write—not this one, but most of them—are second looks and third looks and fifteenth looks at things I have studied before.   This is not ideal, from the perspective of the reading public.  I often feel that I should apologize for writing again articles that I have written before, but that is who we are.   We move the ball an inch at a time.  We pick up things we have studied before and study them again, and we try to make progress. This was a first look, but to have real value, it will require a second, a third, a fourth.  

I would suspect that the number one thing that will need to be adjusted is to get better values for the positions.  I think the order of the batting order positions (and the order of the fielding positions) is about right, but the relative values need to be refined.   It may be that the proper scale goes 20-19-17-12-10, rather than 10-9-8-7-6.  

It doesn’t seem right that a catcher and cleanup hitter HAS to rank ahead of a shortstop and leadoff hitter.   It seems there should be some leeway somewhere, some tolerance, that enables the rankings of top players to vary a little bit.  Yogi Berra was a great player, but not necessarily the greatest player in this group of 196.

If we were concerned with (1) rating seasons, or (2) rating careers, then we would need to weight the player’s performance by his playing time in the season or the length of his career.   Kirby Puckett and Thurman Munson would move down, because their careers were short, and Rusty Staub and Darrell Evans and Ernie Banks would move up because their careers were long.  

That would not be too difficult to implement, but that isn’t really the potential value of the system.  The potential value of the system is in scouting.  The value is in looking forward, not backward.   It’s a common thing.  A scout wants to know one thing:  if there is one player in this league who can play professionally, who is it?  

This, I think, is the system that could tell him that.  If you know who played the key defensive positions and who batted in the key spots in the order and who won, you have a pretty good picture of who the best players are.   High school batting stats don’t mean anything, and don’t tell you anything.   This is the system that, if tracked in an organized manner, can tell you who in a high school league, or who in a college league, or who in a low-level minor league, has the ability to play at a higher level.

To make that work, what I think you would need to do is either (a) replace the 20-point "team performance" system with a mixed system, half team performance and have league quality, or else supplement the system by including a league quality indicator.  It doesn’t have to be anything organized or sophisticated; just a number that tells the system what the observer already knows.   You’re looking at kids from 200-person high schools in Idaho, you know the quality of competition isn’t the same as 1500-person high schools in Texas or Florida or California.   You just need a way to tell the system what you know, and make it work with the rest of the facts.   Given that, this is potentially a way of helping the scouting director organize his reports.

If you were using this system as PART of a rating system for historical players, one thing you could do would be to add ten points for "timeline", and increase that with the passage of the years.   A player playing 1900 to 1909 might be "1", a player playing 1910 to 1919 might be "2", a player playing 1920 to 1929 might be "3", and a player playing 2010 to 2019 might be "12". 

Going back in history, the best players dominated to a greater extent than they do now, simply because there were fewer good players.   Honus Wagner, Ty Cobb and Babe Ruth dominated their game to a greater extent than Willie Mays and Mike Schmidt dominated THEIR game; Barry Bonds may be an exception to this, and we all know why so we don’t need to say why.   But if you put in the 1-2-3-4-xxx-12 for the strength of the decade, then that compensates Mike Trout, which enables the later players to be rated against the earlier players, without either one dominating. 

Thanks for reading.