There have been some grumblings about position ratings for this 1907 season, so I wanted to explain to league members why players are rated as they are.

Believe me this is one of the most difficult and time-consuming aspects of developing a league disk, and none of the ratings was arrived at lightly. That said, nothing is necessarily beyond criticism, for two reasons. First, our statistical understanding of fielding performance and how we understand it is incomplete, even with more modern eras. Second, an argument can be made, though we have to be careful making it, the reputation of a player as a strong defensive performer may actually depress the statistical chances a player may have to put this defensive skill to the test. This wild card comes into play when we are dealing with a skill the opposing offense may elect to test, such as outfielder or catcher arm strength/accuracy. How much of an impact reputation has is difficult to measure since we do not have the statistics to measure it. Conversely, there may be players who accrue more assists because their reputation (and perhaps actuality) is poor.

Why are players rated in some positions and not others? As Michael has pointed out, players in this era did move around a lot. That said, however, when I developed the disk, I rated the players for every position they played in real life. We actually know where players played in the outfield (even if we do not have accurate breakdowns of their putouts, assists, errors per outfield position), and a corollary of “moving around a lot” is that occasionally they moved and stayed at a position they did not play the previous season. It is the SPB rule that, rated or not, players who have an outfield rating may play anywhere in the outfield. No outfield ratings were eliminated unless the player had at least 400 at bats but fewer than 5 OVERALL games played in the outfield. Infield ratings were eliminated for players with more than 400 at bats but fewer than 5 games played at a position.

Why might a player have a worse rating than I expected?

But he was supposed to be a good fielder with a great arm! Because of relatively low statistical samples, performance may vary wildly from year to year. That said, we can only go with what we have statistically. To cite one example that has been raised, Roger Bresnahan certainly had a wonderful reputation. He also had a strong arm rating the year before (actually, the 1906 ratings were much less well researched than I was happy with and were thus not entirely accurate). However, in 1907, he had the lowest assist rate of any full-time catcher in the National League, and only Mike Heydon had a lower rate in the American League. An assist per game may seem great, but most catchers in the league were doing better. What are we to make of this? I can only see a low rating. One should also note that the Giants had a bunch of young players coming into the league that season, almost all of whom were hackers in the field at the outset of their careers. It is entirely plausible that Bresnahan’s assist rates were depressed by not having people in his middle infield who were reliably in position. Of course, this is something we cannot pin down from the stats but merely extrapolate from the fact that Doyle and others had such evidently poor ranges based on how many balls they were able to get to. Might Bresnahan have done better with a better supporting cast? Certainly. But that is a speculation we just don’t have the stats to support.  There are some indications that he might have earned a higher range rating, but to the extent we can isolate his arm it is difficult to raise it at all.

This leads to a very important point. Fielding performance, though recorded in individual stats, is a reflection of the team. In developing defensive ratings I tried to balance Bill James’ fielding win shares against the fielding values recorded in Total Baseball against more traditional range ratings based on assists/put outs per 9 innings played (I also estimated fielding innings for all players). Also, the most important statistic in fielding is one that has never been kept: balls that became hits because they were not fielded. This is an area where Bill James has done a lot of research, and I have tried to have some of the same concerns played out in my rating system, but we do not have perfect knowledge.

In Diamond Mind, for the game model to work, there also must be at least some Pr rated players at every position and no more than 1-2 Ex full-time players at any position in either league.

For further discussion of the role of reputation in the setting of statistics I refer league members to Tom Tippett’s articles on the subject in the DMB archives, which you can find at the company website.

If you have concerns about a fielding rating and can make a strong statistical argument for changing it, please let me know.  I just wanted to let you know that I hadn’t got to this point without a lot of thought and research.

Chris Williams