However, once everyone’s UZR is calculated relative to those 4 year samples, each year is “zero’d out” per position. So, as Eric says, a player’s UZR is always relative to the pool of players in both leagues for that year only.

I did not have to do that. I would have left everyone’s UZR as relative to the 4 years that I use as the baselines for each bucket. That actually would be much better if the data did not have biases in some years as compared to others. But I am afraid it does, so that I am forced to make a “Sophie’s Choice” (not quite as serious of course) and I chose the zero everything out per year. Remember that each position is zero’d out so that for a “Rollins effect” to exist for a particular SS (like, say, Rollins) only the SS overall talent has to change from one year to another. And that is easy to do because we are only dealing with one position out of 7 or so.

And it is certainly possible that one can estimate the relative strength of each position from year to year. One can do as Colin suggests for each position, making the appropriate age adjustments of course (some positions lose as much as 1 to 2 runs a year per player per season with age). One can also look at who arrives and who leaves. For example, if for SS, every player had the same number of chances in year X+1 as in year X, and nothing else changed, we could assume that the SS position got worse overall since everyone aged by a year. Let’s say that Jeter retired after this year and some slick fielding SS from the minors “took his place” and those were the only changes. Other than the aforementioned age considerations for all the other players, we could infer that the SS position got better. Etc.

So, yes, Eric makes a very good point that one should be aware that all UZRs you see are relative to league average (both NL and AL combined) at each position separately. Because of that, one has to be careful when comparing players from different years and even when combining years for the same player.

We actually have the same problem with other metrics, but the chances of an entire league changing from one year to the next are not nearly as great as one position or another. On the other hand, how do you think the AL got so much better than the NL, pitching and hitting-wise, just a few years ago? What I mean is that even overall league-wide hitting and pitching talent can change fairly significantly from year to year.

Just to make one more thing clear. You are going to see much more random fluctuation on a player’s UZR from year to year than you will from a change in league-wide talent, so when you see a player with +5, +3, +7, -6, while you can probably infer a league-wide talent increase in that -6 year, on the average, don’t assume THAT much of a talent change. Most of that change is likely to be random fluctuation, secondly, some of that change is likely a change in that player’s true talent, and lastly, the smallest amount of that change is likely to be a change in league-wide talent at that position (an increase in this case).

2) I think it’s generally perceived that major league level hitting is more rare than major league level fielding. Is this true? How rare is the MVP-level hitter/fielder? Average hitter/fielder?

Year AVG_DIFF
2003 0.37
2004 -1.07
2005 0.30
2006 0.64
2007 0.74
2008 -0.01

I’m not seeing any evidence of a Jimmy Rollins Effect here.