It's park-adjusted for run value at Fenway - not for distribution of balls in play and expected zone coverage based on the park dimensions.

Actually, the park adjustments are for the distribution and "catch rates" of the balls in play. IOW, Manny's data are compared to the data for all LF'ers at Fenway over the last 10 years or so.

Manny did in fact have a great "arm lwts" in 05 (park adjusted). Around +6 runs.

Kotsay's 05 was way out of line with previous years, such that his defensive projection is still pertty good, although not nearly as good as it was going into 05.

There is no "math" or "calc" error for Swisher, but as I said in one of my last posts, there are many sources of potential sample error in UZR. Just because someone's UZR is good or bad does not necessarily mean that he is really a good or bad fielder. Certainly if other metrics (that have somewhat independent methodologies AND add information or at least a certain methodology to UZR) disagree, and especially if reliable "scouting/observation" disagrees, that suggests that the "truth" lies somewhere in between, where in between depending on the sample size of the UZR data and the other metrics, and other "keys", such as the raw speed, agility, and body type of the player in question, etc. In Swisher's case, the proper meld might be 20% UZR and 80% fan scouting - I don't know. In any case, in a large group of players and one-year worth of UZR data, we are going to find X amount of "outliers" by chance alone, simply as a function of the standard error of the UZR measurement process, as UZR is essentially a binomial with a Gaussiam distribution.

Manny's context-adjusted offensive lwts were +48 in 05, around 40 runs better than the average LF'er. That still put him at around league average or less, overall (offensive, defense, etc.) for LF'ers. Given that his defensive projection is not -40 or so (last 4 years, his UZR was -12, -20, -20, -47; as I said, his defensive projection is more like -22), his value next year as a LF'er is still around one win better than the average LF'er, which is good, but that's about it. His real value is as a DH, unfortunately for the team that acquires him for a lot of money and puts him in LF...

if reliable "scouting/observation" disagrees, that suggests that the "truth" lies somewhere in between, where in between depending on the sample size of the UZR data and the other metrics, and other "keys", such as the raw speed, agility, and body type of the player in question, etc. In Swisher's case, the proper meld might be 20% UZR and 80% fan scouting - I don't know.

WELL, COUNT ME AS ONE WHO IS COMFORTABLE WITH A LACK OF OR INSUFFICIENT PRIOR DATA... BAYESIAN TECHNIQUES CAN BE EMPLOYED, AND THE HARDWARE AND SOFTWARE IS CHEAP...


IN SWISHER'S OR THE CASE OF THE LIMITED DATA SET, IT WOULD BE INTERESTING TO APPLY SOME COMBINATION OF FAN SCOUTING, SCOUT "GUT", UZR-ESQUE DATA FOR THE SEASON, AND THEN SOME SORT OF AVERAGE PROJECTION ANALYSIS (SIMILAR TO JAMES' "TOY") FOR THAT POSITION TO EITHER HONE THE DATA SET OR CREATE PROJECTIONS....

I'M NOW EVEN MORE VERY GLAD WE DIDN'T GO OUT AND GET LAWTON, MGL....

WHAT ARE WE GOING TO DO ABOUT THE OF CORNERS AND 2B???!!!! I'M HONESTLY TRYING NOT TO PANIC GIVEN THE DEPTH OF THE STL FARM SYSTEM AND THE FA MARKET...

The problem I have with accepting that Manny Ramirez could be an "average" player overall is that I don't understand, then, how the Red Sox could win 95 games. Just kind of eyeballing the numbers, it seems that in 2005 Boston's pitching stunk, their infielders were tepid hitters, they had assorted outfielders who could hit a little, and then they had Varitek, Ortiz, and Manny, who were great. So how did they win so much if Manny was actually giving away everything he contributed? Is there some procedure whereby you can "balance" all those calculations and account for how the actual Red Sox could get to +105 runs with Ramirez being a wash? I am happy to be tutored :)

he's not an average player overall. so you don't have to accept it. just like jeter wasn't a below average player when MGL's numbers said otherwise.

Just kind of eyeballing the numbers, it seems that in 2005 Boston's pitching stunk, their infielders were tepid hitters, they had assorted outfielders who could hit a little, and then they had Varitek, Ortiz, and Manny, who were great.

Well, one thing is that, if Manny isn't so bad in the field, their pitching doesn't stink so bad.

If that's unclear, I meant "if Manny hadn't been so bad..."

I'll give the NL UZR "GG awards" in the NL GG thread.

Tolaxor, as far as the OF corners and 2B (and pitching), we are working on that right now as we speak. As you know, as long as we have a healthy Rolen, Pujols, and Edmonds, we don't need a whole lot at the other positions to still field a very good offensive and defensive team...

The 2005 Red Sox scored 910 runs while the average AL team scored 771. That's +139 runs, and adjusting for the 101 park factor reported by BB-ref gets you to ... um, +138.

The 2005 Red Sox allowed 805 runs while the average AL team scored 758. With the park factor, that's still -47.

So, outside of Manny, the Red Sox were ~+91 on offense and possibly exactly average on run prevention.

That -47 sure does seem fluky. I'm not saying UZR missed one -- I really have no idea -- but defense at such a low level is pretty hard to fathom, I guess ...

In any case, in a large group of players and one-year worth of UZR data, we are going to find X amount of "outliers" by chance alone, simply as a function of the standard error of the UZR measurement process, as UZR is essentially a binomial with a Gaussiam distribution.

But that's not what is happening here. Your data differs from ZR data by nearly *50* plays made. It differs from DSG by *70* plays made.

That's not an outlier - that's an error in the methodology.