“That is why I have such a hard time with defensive metrics, for some players they are all over the place. UZR has not liked Hudson for four years while +/- shows him being excellent until last year.”

The two systems are a lot closer than you are giving them credit. You just don’t seem to understand what each represents. What ever they represent, they will always be better than your (or anyone possibly short of an MLB scout) “subjective opinion” since your subjective opinion does not involve enough data nor are your heuristics sufficiently tuned from experience (in the way an MLB scout may be).

UZR/150 is a rate stat. It tells you how many runs above or below average the fielding gained or cost the players team for each 150 defensive innings. It is NOT position adjusted.

+/- is a counting stat. It tells you how many plays the player gained or cost its team above average. It IS position adjusted (as the average is referring to average of that position).

Let’s take just one year – and I’ll pick it at random and you can complain about that and suggest another year and I’ll do the same for it.

2007:

UZR/150 – 1.3 +/- – +20

Ok, lets now convert both into a position adjusted counting stat for wins above average.

We’ll start with +/- since it is easiest. Basically, every play at 2B converts a single into an out. So, being +20 plays over average = 20*.475 = 9.5 runs/10 = .95 wins above average.

Now UZR/150. First, we convert it to a counting stat. Hudson played 137 defensive games in 2007, giving him a UZR of 1.1. Now we add in the position adjustment. A second-baseman gets 2.5 runs/150 defensive innings, for 2.5/150*137 = 2.3. Add that back to his counting stat runs and we get 3.4 runs/10 = .34 wins. So UZR/150 and +/- are separated by a little over 0.5 wins.

BOTH systems would agree that players with in about a win are equivalent. These two systems, therefor, rather than being “all over the place” are in complete agreement – at least as it comes to 2007. You can complete (or if you can’t – you can complain and I’ll complete) the analysis on the other years and most likely find they too are in close agreement.

I think you would find general agreement by the creators of both systems that they would place their margin of error near 1 win (10 runs). So yes, if you have a true improvement from player X to player Y of 0.61 wins, that is significant. But if you have two players who measure within 0.61 wins of each other and the measurement of uncertainty is approximately 1 win, then those are essentially the same player.

I don’t think what Phil said was mere overstatement. It was substitution of data with completely valueless opinion.

There is nothing remotely interesting about this statement: “I, Phil Gurnee, consider Orlando Hudson one of the best defensive 2B.” And, to be very clear, that is not a comment on Phil. Equally uninteresting is this statement: “I, Paul Scott, consider Orlando Hudson one of the best defensive 2B.”

Further, as to that statement, no matter what defensive metric you use, that statement is not supportable. Phil claims to see a very large difference over the last four years between +/- (“being excellent”) and UZR (“does not like him”). While I disagree (for the reasons posted above) about the degree of this purported departure, one this is very clear – neither system considers him “one of the best defensive 2B ‘ever seen.’” To the contrary, both “over the last four years” rate him as above average – that is certainly good, but hardly “best ever seen.”

What Phil is attempting to do is first, discredit defensive metrics generally and then assert that because defensive metrics fail (if you accept his first argument) he can substitute his own subjective judgment on equal or better footing. This logical fallacy is untenable. It also leads to a valueless blog, as no one should be interested in post after post of “this I believe.”

… was lead by Counsell at +35. Hudson was tied for 3rd with +20. Not sure why it is not online, but it is in the FB book.

the rankings granularization may hide it somewhat, but +-1 (like UZR) does not find Hudson to be the “best ever seen” over the last 4 years.

Now, from 2003-2005, he is, by +/-, the best. +/- has him at +77 over that time with his closest competitor (Marcus Giles and Adam Kennedy) at +38. More recently, however, it’s Chase Utley that +/- sees as “clearly the best” with 2006-08 having been dominated by him at +85 with Hudson tied for fourth at +29.

don’t confuse any of this with me being down on Hudson. I am not. I just strongly object to:

1. replacing definable, arguable metrics with completely useless value judgments; and
2. suggestions that defensive metrics are useless because they lack agreement when no effort was made to actually analyze the situation.