This, coupled with the hyper-attention given to defense, which in my opinion, has an underlying motive of proping up replacement-level, or AAAA roster building. Measuring every players’ contributions equally defensively and offensively is biased, and doesn’t leave enough room in value for guys that hit incrementally above the norm.

Snarfy, you are a funny cat, but its way too easy to ferret out the lookout landing nerd who is so desperate for even the slightest bit of positive social interaction. Sad.

Anyway…

What we are trying to do here is not to predict Longoria’s next UZR, we’re trying to figure out just how good he is. So, if in one season he’s +20, then he regresses to the mean, maybe to +10, he’s still a +15 fielder. And at that point we’re fairly confident the +15 is repressentative of his skill level. That’s the basic reason why we need to keep that +20. Its part of the information that helps us figure out how good he is. What’s happening here is we’re trying to answer two fundimentally different questions. You are asking, “how well does the +20 season predict the next season?” I’m asking, “how well does a +20 season repressent the player’s true talent?”

Then, why might we keep Andruw Jones’ 5 +20, and Utley’s 2? Because our question slightly different still. There I was wondering how well any +20 season predicts the talent of a player. The fact that about 1/3 (going off memory here) of our + 20 seasons by UZR/150 in 130+ games, is actually a good thing. That means the best players are repeating those seasons, so to take them out is skewing the sample towards the guys that actually just had a lucky season.

The basic idea here is how you specifically frame your question changes how you analyze the data greatly. There are a lot of ways to put these numbers together, and you could do everything right mathematically, but be trying to answer the wrong question and its all worthless. On the flips side, you can put them together in 100 different ways and ask 100 different questions, and be right, everytime. Any, this is a very big problem in my field were researchers analyze the data in all the right ways, but then go on to make conclusions that over reach that analysis, or are even entirely different from that analysis.

So, what you’re doing answers one question. But, I’m not particularly interested in that question, nor do I think it is the question that is relavent to this situation. Regardless, I think the final nail is in the coffin here. I will be interested to see how UZR plays out, and its ability to predict a player’s true skill, as we get more seasons worth of data.

Been fun and educational, see you in the next thread.

I could look at fielders a season between 0 and 1 run and see what that season says about a players talent level. In the available sample, they regress very slightly to the mean. Players with a season between 4 and 5 regress a little bit more – about 2 runs. Players between 9 and 10 runs regress about 4.4 runs. Players between 13 and 17, about 5.5 runs. When you only have one season of UZR, you have to regress to the mean in considering the players true talent. The further from 0 the single season of UZR is, the more runs will come off with regression. When the predictor sample rises above 20, then sample bias kicks in pretty hard in selecting primarily outliers, and regression is even stronger. But apparently you don’t get that (you demonstrated in another post that you have no clue how to regress to the mean or even what it means). I’m sorry. Fortunately, as Longoria continues to play for the next few years, this aspect of analyzing his UZR ratings will become less important, and your pseudo-statistical approach will be less detrimental to your results.

I’ve explained the problems with that at least twice now, and they remain the same for +15, or 13-17 or -2…or any other random numbers between -20 and 20 you want to pick. We aren’t asking the question of “what are the career averages of players with a +20 season when not including that year?” I can’t say it any more clearly than that. We are asking “how representative of a players true skill is a single +20 season, specifically going forward for a young player?” When doing that, we don’t want to throw out ANY season, because ALL seasons are needed to establish a player’s true talent. You can keep repeating yourself, but it isn’t going to change that.

“The whole reason you look at only the next season is to avoid what you are talking about with Rolen (who is not retired, by the way). This was your method. Is the only reason you wanted me to use it so that you could poke holes in what you knew was flawed?”

If you think that’s what I wanted you to do, you missed the point (and no, only looking at the next season isn’t telling for reasons I’ve already stated). I did an admittedly simplistic analysis at how close those +20 seasons where to career averages and stated some problems with that (and continue to do so), in an honest discussion about how to approach this problem. You then modified my simplistic analysis to throw out the +20 season (and repeated +20 seasons from a single player, which again changed the question we’re asking, but fine, we’ll go with that). I’ve stated problems with that, the main one being that, that season counts in establishing a player’s true talent. I then discussed what information we need to gather to really be able to answer this question (aging patterns), which neither of us is likely to do, and we probably don’t have the available data to do it anyway.

“Now you’re going back into the subjective looks at individual players and ignoring the aggregate results. Weren’t you shunning that method earlier?”

What I did was go player-by-player through our entire sample of +20 fielders to illuminate problems with the way we’re doing this analysis (our sample wasn’t indicative of the player we asking the question about), and why we need aging patterns. For example, Vizquel putting up a +20 season at the age of 40, doesn’t tell us much about how likely a 23 year old is in repeating a +20 season, or what his true talent level is. Think of the hitting equivalent. Does Luis Gonzalez posting a .453 wOBA, career best by about .050, at the age of 33, and how well he was able to repeat that, give us much information about how well a 23 year old with a .453 wOBA will repeat that? Aren’t their projected career paths much different, and thus the likely hood of each of them repeating it very different? Now fielding might “age” very differently than hitting, but we don’t really know yet, do we?

Now, its been a fun discussion, and through it I think we’ve learned first hand the limitations with this fielding data, but we’re going in circles. We’re now stuck repeating ourselves (ok, really you are, and my response to that same stance isn’t going to change) over leaving out this one season. If you’re not willing to accept the problems with that approach, and move on, I suppose this discussion is over. Regardless, we don’t really have the data to get around these problems, especially given the simplistic analyses we’re willing to do, and its very possible we don’t have enough historical data on UZR to do the required analysis anyway.

Lastly, to reiterate my final basic point: The young, all-around talented players on our list of +20 fielders were able to maintain these very high levels in UZR (Crawford and Jones, which I would consider better comps to Longoria than the likes of Vizquel or Cabrera.), so I suspect Longoria is a better bet to repeat high level UZR than our sample’s average. You want to slam me for subjective looks? Fine. However, in the absence of conclusive quantitative data, its absolutely proper to make informed and rational decisions or predictions on subjective information (or do you think scouting doesn’t have its place?). I may have “shunned” subjective analysis somewhere (ref?), but at times it has its place and in others it doesn’t.

PS – Keep the irresponsible posts to yourself, please. People come here to become more educated on baseball, not your opinions.

The whole reason you look at only the next season is to avoid what you are talking about with Rolen (who is not retired, by the way). This was your method. Is the only reason you wanted me to use it so that you could poke holes in what you knew was flawed?

Now you’re going back into the subjective looks at individual players and ignoring the aggregate results. Weren’t you shunning that method earlier?