I'd be interested in seeing the breakdown for batted ball types for teammates that played next to each other. For instance, if you could show all the infielders of one team and how they did on batted ball break downs. Maybe we could try to see if one doing well at flyballs hurt another.

As someone who knows little about statistical defence analysis, but finds it fascinating, I have noticed that for these types of studies (PMR, UZR etc.) there is alot of analysis on speed, distance , slice, zone, fly, grounder etc. for each BIP, and I have read that we still are not there yet on what we should be measuring. Why do we not do this same analysis for offence (quality of pitch, night / day game, quality of defence behind pitcher etc. etc.) It seems that we are satisfied with applying set wieghts to each event (single , double, HR), and then we move on. Is this same type of analysis for offence being looked at.

I didn't see any account of how the "predicted" columns are derived (probably not paying attention.) I assume they are some sort of "average" as they seem awfully high for "minimum competence" even if you can shake 20 gloves out of the nearest tree?

I'm impressed by how closely these rankings correspond to my intuitive "who's better than who" lists. I'm not sure if that means the system works perfectly already and I'm brilliant and should be offered a scouting job, or we're all victims of the same class of delusion... but I'm inclined to assume that something good is happening here...

David,

Thanks for mentioning my idea, even if you don't ultimately agree with it. ;-) We may be able to find a middle ground.

Yes, Derek Jeter is probably good at fielding pop-ups--his putouts have been high over the years. And there might actually be a benefit in the case of the Yankees because Jeter overs some ground that Bernie Williams no longer can cover. That might explain Bernie's shockingly low rate last year--Derek probably took "Bernie" chances that a centerfielder would normally take. You could also, however, consider the possibility that Bernie's weakness forces Derek to take discretionary chances that increase Derek's rating. Derek might look good because Bernie is trying not to put unnecessary stress on his knees, and so lets Derek every possible chance.

Please tell me if I'm missing something, but I believe the way PMR works on all fly ball in left center is that it determines the respective probabilities that an average shortstop, left-fielder and centerfielder would catch each one, and sums them up. If Bernie let's Derek take more short fly balls than normal, Derek's rating compared to shortstops will go up and Bernie's will go down--but the _team_ performance will be unchanged.

The key question for determining the true quality of fielder is the number of batted balls he converts into outs (a) above and beyond what an average fielder at his position would and (b) that nobody else fielding another position on an average-fielding team would normally be able to cover.

Derek is clearly failing to field ground balls that _only_ a shortstop is in a position to field; my guess is that Derek is catching many pop-ups that would have been caught by _somebody_ (even on the Yankees) 90-95% of the time.

The great thing about your model, David, is that it is the best method yet devised for _measuring_ the extent to which Derek is padding his putouts with discretionary chances.

Just calculate the weighted average probability that the flyouts Derek actually caught would have been caught by somebody at a different position on a normal fielding team. Then compare Derek's "weighted discretionary chances" _rate_ with other shortstops. Let me know if I haven't been clear about this.

Regarding line drives, the sample sizes are so small that I wouldn't pay attention to them, except possibly over the course of a career.

Guzman is interesting. His 2003 PMR last year was OK, his ground ball 2004 PMR just barely below average, his STATS ground ball zone rating is only slightly below average, yet UZR has ranked him for 2001-03 as a terrible shortstop, with maybe only Jeter being worse.

oh, the "we must prove Jeter is worse than he is" crowd is back! Why can't someone at least acknowledge that Jeter is good at fielding popups.

You know, I think these numbers are pretty interesting, but I was kind of wondering what the MOE was. I mean, I watched enough Yankees games last year to reflexively wonder if Jeters plus putouts are because he went into the stands a few times, showed that over the shoulder catching abiity, or becaus of Bernie's weak knees (the least likely). Or is it because Rivera breaks all those bats resulting in little popups to short and second. In any case, the difference between predicted outs and outs is only 11 putouts. Given the many variables available I wonder if that is even outside of the MOE.

I havent had time to read the earlier posts in too much detail but I'd be interested in looking closer at what is considered "In Play". I mean, it seems to be here the fielding version of at bats but much less concrete. My gut instinct suggests there is some validity to the numbers and yet the putout per liner ratio is so small I don't see how the differences are significant in any way. I ask because this will affect the MOE on these numbers.

Eckstein has the worst liner numbers, but that seems so trivial to be between what percentage wise is the difference between a batter who hits .027 and .038. If that was a hitting stat we'd simply note that the numbers are so small that the difference is unimportant. That is a standard sample size issue.

Regarding David's point about not caring how he gets the outs, but does he.

This is the same issue with pitchers and DIPS. The point is to separate the different types of ways to get an out, and see if there's any skill in that metric, or if it's polluted with noise. A pitcher's BB,K,HR numbers are much less polluted than his hits on BIP.

Similarly, if you've got an IF with tons of lineouts and another not, does that relate to skill? Probably not. He's just lucky to be where he was. When you talk about positioning, it's not like Jeter and Eckstein's positioning would be 20 feet different. We're talking about a few steps here or there. On ground balls, this makes a big difference, as running 10 feet or 30 feet for a ground ball is huge. For liners? I don't think the SS positions himself to set up for the liners. Those are probably dumb luck.

Popups have their own issues. Popups that remain in the infield are completely useless. If the fielder doesn't need to move his feet, and he's waiting for 3 seconds for the ball to come, then that's pure luck.

How about charging back to get a pops to shallow OF? Again, most of the time, the OF could also have made the play. Jeter is nice to watch to make that play, for sure. But, does he allow Bernie/Matsui to play deeper, or is he just making a play that the OF could have made? Again, lots of noise in that metric.

Groundballs however have alot less noise than liners and pops.

For this reason alone, you'd rather have a metric that concentrates on GB.

You see, all metrics are observations. But, to try to get a true rate from an observation, you have to regress the performance. And, if you have to regress the pops and liners almost 100%, then they are noise, and should be treated separately.

Until we know the answer, they should remain separate, so we can determine the answer.

(Maybe the Yankees should move Jeter to centerfield and find a shortstop who can field grounders.)

A shortstop who can field grounders? If only the Yankees had one of those! But then moving Jeter would only make sense if they actually needed a new CF, and given the team's offseason moves they clearly don't.

So I don't know what you're smoking, David.

Tango,

Thanks for making my point better than I had been able to.

Tango,

If popups on the infield are completely useless, that would show up in the data as having a very high probability of being caught. If Eckstein doesn't catch a lot of balls that a normal shortstop would catch, that says one of two things

1. He's not getting to the ball.
2. Someone else is taking charge.

Either way, it's a negative about his fielding. I have not problem breaking down the data, but if David is not catching pops that other shortstop do regularly catch, it's a negative that should effect his overall numbers.

If Eckstein is not catching popups because his LF is more likely to take charge, that should not hurt his rating.

He should only be penalized when a ball that a shortstop could catch falls for a hit.

Can you compare popup and line drive conversion rates for 2003 to 2004? Is there a correlation? I'd be happy to test the correlation if you send me the numbers.

I have to agree with Tangotiger, Michael Humphreys, and rallymonkey. Infield pop-ups only indicate whether or not the IF likes to call for the ball and not his skill level. The only "flyballs" that should be included are the ones that fall for a hit above a certain level. I'm not sure what that level should be but it should be a lot higher than infield pop-ups. Then you could use that same "not caught" level for flyballs in foul ground as well.

The line drive stats seem more like noise than anything else to me. It just doesn't make any sense that Guzman's skill level at line drives is that much better than an average shortstop. He recorded 41% more outs on line drives than expected.

If you posted just the groundball stats for infielders I think it would give a much more accurate reading of their true range. It wouldn't give a reading on how they are at fielding pop-ups or line drives but it doesn't look like your system is doing that now anyway.

i'm not gonna get into the stathead argument other than to go on what i've seen; jeter's belaboring as an ss is overdone, and his ability to field pop flies outdoes anyone's i've seen; but this isn't about that part of a move to cf, it's about what it would do for nyy; he can run, has great instincts, and would remind many of us in the transition of robin yount's successful switch; i like it because i love nyy and i would think it would allow them the flexibility of finding a player at several positions; i don't think it would necessarily mean shoving arod back to ss, but that would certainly be a possibility for a coupla years; i'd like a guy like huff, who could later go to 1b if necessary, and there are others out there who could strengthen nyy...but the point remains, it would be a good thing made possible by dj's ability to become a plus centerfielder; one more thing: it's moronic to post that 90-95 percent of the pops jeter has nailed would be caught by others, particularly since lf has been an adventure for most for years until matsui and is still THE toughest of lefts around...