Overrated = Brian Roberts?

[Frobby]

My use of the word "secret" was a comment on the fact that somebody knows the relevant info and won't tell us what the stats say for a whole year.

That is all you really needed to say. I didn't know that's what you meant, and I'm sure Baltimoron didn't either.

[Migrant Redbird]

Now, as you may have noticed, the fact that I don't like the secret aspect of it was just parenthetical. The main point was that we need good-D numbers that are absolute, not relative. What's your opinion about that? ;-)

That we may "need" them, but we ain't gonna get them. There are just way too many factors involved which defy accurate measurement.

I think that the beauty of the +/- system is that they've defined it as objectively as they possibly could. A ball gets hit in play and someone reviews the path of the ball and assigns it a vector, a relative velocity, and a type (fly, liner, "fliner", ground ball). Yes, certainly there is some subjectivity in deciding what is a fly ball and what is a line drive, and the invention of the "fliner" only increases the granularity of the decision which has to be made, but I suspect that, after a reviewer has done it for a while, he or she becomes pretty consistent. (I would also hope that BIS has some procedures in place to assure maximum consistency among different reviewers.)

Once the ball in play has been characterized, then the observer records whether it results in an out or not. It doesn't make any difference whether that was because of an error by the fielder or a defensive alignment that gave the fielder zero chance to make the play -- did the play result in an out or not? Cut and dried! Simple! There may have been a subjective judgment by the umpire who made the call, but there was none in how the observer recorded it.

The example which you gave of Brooks making a play on a screamer down the line just calls for too much subjective judgment to suit me. By the +/- system, only one time in a hundred does a third baseman manage to field that ball and get an out so, if Brooksie makes the play, he gets the absolute maximum +/- rating boost for doing it. Conversely, if Brooks can't handle the play, he loses an extremely tiny amount on his +/- rating if any other third baseman somehow managed to make that particular play that year. We don't make the subjective judgment on whether that play was difficult or not; we let the large volume of data make that judgment for us.

[RShack]

That we may "need" them, but we ain't gonna get them. There are just way too many factors involved which defy accurate measurement.

I think that the beauty of the +/- system is that they've defined it as objectively as they possibly could. A ball gets hit in play and someone reviews the path of the ball and assigns it a vector, a relative velocity, and a type (fly, liner, "fliner", ground ball). Yes, certainly there is some subjectivity in deciding what is a fly ball and what is a line drive, and the invention of the "fliner" only increases the granularity of the decision which has to be made, but I suspect that, after a reviewer has done it for a while, he or she becomes pretty consistent. (I would also hope that BIS has some procedures in place to assure maximum consistency among different reviewers.)

Once the ball in play has been characterized, then the observer records whether it results in an out or not. It doesn't make any difference whether that was because of an error by the fielder or a defensive alignment that gave the fielder zero chance to make the play -- did the play result in an out or not? Cut and dried! Simple! There may have been a subjective judgment by the umpire who made the call, but there was none in how the observer recorded it.

The example which you gave of Brooks making a play on a screamer down the line just calls for too much subjective judgment to suit me. By the +/- system, only one time in a hundred does a third baseman manage to field that ball and get an out so, if Brooksie makes the play, he gets the absolute maximum +/- rating boost for doing it. Conversely, if Brooks can't handle the play, he loses an extremely tiny amount on his +/- rating if any other third baseman somehow managed to make that particular play that year. We don't make the subjective judgment on whether that play was difficult or not; we let the large volume of data make that judgment for us.

So, once there's a large volume of data, why couldn't you use that as the basis for assigning absolute rather than relative values? If the large volume of data tells you that the ball I described is almost always a double, then what's so if-fy about crediting Brooks with saving 2-bases plus an out? Once you have a frame of reference, what's the difference in accuracy about? It seems to me that the large volume of data would permit you to do it just as accurately as the whole +/- thing. Why would the accuracy of the two be any different? Either way, you have an observer doing data-entry about how the ball was hit and where it went, and the machine spits out the answer.

[Migrant Redbird]

So, once there's a large volume of data, why couldn't you use that as the basis for assigning absolute rather than relative values?

If it's the "large volume of data" which makes that determination, rather than the judgment of the observer, then I have no problem with it. It's akin to a discussion I had with Baltimoron a day or two ago about assigning different values to the +/- ratings by position. Of a ball that gets by a 1st baseman ends up being a triple in a particular ballpark and a ball that goes up the middle is only a single, they shouldn't be scored identically.

[RShack]

If it's the "large volume of data" which makes that determination, rather than the judgment of the observer, then I have no problem with it. It's akin to a discussion I had with Baltimoron a day or two ago about assigning different values to the +/- ratings by position. Of a ball that gets by a 1st baseman ends up being a triple in a particular ballpark and a ball that goes up the middle is only a single, they shouldn't be scored identically.

Right. All you want the observer for is to judge how the ball was hit and exactly where it went... plus to make notation of anything noteworthy that the system might miss. The observer just puts that stuff in. You don't want a person trying to remember everything the db knows about what happens when balls are hit like that. That's what db's are for.

As for the other part, I don't know who was taking which side of the argument, but a ball that goes down the line for a double or a triple should definitely count more than a ball that goes up the middle for a single. The value of all D-plays are not equal.

[El Gordo]

I can imagine a better one. Which is not to diss the "+/- Play" idea, I agree that it seems like progress. The main problem I have with it (in addition to it being a secret) is that it's all relative. It's like having OPS+ but with no OPS. What's missing is something that tells you how a guy did, what difference he made, apart from how other guys at his position did. What we need is something that can tell us how each guy did on his own, and then compare different guys based on their numbers about that. That's what we do with hitting and pitching, and that's what we need to do about fielding too.

I can imagine a way to do that, one that would replace all the insane silliness about what is-and-isn't an E with something that is both better and a lot more meaningful. Here's an example scenario. Let's say Brooks is still playing, and somebody hits a screamer down the line that would be a double unless Brooks makes one of his hero-plays, OK?

• If Brooks dives and lands on his nose, but it gets by him anyway, then it's a double, just like God intended. Brooks gets no credit but he also gets no penalty.
• If Brooks grabs it and throws the guy out, then Brooks gets credit for saving 2 bases, plus he gets credit for creating an out that wasn't supposed to be an out.
• If Brooks grabs it and his throw is late (or bad, or whatever) and the guy is held to a single, then Brooks gets credit for turning a double into a single.
• If Brooks grabs it and throws it away, such that the batter winds up on 2B, then it's just like it got by him, Brooks doesn't get any credit, but he doesn't take an E either, because what he did made no difference.
• If Brooks grabs it and throws it away, such that the batter winds up on 3B, then Brooks takes a hit for making everything one-base worse than God intended, he turned a double into a triple.

To me, that seems fair. It certainly paints a better picture about the impact Brooks had on that play than anything else I can think of. Of course, the main problem is in how to do it. But we already accept the idea of trusting the judgment of observers. We trust Untrained Observers for E's. We trust Trained Observers to be umpires. And we seem to be adding a new kind of Trained Observers for +/- Plays. If we're gonna trust Trained Observers left and right, I'd prefer to trust them to tell us what we want to know in real-time, and put it in the box score, instead of waiting for some company to crunch numbers for a year and then sell us the secret results.

EDIT: I think we prolly need a few years of "+/- Play" data as a basis to go by before it would be possible to do this decently.

The problem with this methodology is that no one can say for sure what God intended, except for maybe SG, and we know that he intends for only offensive data to be considered. I think the +- does me just fine, if only I could afford a copy of the Fielding Bible.

[RShack]

The problem with this methodology is that no one can say for sure what God intended, except for maybe SG, and we know that he intends for only offensive data to be considered. I think the +- does me just fine, if only I could afford a copy of the Fielding Bible.

Well, in the very limited context of this particular problem, I guess I'm trusting the db to play the God-role about what a given kind of ball-in-play is supposed to turn into.

That's what the entire +/- system is doing anyway. If you think of each BIP as having a certain kind of force-and-trajectory directed at a certain location, and if we call that combination of force, trajectory, and location the BIP's "profile" (we might say "vector", but vector can mean different things), then the +/- system is based on recording each BIP's profile along with the outcome. That's how their db determines the +/- score of each individual SS relative to all SS's. Once you have a huge db of that stuff, then we can only hope that they've been recording more than just "hit" or "out" as the outcome of each hit-ball (which I'm sure they have been). As long as their db knows which ones are singles vs. doubles vs. triples, etc., then if you tell the db about the profile of the next BIP, it can tell you the probability of that ball having various outcomes pretty precisely.

AFAIK, the justification of releasing the +/- numbers once a year is that they have to construct the whole 2007 db before they know how to say anything about the performance of 2007's players. But that's not entirely true. They have data going back for as long as they've been doing this. They're not really starting from scratch each Spring. For the data-sets they're using for each season's +/- score, yeah, they are starting fresh. But that's not the only thing they can do. They can look at all the data they have from all the years. The +/- results might change some from year-to-year, just like OPS+ does. But I bet there's lots of relatively consistent and continuous data that they could go by. I don't know how many years of that data you'd need to do it right; somebody would need to know what the aggregate data says in order to tell how much you'd need. But whatever that amount is, my completely unsubstantiated guess is that they're probably not far from there by now. Once they get there, then they should be able to give meaningful D-scores to guys just as soon as each game's data has been input, just like OBP and SLG get updated all the time. It wouldn't have to be relative, and it wouldn't have to be just a-year-at-a-time either. If he's selling this stuff to MLB teams, I bet they don't wait a-year-at-a-time to get it.

[Migrant Redbird]

I think the +- does me just fine, if only I could afford a copy of the Fielding Bible.

New and used from $9.99 plus $4 shipping at amazon.com

Problem is, TFB only covers the 2003-2005 seasons. If you want the data after that, you'll need to buy the The Bill James Handbook each year, and that's another $15 plus shipping. Got mine; haven't gotten around to studying it yet.

You can get the leaders and trailers at their website but that's just the tip of the iceberg. Most of the players you'll want to check out won't be listed.

Hey, they got to pay all those game reviewers somehow, don't they?

[RShack]

Hey, they got to pay all those game reviewers somehow, don't they?

They certainly do. I've been kinda assuming that they're selling stuff we don't even know about to ML teams. But I don't know if that's true. (If I was them, that's what I'd be doing, but I don't know exactly what they're doing.)

It prolly won't happen, but here's what I wish would happen: that MLB would just hire them as an information-provider, have them do it for everybody, both ML and MiL, and just put it all out there for everybody to see.

[Migrant Redbird]

Once they get there, then they should be able to give meaningful D-scores to guys just as soon as each game's data has been input, just like OBP and SLG get updated all the time. It wouldn't have to be relative, and it wouldn't have to be just a-year-at-a-time either. If he's selling this stuff to MLB teams, I bet they don't wait a-year-at-a-time to get it.

Well, there might be some debate about it being "meaningful" or not. You have to consider that the +/- ratings are affected by the randomness of where balls get hit, so while you might be able to assess the difficulty of a particular play in a generic sense, it might not mean much for a specific BIP, especially if there's some kind of a defensive shift on.

My impression of the +/- ratings is that they vary a little bit from season to season due to "luck" and varying distributions of where BIPs cluster, so 2 or 3 years of data is much more reliable than a single year's worth. How much then would a single game's data be worth?

[RShack]

My impression of the +/- ratings is that they vary a little bit from season to season due to "luck" and varying distributions of where BIPs cluster, so 2 or 3 years of data is much more reliable than a single year's worth. How much then would a single game's data be worth?

It's not that the single game's data is worth much. It's that the multiple years of data is worth much. Based on the big data set, if you tell the db about the BIP from todays game, it can look up what has happened to all the balls that were ever scored/rated the same. It will see X% outs, Y% singles, Z% doubles, and so on. That let's it assign value to whatever the defender did on today's BIP, and figure that in to his Defensive Goodness scores for the season-to-date. Just like BA, the player's DG scores will be goofy early in the season, and then gradually stabilize into something you can trust as being more-or-less real. Meanwhile, just inputing the score/rating and outcome for today's BIP just adds to the data the db has available to refer to.

[mweb]

Regardless of ones opinion on defensive stats, there is no good reason for every thread where such stats are mentioned to make a point to turn into this. Many threads get derailed due to this reason, and I don't think it's good for the board. So is there anyway we can avoid having this happen the next time someone says player X was +8 in UZR? Is it that hard to resist turning that post into a full on war against defensive stats?

[Roll Tide]

News and Notes: Braves Talking With Teixeira

The Braves and Mark Teixeira agreed to a one year deal today worth 12.5 million dollars, but talks of a long-term deal continue. Both sides confirm that they are moving towards a deal, but no deal appears imminent just yet. I think Teixeira could earn 20 million on the open market if he puts up similar numbers to last year and I think the Braves are hoping he will take a discount. In 132 games last season, Teixeira batted .306 with 30 HR and 105 RBI. Personally, I think Teixeira will stay in Atlanta. Interest will definitely be high for Teixeira, especially from teams like the Mets, Giants, and Orioles, all of whom are looking for first baseman.

Dodgers May Trade Pierre: The Dodgers have four outfielders capable of starting and may look to trade Juan Pierre. Pierre signed a huge deal in 2006 but with the acquisition of Andruw Jones, he is the odd man out. I think the Dodgers will have trouble trading him, mostly because of his high salary.

Cubs Eying Marlon Byrd: The Cubs Convention started this weekend with the big suprise being the signing of former White Sox closer Shingo Takatsu. However, rumors are flying that they are looking to bring in a right handed outfielder to platoon with Felix Pie. A trade for Marlon Byrd might cost them something like Kevin Hart, but that at the most.

Phillies Begin Talks with Howard: The Phillies have begun preliminary talks with Ryan Howard about a long term deal. However, they are expected to table discussions before Spring Training and wait till next offseason.

Rockies Talking with Tulowitski: The Rockies and Troy Tulowitski are working to complete a long term deal that could lock him up through 2013.

Coco Crisp Still On Block: The Red Sox have yet to trade outfielder Coco Crisp but plan to do so before Spring Training. A lot have teams have already filled the void in center so it might be hard to trade him, but I think it will get done. If the Rangers trade Marlon Byrd to the Cubs, there will be an opening for Coco Crisp. Also, the Padres are a possibility if they can get a good deal.

Orioles trades dead?: Despite reports from ESPN 1000 that said a Brian Roberts trade would be done this weekend, it appears that talks have stopped. Once source said both trades are slowing down because of the Orioles 'unwillingness' to make these deals happen. However, Roch from the Baltimore Sun believes there is still a 50-50 chance a deal gets done.

Nationals Remain Interested In Benson: A source close to Kris Benson confirmed that the Nationals have remained interested in Kris Benson since his workout down in Arizona.

Wow now its the O's noy getting the deals done ... Back & Forth.

[RShack]

Wow now its the O's noy getting the deals done ... ...

Well, the article said that was according to "one source". For all we know, the source could be JTrea. He got quoted (or interviewed, or something) by somebody in the media once, didn't he?