THE BOOK--Playing The Percentages In Baseball

I generally liked the article by Dave, but I had two nitpicks with it.

One was that there is no reason we should expect that the distribution of the star players should be equal across positions or even roughly equal.  There might really be a concentration of superstars at first base.  It’s all fine and well to note that Zobrist ranks higher among second basemen than Fielder does among first basemen, but that doesn’t prove anything about their relative worth.  It’s fine for establishing plausibility but falls well short in establishing proof.

Second was the “side note” that you list above.  Dave’s addressing a straw man there.  Buster Olney specifically said that he was talking about ranking players based on 2011 stats only in regard to his GM comment, not that 2011 stats alone were the best measure for ranking players.

I don’t think the reader is going to appreciate the nuance of “if all you knew were his 2011 stats, and nothing else” implication.  When he says “Prince Fielder”, he is not only saying to take his 2011 stats, but he’s also asking you to look at his body of work in his career and, basically, his actually body.  If Dave is creating a straw man, then Buster is creating a biased incomplete scenario.

***

One thing that Dave didn’t do well enough to address Mike’s legitimate concern: the positional adjustments do NOT force the average 1B = average 2B.  Dave’s article certainly leaves the impression that that is the intent of the positional adjustment.

The average 1B (in 2011) has created something like 20 more runs than the average 2B (per 162 games).

However, the positional adjustment is actually 15 runs. 

Basically, the average GM will pay the average 1B more than the average 2B, because the average 1B is more valuable than the average 2B.

Tango/2, it wasn’t an implication/nuance.  Olney stated it flat out:

“If you asked 30 GMs who they would pay the most among these players–Victorino, Yunel, Kendrick, Fielder–off ’11 stats”

That’s a perfectly reasonable and straightforward thing for Olney to ask of WAR because that’s exactly what it’s designed to measure, is it not?

Now, perhaps you think readers would ignore what he said and ignore what WAR is intended to do because perhaps they don’t understand it.  Fine, but then say that you’re not addressing what Olney said, but rather addressing something else.

I also think it’s interesting to see how the five players mentioned by Olney rate in three different WAR systems.

FanGraphs:
5.6 Zobrist
5.3 Victorino
4.4 Kendrick
4.2 Escobar
4.2 Fielder

Baseball Reference:
4.5 Zobrist
4.3 Victorino
4.0 Fielder
3.8 Escobar
3.1 Kendrick

Baseball Prospectus:
4.3 Zobrist
4.0 Victorino
3.8 Escobar
3.7 Fielder
1.7 Kendrick

Pursuant to that comment:

If we remove UZR, assuming 9.4 RPW:

Are we really so confident in 2/3rd of a season’s worth of UZR that we can say that Buster’s point, that professional talent evaluators would consider Fielder’s 2011 performance (not his future projection) to be more valuable than the other three players listed?

No, that’s not what WAR was designed for.  WAR was designed to interpret past performance numbers, just as OBP, SLG, ERA and anything else was designed for.

In no way should you then look at WAR numbers in a vacuum, and then ask “which player would you sign” based on “11 stats”.

It’s as crappy way to do it as using 2011 ERA.

Buster’s question is the kind of thing that makes politics unappealing to discuss, because it’s intent is to stake a position, without really wanting to discuss the issue.

I should have added in Adrian Gonzalez, whom Olney was tweeting about earlier in the sequence as being superior in value to Zobrist (though not in the tweets that Dave referenced in his article).

FanGraphs:
5.6 Zobrist
5.3 Victorino
5.1 Gonzalez
4.4 Kendrick
4.2 Escobar
4.2 Fielder

Baseball Reference:
5.2 Gonzalez
4.5 Zobrist
4.3 Victorino
4.0 Fielder
3.8 Escobar
3.1 Kendrick

Baseball Prospectus:
4.5 Gonzalez
4.3 Zobrist
4.0 Victorino
3.8 Escobar
3.7 Fielder
1.7 Kendrick

Tango, you’re dead wrong here.  He didn’t say “sign”, as in a multi-year deal, and I don’t know why you changed the word he used and kept it in quotes as if Olney said it.  He was talking about value for 2011 and how much a GM would pay for 2011 stats.  If that isn’t what WAR measures, then you’re in a different sabermetric universe than I am.

E.g., Olney said “Shane Victorino spent a good deal of time on the DL—and he’s ahead of Adrian Gonzalez (who is excellent defensively).”

Why would that matter if he was talking about signing to multi-year deal?  It makes perfect sense if he is talking about WAR as a one-year measure of overall value to a team in that year, which is quite clear that he is from all his other tweets.

Similarly, he said the following:
“Do you think Shane Victorino has more value than Adrian Gonzalez? I like Victorino, but that makes no sense.”

“But ERA is not purported to be an overall measure of value in the way that WAR is.”

“But there is no front office—progressive or not—who would assign more value to Victorino over Gonzalez, or pay him thusly.”

“Yes; Zobrist is an excellent player. Everybody loves him. But greater value than Adrian Gonzalez?”

When Olney is using the language of WAR exactly as it is designed, why would you want to assume based on things he didn’t say that he has a misunderstanding and then criticize him for that imagined misunderstanding?

"Imagined” misunderstanding? You are assigning an intent to me that is not there. 

I think it’s a fair enough interpretation that if someone “pays” for stats, that he is “signing” that player.  I’m not imagining anything.

But, let’s say I misinterpreted for the sake of this discussion.

Now, if we want to be as fair as possible to Buster, and speak strictly in terms of “produced value”, then let’s just deal with it on that basis.  So, we’re going to interpret Buster as saying “paying for produced value after-the-fact”.

What are we supposed to think when he says “Zobrist is an excellent player”?  That’s not a “performance in 2011” statement.  That’s a true talent statement.  So, he’s now intermingling true talent statements with past observed statements.

So, when I see stuff like this:
“who would assign more value to Victorino over Gonzalez, or pay him thusly”

The correct statement to make is:
“who would assign more value to Victorino’s 2011 performance irrespective of his prior performance over Gonzalez’s 2011 performance irrespective of his prior performance, or pay for that performance thusly”

And his point of contention, one issue that he’s talking about, is the positional adjustment.  So, he’s got a problem with the positional adjustment.  That he disputes WAR on that basis.  And, what can I say that I haven’t already said?  I’ve got good positional adjustments.  I have the gap between 2B and 1B as 1.5 wins.

Tango, Fangraphs includes WAR translated into dollar values right there on the player cards. I don’t see why discussing WAR in terms of monetary value should imply a contract, rather than discussing the values as they are presented. Given the limitations of Twitter as a communications medium (140 characters is NOT a lot, trust me) I’d err on giving Olney the benefit of the doubt.

Or put it this way. Fangraphs says that, given season to date stats, Howie Kendrick has been worth $19.60 million and Prince Fielder has been worth $18.90 million. My reading of Olney is that he’s saying that no GM would put a higher value on Kendrick’s performance (not projection) than Fielder’s, contrary to the dollar values listed on Fangraphs.

Colin10/11: I already replied in Tango/9.

Buster mixed up his statements, saying things like:
“Do you think Shane Victorino has more value than Adrian Gonzalez?”

That is NOT about performance in 2011.  “Has more value” means “has more value”.  That’s a statement of true talent.

If he meant it in terms of performance in 2011 that we can attribute to Victorino and Gonzalez, then that’s how we should talk about it.

***

Let’s just agree that Twitter doesn’t lend itself to accurate representation of thought because at this point, we’re just parsing the heck out of tweets.

If we want to discuss “performance in 2011 we can attribute to player X”, then let’s talk about that.  If we don’t want to talk about this, then, well, we’re going to get nowhere fast.

If we want to discuss “performance in 2011 we can attribute to player X”, then let’s talk about that.

Yes, let’s!

Cameron brings up things like ERA and batting average not passing the smell test. What he’s really doing is uncovering very common sabermetric criticisms of those metrics - ERA likes Vogelsong because it doesn’t know how to account for a pitcher’s defensive support, for instance. This is why Fangraphs uses FIP instead of ERA for fWAR, for instance. So I don’t see how this is a defense of using ERA to measure performance (rather than talent) in an arbitrary period. And if you can’t defend ERA as an adequate measure of Vogelsong’s performance, it seems to me like Dave’s analogy breaks down.

A GM has $X million (say 20) to buy a player’s performance to date in 2011 that he could magically tack on his own team’s performance.  Who’s production does he buy?  Or asked another way, a GM has a replacement level player at every spot on the field.  He can retroactively choose a guy from among those listed to improve his team.  Whom would he choose?

Realistic it’s not, but that seems to be the question Buster is asking.  In in that context, most GMs would spend their money/pick on Fielder.

Realistic it’s not, but that seems to be the question Buster is asking.

It’s not just some random silly question.  It’s EXACTLY the question WAR purports to answer.  So if one were to question WAR, as Olney was doing, it’s a perfect question to ask.  He was meeting WAR on its own terms.

Not to be too much of a nit, but starting with an all-replacement team and tacking on one player is not what WAR is about.

In any case, I’ve already said that we can fairly interpret Buster as wanting to say this:

The correct statement to make is:
“who would assign more value to Victorino’s 2011 performance irrespective of his prior performance over Gonzalez’s 2011 performance irrespective of his prior performance, or pay for that performance thusly”

Each side has a fair interpretation of what Buster actually said.

But that doesn’t make sense to me.  I’d rather take Zobrist at an up-the-middle more premium position and find a serviceable replacement at first base.  It’s much harder to replace a Zobrist at second base.  Plus, projecting Fielder to age well w/his body type is another consideration altogether.

Ok, so onto Victorino v Fielder (all data from Fangraphs).

Hitting: Fielder has produced 15 more runs than Victorino.  Is this what Buster is disputing?  I don’t know.

Baserunning: Victorino has produced 8 more runs than Fielder.  Is this what Buster is disputing?  I don’t know.

Fielding+Position: Victorino has produced 21 more runs than Fielder.  Is this what Buster is disputing?  I don’t know.

Playing Time: Fielder has produced 4 more runs than Victorino. Is this what Buster is disputing?  I don’t know.

Add it up, and Victorino has produced 10 more runs.  This is presumably what Buster is disputing.

I’ll be glad to have the discussion with Buster, if he’d like.  Or he can tweet his conclusions.

I should not say “had produced”, but rather “estimate to have produced, with a certain amount of uncertainty level”.

Olney says he was talking about WAR, but you want to act as if he doesn’t know what he was talking about and actually meant to say “true talent” and then proceed to show how his understanding of true talent is wrong.  Fine.  It’s not an accurate way to engage with Olney, but whatever.  Good thing sabermetricians aren’t ideologues.

Based on Dave’s selection of quotes it seems the basic takeaway is that GMs place a greater emphasis on elite offensive performance than playing very good defense up the middle.

So from your list of components Tango, I think Buster believes that GMs would take issue with the positional adjustment most of all.  I would add that I imagine they fail to sufficiently account for base-running in their mental calculus

"ERA likes Vogelsong because it doesn’t know how to account for a pitcher’s defensive support, for instance.”

That’s not it.  His defensive support is only a tick above average, and his BABIP is a good, but not extreme .280.

He’s given up a lot fewer runs than you should expect from his hits, walks, and homers allowed.  The big thing that stands out is the timing - people are only hitting .160 off him with RISP.

Based on Dave’s selection of quotes it seems the basic takeaway is that GMs place a greater emphasis on elite offensive performance than playing very good defense up the middle.

Almost everyone does.

Saying that Ben Zobrist is more valuable in 2011 than Prince Fielder puts you in small company ... even if it’s true.

Prince Fielder, the guy who’s about to land a mega free agent deal is not as good as what did you say his name was again? Zenith? Zorbet? Zombie?

It’s not hard to understand how anyone could have an issue with the reality, especially when considering what GMs typically value. The problem with that is that there may be up to 7 or 8 GMs that would pick Victorino’s stats. This isn’t 1990. But, these are the same GMs that if they could pick one guy off the NYY, might pick Brett Gardner and not Cano or Teixeira, etc.

But Olney does ask a very good question when he is asking about the positional adjustments. That aspect was not addressed at all.

#18, it is extremely obvious to me from reading the quotes what he’s disputing is the position adjustment.

If Buster wants a conversation, I’d be interested in what he thinks it should be.  But it looks like the only way you’d get Fielder to rate comfortably ahead of all those players in the comparison would be to use no position adjustment at all.

Rally/18: excellent point.  Since Fielder is behind by 10 runs overall, but Victorino has a 15 run positional adjustment, then simply removing the positional adjustment will put Fielder just 5 runs ahead, which would still seem to be too close to Buster’s comfort.

Basically, how many runs or wins does Buster have Fielder above Victorino?

Since Fielder is behind by 10 runs overall, but Victorino has a 15 run positional adjustment, then simply removing the positional adjustment will put Fielder just 5 runs ahead, which would still seem to be too close to Buster’s comfort.

I come up with a 10 run difference based on positional adjustment, and an 11 run difference in UZR, from the sortables at Fangraphs. Where does 15 runs come from?

My bad… 15 runs is the per 162 game adjustment.  Obviously, with two-thirds of the season, the 10 run adjustment is what we are really talking about.

So, yeah, removing the positional adjustment altogether, and they are even. Thanks, good catch.

Presumably, Buster is thinking that Fielder’s performance in 2011 must be far enough ahead of Victorino’s that the two can’t be compared.  That would imply at the minimum a 10 run difference in support of Fielder (if not 15 or 20 runs of him being ahead; but, let’s be nice to Buster and presume Fielder should be at least 10 runs ahead).

Really, Buster’s issue might be with the replacement level being so high, that Fielder having, what, 100 more PA and/or 25 more games, isn’t given enough credit.

Sure, it’s possible that if you are evaluating performance in any given season that you want to give more credit for simply being on the field.

The way replacement level is setup, you get 22.5 runs per 162 games.  So, if you have 25 more games than someone else, you get 3.5 more runs, which seems like a ludicrously small number.

But still.  You can make the replacement level 40 runs per 162 games, and that will still only get you at 6 more runs ahead.

Something to think about:

Most benches contain part-time or utility players who are solid defenders and/or baserunners, which means that situationally, you can often plug in better-than-replacement level performance from a replacement-level player, e.g. a pinch-runner, a defensive sub.

Most benches do not contain anyone with power like Prince Fielder.

So, perhaps instead of comparing Fielder to Zobrist, if we’re trying to think like GM’s, we could compare:

Prince Fielder + an average middle infielder + a decent pinch-hitting bat + a good fielding/good running utilityman

vs.

Ben Zobrist + an average first baseman + a decent pinch-hitting bat + a good fielding/good running utilityman

There’s enough moving parts in here to support any argument, but what I am getting at boils down to this:

If you have Fielder but not Zobrist, you can situationally substitute in the best aspects of Zobrist’s game (fielding, running), as well as situationally substitute out the weakest aspects of Fielder’s game (fielding, running).  You have lots of options for making your on-field lineup better when the situation (leverage) favors it.

If you have Zobrist, but not Fielder, you cannot pinch-hit a 50-HR bat for your average-hitting 1B (nor would you PH for Zobrist himself, he’s too good for that), and you (generally)cannot sub in a better fielder/runner than Zobrist.  You have limited ways of making your on-field lineup better.

That is why WAR-comparisons like this are missing something: they do not / cannot model situational substitutions that strongly impact the game…

Victorino has 3 games he started that he didn’t complete, while Fielder has 9.

I don’t know how much leverage you can get out of 6 games from your bench that will minimize the negatives of these starters in favor of the bench.  All that might be worth 1 run, maybe 2?

I appreciate the idea behind it, but if you sit down and work it out, it just won’t have the impact.

Benches these days often have 4 players.  One of them is a backup catcher.  You can only substitute once per player per game.  That really limits the impact of replacing a Ben Zobrist type with bench players.  In the ideal game you can start John McDonald at 2B, play him for 9 innings, then pinch hit Jason Giambi when his spot comes up with bases loaded in the bottom of the 9th. 

But few games follow the script.  You’ll have to bat Mac in the 3rd inning because you’re not sure if this is the best spot for Giambi - once you use him he’s got to leave the game for another competant fielder.

Victorino has 67 R, 40 RBI (which with 11 HR means 96 runs participated in).

Fielder has 71 R, 85 RBI (and 130 runs participated in).

It’s a good guess that Olney thinks that there must be at least a 30 run difference on offense between the two, and that that’s his starting point.  And I presume 90% of baseball fans would think this way.

I’m dying to get to transparent defensive metrics. 

What I mean by transparent is, if someone says player X is 25 runs better defensively than player Y, it would be great if you could easily look at a list of plays and identify the ones that each player was involved in where they saved or coughed up runs.

Unfortunately, we can’t do that right now.  No doubt a lot of skepticism about metrics like UZR, and thus WAR, derives from this lack of transparency.  Or maybe I should refer to it as a lack of traceability…

In Chicago we have a typical sports radio station called The Score. I was listening a while ago to the Boers and Bernstein show. Dan Bernstein brought this up and actually read the Buster tweets and the Cameron post in its entirety. It’s cool that the younger sportswriter like Bernstein are growing up with saber, and actively trying to get it in there whenever they can.

At least Olney didn’t call Fielder a “50-HR bat”. Wow.

------------------------

@34

Bernstein will do that from time to time in order to get a good discussion going. But, Terry Boers will end it all with a “Ah, I don’t like him.”

A conversation of one, isn’t much fun.

IMO, most people enjoy being like Boers. They enjoy their opinion based on their vast experience and knowledge gleened from their days of playing Little League. After all, Major League Baseball is just grown men playing the same game we all played as kids (my ass).

When it comes to arguing things in baseball, I don’t think people want an objective answer. I don’t think they want to be able to look it up or calculate it out to end in an objective answer. I think they enjoy being their own expert based on their observations, experiences, preferences, and memories.

By providing data, stats, and analysis like this, one is - in fact - telling a person “what you know about baseball is wrong”, and they don;t want to hear it.

I’m the exact opposite, but I will demand that you prove it. BP’s book “Baseball By the Numbers” will never be read by many fans for its subtitle “Why everything you know about baseball is wrong?”

You think folks really want to acknowledge the real value of sac bunts, stolen bases, relievers, and other aspects that “everyone knows” are awesomely awesome? Heck, even the people whose livelihood depends on such things don;t really want to know. It’s stunning.

@34/Circle- It makes more sense when viewed through the prism “who gets to make money analyzing baseball?”.  Then it’s less surprising to hear Dan Shaughnessy or Murray Chass talk about mothers’ basements.  Or that the internets are full of articles about how pea-brained the BBWAA is.

Olney’s an interesting case.  He’s a reporter, really, and shouldn’t really have an interest one way or the other.  It seems like he’s just starting to think about these issues (odd, I admit), and I’m curious where he’ll land.

With the average fan, I think inertia is a big issue.  Some folks just can’t be bothered to unlearn/re-learn stuff that’s already settled in their minds.

CC #35:

At least calling Prince Fielder a 50-HR bat isn’t as bad as calling Ted Williams a .400 hitter; that would have been a double-WOW! (with the extra 2-minutes for uttering a phrase that acknowledges the existence of “batting average” without the obligatory derision...)

I liked Dave’s article, a good explanation of why position is important.

If I had $20m or so to spend on a player, on average Zobrist is going to represent a much larger upgrade over the player I currently have at 2b than Fielder is going to be an upgrade at 1b.

Buster might like THT’s rankings better, here’s my list of the mentioned players

Hardball Times
4.5 Gonzalez
3.6 Escobar
3.5 Fielder
3.4 Victorino
3.0 Zobrist
1.7 Kendrick

Oliver has Zobrist as the 2nd best 2b year to date in 2011, behind only Pedroia. Oliver doesn’t like his defense as much - after assigning Zobrist a huge +20 runs in 2009, it’s given 0 and +1 the past two years.

Oliver doesn’t like his defense as much - after assigning Zobrist a huge +20 runs in 2009, it’s given 0 and +1 the past two years.

Where I live that would be called an “Oliver problem”, not a “Zobrist problem”.

@37

Lighten up, Scrabble.

*Big Grin*

Fair point.

While I agree with Dave re: the positional adjustment, I’d point out that Olney and others do have a point with regards to the use of fielding data from a partial season in WAR. 

Such fielding data - from ANY fielding metric, not just UZR, available - is not just too small of a sample size to be predictive, but due to the nature of how the data is collected, just not enough that we can even be sure that the fielding rankings are correct.  So how can we plug that in to WAR?

That’s always bothered me: MGL and others claim Fielding metrics are good after three years, but then say nothing about their inclusion in one year WAR measures. 

So to Olney I’d say that the positional adjustment is correct for reasons not having to do with positional scarcity as much as the difficulty in playing positions (though the two often amount to the same thing), but that the fielding #s may be misleading.  Even so, according to Fangraphs, Zobrist is basically on par with Fielder once you discount the fielding data - because of the position(s) he plays.

Comment I was going to make to CC/39 also applies to 41.

I was careful to say that Oliver assigned Zobrist x number of fielding runs. None of us no exactly, but I have developed my own model of what to look at and how to count it up, and sometimes I get different results than others.

Very true, a single season of fielding can be unreliable, but hat’s not really unique, as I can say the same thing about batting average. Many guys can be consistent, while others jump around from season to season.

As was discussed in the direct standardization thread, you shouldn’t take a rate and extraplote it over a larger sample size without first regressing, but that’s what a single season UZR150 does. I think it’s fine to say that so far this year a player has been evaluated in a certain way, as long as we understand that it’s a season by season analysis. Those can then be combined to form a true talent estimate.

At SS, I’ve got Andrelton Simmons best at +14.7. That’s based on +21 in 61 games in 2010, +13 in 102 games in 2011, then regressed. Zach Cozart second at +14.2 (+13,+25,+7), Brendan Ryan +13.3 (+6,+9,+14,+12) then Tulo +12.8 (+14,+14,+12,+12)

"MGL and others claim Fielding metrics are good after three years, but then say nothing about their inclusion in one year WAR measures.  “

I don’t think that’s true.  In any case, from MGL’s perspective, he doesn’t care at all about valuing past performance.  He cares about true talent levels, and so, he’d regress everything anyway.

MGL addressed the one-year fielding issue very well here:

http://www.insidethebook.com/ee/index.php/site/comments/does_one_year_uzr_tell_us_exactly_what_happened/

----"Very true, a single season of fielding can be unreliable, but hat’s not really unique, as I can say the same thing about batting average. Many guys can be consistent, while others jump around from season to season.”
******************

That’s not a good comparison to use, because the variability in BAvg is pretty much due to a different reason than the variability in UZR.

With BA, we are not looking thru a cloud and estimating, as we already know the exact numbers of hits and at bats - but my point is that those underlying “true” counting stats themselves have a lot of year to year variance, so the effect is not exclusive to UZR or any other defensive metrics.

Defense has that plus the uncertainty of estimating unknown counting stats, or expected values.

----"Defense has that plus the uncertainty of estimating unknown counting stats, or expected values.”
****************

How do we know how much of the former that fielding has? What part of fielding is analogous to the batter’s inability to aim the ball when he hits it, while still swinging the bat hard? That’s why people commonly think that fielding is more stable, because their eyes don’t see that analogous component to fielding. Maybe it’s mostly the ‘estimating values’ that makes fielding look so variable.

@46

Very good point Brian.

If I am not mistaken, part of the initial “sabermetric” quest was in wonder about how much of each hit should be credited to the batter, since not all hits are created equal.

Currently the hitter gets full credit for a hit, even if its just a blooper that falls in front of TCQ in RF, or bounces in front of the wall in CF just out of the reach of Colby Rasmus who gave up on the play thinking it was a home run.

KJOK: great stuff, thanks for find it.  That was a good thread.

"What I mean by transparent is, if someone says player X is 25 runs better defensively than player Y, it would be great if you could easily look at a list of plays and identify the ones that each player was involved in where they saved or coughed up runs.”

For most people that would not be pretty to look at.  It would be a list from several hundred to a thousand lines long for each player, and would show for each what the estimated chance of making the play was, given the buckets it fell into, whether or not the player made a play, and the run difference between the two.  Add up all the +.15’s and -.08’s, and you get the fielding rating.

It would be great for public understanding if you could say “Brick Stonehands, rating = -30 plays, and then show the 30 plays he failed to make that an average player would, but that’s not how it works.

As for producing the type of list in my 2nd paragraph, before I started collecting paychecks for my work I would not have objected to showing the detail of TotalZone on a play by play level on any proprietary information grounds.  I may have objected on workload or bandwidth grounds.

At a team level, for every 4000 balls in play there are about 2800 outs.  Each one of those outs will show a player saving runs for his team.  Some a little, some a lot, and varying greatly based on the inputs of the defensive system.  Then there are about 1200 hit allowed - for each one some player (or multiple players) will be charged some portion of runs allowed.

It’s not pretty.  But databases are good at adding that all up into a number the human brain can make sense of.

Rally,

Most people don’t look at a list of every one of a player’s 600+ plate appearances in a season, either, but knowing that it is there, and that there is a traceable source for the run values being quoted, is comforting to anyone who uses the numbers, whether they delve deeper or not…

You’re right, of course, that defensive plays are not quantized the same way offensive plays are, but we will still be able to go down the list and identify the most significant plays both positive and negative, and watch the video to confirm them.

The difference between an offensive play and a defensive play was described in post 2 here:

http://www.insidethebook.com/ee/index.php/site/comments/does_one_year_uzr_tell_us_exactly_what_happened/#2

Recapping, for an offensive play, the “baseline” wOBA for any play is .250 to .400, if you consider all the conditions (pitcher, park, etc).  So, if someone gets a HR, he’ll get +1.6 to +1.75 extra wOBA points.  It’s all pretty close.

But for a fielding play, the baseline for any play will be .000 to .999.  So, when a guy makes an out, he’ll be +.001 to +.999.  It’s a huge gap.  Estimating his chance at making an out requires far more than just knowing the pitcher and park (like for a batter).  You have to know where the fielder was, where the ball was hit, how much time he had to get there, in addition to knowing runners on base, and so on.  To estimate that would require details that we just don’t easily have, so we make due with what we have.

Whereas for a hitter, your uncertainty is whether that HR was +1.70 or +1.65, for a fielder, your uncertainty (based on available data) is whether it was +.050 or +.350.  It’s just so much wider.

That’s why HITf/x and FIELDf/x will be a godsend here.

Recapping, for an offensive play, the “baseline” wOBA for any play is .250 to .400, if you consider all the conditions (pitcher, park, etc).  So, if someone gets a HR, he’ll get +1.6 to +1.75 extra wOBA points.  It’s all pretty close.

Tango - I think you are creating a larger difference between offensive and defensive metrics than there actually is by arbitrarily choosing your baselines and units.  If you are considering park for an individual play than a fly ball hit with the same speed, vertical and horizontal angles, and spin could be a HR, extra base hit, single or out depending on the park.  That’s about as huge a swing as you can get, from 1.4 runs to -.25 runs in linear weights.  That’s the same range that an outfielder could have on a defensive play where he catches a ball that would have gone over the wall.

For balls in play if one uses runs for both offensive and defensive metrics the range is actually smaller for defensive metrics for any individual play.  Convention for offensive metrics gives the total value of a hit or out to the batter.  Defensive metrics only give a portion of the hit balls value to the defensive player making the range of possible run values smaller.  Both offensive and defensive metrics know that they are not assessing the run values exactly right for any indivdual play and are depending on aggregation over a period of time to “even out” a player’s offensive or defensive total value.  How well this works for each type of metric is still a very debatable question.

Hit Fx and Field Fx will be a big help in make the assigning of run value more accurate on any individual play, but whether they will offer significant improvement on a year’s worth of aggregated data remains to be seen.

There’s nothing “arbitrary” to what I did.

One point in wOBA is equal to one point in OBP which is roughly equivalent to one point in outs per play on the fielding side.

To convert wOBA to runs, you divide by 1.20 (or multiply by 0.83).  To convert outs per play, you multiply by 0.75.  The scales are close enough to make the illustration useful.

Secondly, in both cases we are looking at outcomes, results.  A HR is a HR, and not “60% of a HR, and 30% of a double, and 10% of an out”.  And the same goes on the fielding side: an out is “100% out”, and not anything else.

Therefore, I reject your argument against my illustration.

My illustration stands as an accurate description of the issue between hitting and fielding stats: the opportunity for success for each plate appearance is roughly the same for every hitter, but the opportunity for success for each batted ball is nowhere near the same for every fielder.

The entire issue with fielding rests on this very issue.

This is why HITf/x will be a godsend: finally, we will know the “opportunity for success” for each batted ball.

Hopefully this doesn’t muddy up Tom’s illustration, but while I think as Peter’s arguing that defense has some of the same assignment of responsibility issues regarding run value as offense such as park, etc., at least on offense if the batter makes an out it’s only the batter you have to ‘adjust’ and credit/debit, not assign partial credit to the on-deck batter, or the batter who just batted, etc. because you’re not really sure who on the offensive team is responsible for the out - which is what you have to do for defensive run value calculations.

My illustration stands as an accurate description of the issue between hitting and fielding stats: the opportunity for success for each plate appearance is roughly the same for every hitter, but the opportunity for success for each batted ball is nowhere near the same for every fielder.

The entire issue with fielding rests on this very issue

Tom - I respectfully assert that you are dead wrong.  The difference in offensive and defensive metrics lies in several factors.  Much of a batter’s offensive value lies in HRs, strikeouts and walks; outcomes of a PA which we are pretty sure are mostly due to a batter’s skill and thus repeatable from year to year.  We are also pretty sure that a batter has some control over the distribution over the quality of the balls in play that he hits.  Some batters are able to hit the ball harder consistently than others, or with less hang time, and perhaps between fielders.  This too is a skill that seems to be somewhat repeatable from year to year.

In contrast to this a fielder has absolutely no control over the distribution of balls hit into his area of responsibility.  This distribution can vary greatly from year to year due to chance, due to the pitchers on his team, and somewhat due to the batters that hit the balls to him.  His entire skill rating for the year be determined by how he performs on the 60 to 80 hit balls that he gets a year that are hit in such a way that they either are not caught more than 95% of the time by the worst fielder at his position or not hits 95% of the time for the best fielder.  His skill as a fielder may not vary much at all from year to year, but his skill rating will appear to vary because the variance on those 60 to 80 balls “that matter” is going to be much greater than the variance on the 500 to 600 PAs where a batter can exercise some control over the results.

Therefore I reject your rejection.

I don’t think you can appreciate my argument, since alot of what you say supports my argument.

I’m talking about the “opportunity space”.  A batter gets 600 PA, and his chance at success for 600 PA this year and 600 PA next year and 600 PA two years ago is roughly the same.  He’ll face roughly the same quality of pitchers in front of the same quality of fielders in front a random set of parks.  More or less, give or take. 

If you put in an average hitter into Prince Fielder’s opportunity space over the last 3 years, that average hitter is going to have about the same wOBA each year.

But for fielding?  No, not at all, and the reason was articulated by you yourself.

You give Chase Utley 600 balls in play in 2011 or 2010 or 2009, and those 600 balls in play are not going to look anything alike each year.  His pitchers play a role, his fielders play a role, the batters play a role, the parks play a role, where he positions himself plays a role.  All these things conspire against making the 600 BIP in 2011 anything close to equivalent to the 600 BIP in 2010.  The opportunity space is not the same.

You put an average fielding 2B in Utley’s spot in 2011, and he might get 400 outs.  You put him in Utley’s spot in 2010, and he might get 350 outs.  You put him in Utley’s spot in 2009, and he might get 450 outs.  We have limited knowledge as to what kind of opportunities are presented to each fielder.

The entire thing of hitting v fielding lies in the opportunity space of the hitter and fielder, and our ability to determine what that opportunity space is.

Therefore, I dismiss your rejection of mine.

This is fascinating stuff and I want you both to continue, but I’m getting confused about what exactly each of you is standing behind - perhaps could you both clarify your positions?

My point is thusly:

1. A batter is going to see some 2500 pitches per season.  Those pitches are going to come to him from focused angles, focused speeds, focused movements, and focused locations.  The fastballs are going to be 88-98 mph, he’s going to see 60% - 70% fastballs, the strike zone is 4 square feet, and so on.  It’s all a very controlled environment.

If we had to estimate what an average hitter would do with the environment faced by Prince Fielder in 2009 and with the environmenet faced by Prince Fielder in 2010, we’re going to come away with the fact that Prince Fielder’s environment to hit has not changed.

His opportunity space is going to be similar year to year.

2. A fielder is going to see 4000 balls in play, spread out 90 degrees (or more), spread out from 40 feet to 400 feet, going up at launch angles from minus 20 degrees to plus 40 degrees.  He’s stationed in one part of the field, and so, will only be involved 10% of the time.

The environment that Chase Utley faced in 2009 has virtually little to do with the environment he faced in 2010.  What an average player would do in each environment will bear little resemblance. 

The opportunity space is not the same whatsoever.

3. Other than in the parts where Peter disagrees with me, I agree with him entirely.

I wonder how true this is.  I.e. I bet Utley 2010 mix of trajectory, speed, direction is pretty similar to the Utley 2011 mix.  When you say an average fielder in Utley’s spot might field 350-450 balls/yr depending on the year, is that data-based or made up for sake of example? 

Semi-related:  wouldn’t UZR be improved/enhanced if it were expressed on a per opportunity basis?  Seems like this would eliminate some bias created by pitching staff tendencies (Ks, GB/FB) and just plain luck affecting how many balls are hit into the zome of a particular fielder.

I don’t know about Utley specifically (just an illustration), but there’s a huge difference in quality of opportunity year to year.  If there wasn’t, then we’d be happy with assists and putouts per ball in play.

As for UZR, MGL has shown it in the past as a per opp (called the UZR rate).  Now he shows it as per 150 “games”, where a “game” is the typical number of outs a fielder would get at that position.

Thanks.  I thought UZR/150 was just UZR extrapolated to a full season.  Did not realize it was normalized for the # of chances.  Would there be value in normalizing for the quality of opportunities as well?  I.e. Peter Bourjos comes up and sees an inordinate number of 25/75 chances and his UZR goes wild.  You might normalize to a typical mixture of cans of corn v great running catches.  Or does UZR/150 do this, too?

Right, UZR/150 does this.  Imagine that ALL of Bourjos’ chances were of the 99% variety.  All of a sudden he has the equivalent of “200” games in a season.

I don’t know about Utley specifically (just an illustration), but there’s a huge difference in quality of opportunity year to year.  If there wasn’t, then we’d be happy with assists and putouts per ball in play.

But no one has ever shown that UZR and the like are an improvement over assists and putouts-based metrics.  (If I am wrong, point me to the research.)