[MtGrizzly]

Ok SABR dudes, I need some help understanding this whole "WAR" (Wins Above Replacement) thing. I'm specifically wondering about it in context of this Lookout Landing article on the M's 2009 outfield, as viewed through the WAR lense.

Preaching to the Choir

I'm especially struck by this statement:

Quote

We're looking at a league-average outfield with the potential to be more than that, and while the names might not be sexy and the bats might not be the most potent, it's about equal with an outfield of, say, Ibanez/Ichiro/Dunn, and well ahead of an "outfield" of Ibanez/Ichiro/Griffey. Hell, for all the flak Ichiro gets for not being a classic run producer, over the past three years he's been worth 0.9 wins more than Alfonso Soriano, 2.6 wins more than Vladimir Guerrero, and 4.4 wins more than Torii Hunter. He's an awesome player, even if he doesn't seem like one.

Hmmm, MtGrizzly thinks to hisself, that just don't seem right. So I goes to TangoTiger's site for an explanation of WAR.


How to calculate WAR

OK - honestly, not much help there. I'm less interested in how to actually calculate WAR than I am in its utility, its bias and its weakness. It seems that while a lot of the complicated stats like this do illuminate something that others don't, it also seems like they tend to leave a corner of the room or two in the dark. I intuitively like a stat that accounts for offense and defense, while making position adjustments to come up with an overall value. But this one just seems too...tidy.

So Matt, Sandy, Shields, Doc, anyone...help a fellow out? Is WAR as nifty a stat as it seems or is it, like the song, good for absolutely nothin'. Huh.

[Sandy - Raleigh]

I won't go into the whole thing -- but I will say that one of the major problems with the 478,017-part stats for wrapping everything into a nice tidy package is that with each level of complexity the ability of the common fan to have any hope of understanding it goes buh-bye. And, there is also the problem of if you make an error in ONE part of the process, it has the potential to undermine the whole -- even if the other 478,016 parts are fine.

My first beef with WAR.

The positional adjustments are:
+1.0 wins C
+0.5 SS/CF
+0.0 2B/3B
-0.5 LF/RF/PH
-1.0 1B
-1.5 DH


That is the foundation block they lay their entire defensive analysis upon. Guess what. It's COMPLETELY arbitrary.

First off - they give the catcher the #1 defensive bonus. This seems extremely odd to me for a number of reasons. First is that catchers actually play the fewest games at their position during a typical season. They don't put in the innings of your typical SS or CF.

Secondly -- they don't actually make that many plays, (beyond catching pitches). The assist totals for thwarting base-stealers is actually their primary direct defensive impact. That "might" be 100 plays per season. That's a third to a sixth of the total number of players at the other positions.

The *ONLY* reason to throw a +1 on the catcher spot is based on the long-standing BELIEF that catcher defense is critical. But there is simply ZERO mathematical data to support this notion.

I am happy to see the 1B position near the bottom of the list. In truth, while 1Bs make the most plays, they are the recipient of others work in the vast majority of cases.

The concept that there is a 1/2 game difference between 2Bs and SSs is another completely bogus idea. You go and look at plays made per position, and 2B and Short are "nearly" equal. But, go and take a look at the LEAGUE RFs. The Range Factor per position for the AL in 2008 were:

(catchers don't get an RF)
1B - 9.26
2B - 5.01
3B - 2.74
SS - 4.39
LF - 2.05
CF - 2.71
RF - 2.15
P -- 1.60

In truth, you have to subtract out the assists from the 2B, 3B, SS from the 1B stats to get a reasonable view of their ACTUAL contribution. What the RF measures, (in a less-than-perfect manner), is how many outs are recorded PER GAME (or 9 innings), by each position. Currently, pitchers fan about 7 batters per 9, leaving 20 outs to be divided by the rest of the team.

1.5 combined = 1B/CA/P/DH
2 each = LF/RF
2.75 each = CF/3B
4.5 each = 2B/SS

THAT is a more reasonable assessment of the outs per game from defense. It should stand to reason that the players responsible for making the most outs MUST be the ones providing the most defensive impact. Frankly, elevating the position that is responsible for the fewest outs, (1Bs typically get more assists than catchers, btw), to most valuable -- while shoving the position responsible for the MOST outs, (2B), to the middle of the pack shows a gross disregard for reality, IMO.

For *ME* -- it doesn't much matter what they do defensively after this point, because they randomly decided to hand a win per game to EVERY catcher before doing anything else. The concept that SS is *harder* than 2B does not mean it is more important. The throw to 1B is easier for the 2B, of course. But, that doesn't mean they don't produce the outs.

=========

My general view of the "mega" stats, that combine offense and defense to spit out a single number are ALL suspect. I think they *ALL* vastly overstate the value of defense, (not because defense isn't valuable - but because the current systems for judging defense are so volatile and suspect as to render them worse than useless).

I think most of the offensive win projections are probably fairly close to reliable. But, I think the combined systems, for all their intended objectivity, might as well just pull a random number out of a hat and add it to (or subtract it from) the offensive number.

Current day - I think it is completely reasonable to look at "offensive WAR" or "offensive win-shares" - and then look at the defensive side of the equation seperately.

[TangoTiger]

I came up against posting size limits here, so I have created a thread on my blog to address all the issues so far:

http://www.insidethebook.com/ee/index.php/...eptions_of_war/

[Lonnie]

TangoTiger, on Jan 26 2009, 12:46 PM, said:

I came up against posting size limits here, so I have created a thread on my blog to address all the issues so far:

http://www.insidethebook.com/ee/index.php/...eptions_of_war/

Interesting. Thanks TT! Oh, and welcome to MarinerCentral!

Lonnie

[MtGrizzly]

Matt, Sandy, Tango - awesome replies. Thanks. This is one of the things that I absolutely love about this site. I asks a question, I gets my answers.

all around.

Love the debate, even though some of it is whizzing above my head.

[Sandy - Raleigh]

Tiger,

First, let me apolgize for throwing out absolutes so trivially, ("completely", "zero"). That was unfair of me, especially without a complete understanding of your system.

Catcher defensive ratings are a hot button for me -- and I've seen other systems in the past, where the defensive valuing either was arbitrary - or the reasoning behind them was so bizarre as to be indefensible, IMO, (if you'll pardon the pun).

So - to be fair, I'll ask the questions directly, and allow you the chance to explain:

1) What is the rationale that suggests catcher DEFENSE is worth 2.5 wins per season (25 runs?), more than a DH?

[TangoTiger]

Sandy,

Thank you for your reply.

Let's ignore the DH for now, and focus just on the catchers.

I see nonpichers as belonging to pools:
Pool A - Catchers
Pool B - Infielders + Pool A
Pool C - Outfielders + Pool B
Pool D - 1B + Pool C
Pool E - Rest + Pool D

Because of the uniqueness of catchers, I reason that a team's set of catchers should have similar value to a team's set of noncatchers.

As it so happens, MLB happens to pay their set of catchers around the same as the noncatchers. While this doesn't prove anything, it at least shows us that this is how MLB thinks of catchers.

Also, in this year's Hardball Times Annual, I show what happens when you put noncatchers in a catcher position (for emergency purposes), or how much value Mike Piazza has in controlling the running game. It is a big difference, further showing the uniqueness of catcher. In that same article, I show that catchers take a huge hit on their hitting, likely because they are catchers. That once those guys stopped being catchers, their hitting improved.

So, in order not to take an extreme position on the matter, I decided to set a positional value to the catcher such that the catcher position is treated, offense+defense, as an average position.

Of all the positional adjustments I make, the catcher one is the one least grounded in data, and so, the one I have the least conviction on, and so, the one that's harder for me to defend. But, it is tempered with some basis in logic and baseball "feel".

[DaddyO]

Oh my! I LOVE this! Sabermetrics that is also plugged into MLB realities, and the ensuing discussion. Matt, Sandy, Tango, all engaging with logic, demonstrations and analysis. We are truly blessed to be in a creative, learning environment.

[MtGrizzly]

DaddyO, on Jan 26 2009, 01:18 PM, said:

Oh my! I LOVE this! Sabermetrics that is also plugged into MLB realities, and the ensuing discussion. Matt, Sandy, Tango, all engaging with logic, demonstrations and analysis. We are truly blessed to be in a creative, learning environment.

I know, right? I'm going to have to dip into my bank if this keeps up much longer.

This post has been edited by MtGrizzly: 26 January 2009 - 12:26 PM

[TangoTiger]

Matt is correct that there are sampling bias issues in terms of who gets to move from position to position.

(I would not bring Catcher in this particular discussion of moving players, since I did not use that for catchers. They're an exception as noted in my previous post.)

And, it's not just the players who move from position to position.
- I also look at the overall type of players you have there, via the Fans Scouting Report (e.g., these days, the Fans think that the guys who play LF are far more fielders than the guys who play RF... so, I've been mulling over the idea to separate the LF and RF).
- I make an adjustment that LH can't play 2B/SS/3B anywhere close to what an otherwise similar RH would.
- I use the offense at each position to try to gleam a bit more insight.

That said, there are other ways to do this: one was from Sean Smith at Hardball Times, and another by Dan Rosenheck. And, their models, completely different from mine, generally produced similar results to mine, within, more or less 0.25 wins.

***

And to highlight this post:

Quote

But this one just seems too...tidy.

If you were to drop the "too", then I'd consider it a high compliment!

There are alot of very complicated "all-in-ones" out there, and I concur with the other poster who said that if you make a mistake in one spot, the whole thing unravels. This is why I made mine as easy as it looks. Fangraphs was able to replicate most of it. And each component is there for everyone to see.

And this model is supported by MLB, as proven by their free agent signings over the last few years. Most signings (outside of Carlos Lee, Juan Pierre, Mike Cameron, and relievers) could be explained within 0.5 wins of what WAR said.

[TangoTiger]

SABR Matt, on Jan 26 2009, 12:20 PM, said:

Monetary compensation shows you something about scarcity...not about value. It's darned hard to find guys who can withstand catching. The demands of the position are a deterrant to people learning about it and being groomed as catchers.

Right, I am saying that scarcity is value. It's just like in Fantasy Baseball. The last guy picked gets $1. Everybody else gets something above this guy. The scarcity at catcher means that Joe Mauer is tremendously valuable, exactly because of what Matt says that it's darned hard to find guys who can withstand the rigors of catching.

Quote

I also have a major problem with equating uniqueness with real value. One could make the same kind of argument for base-stealers...they are very rare and hard to find these days...but you don't see sabermetricians arguing that base-stealing has twice the value it appears to have because it is unique.

I only meant uniqueness in terms of the overall contributions. Someone may have a unique ability to hit triples, but that doesn't give those triples more value. So, Matt's entire quote here represents nothing at all about what I am talking about.

I am only talking about the ability to generate or prevent runs, and how well you can do that compared to the average player you can put in that player's place and the worst player in MLB that you can put in that player's place. The "uniqueness" argument does not extend beyond that.

[Sandy - Raleigh]

Tiger,

Thanks for the response. One of the reasons that I elected to use the DH in my question is that the DH completely removes the question of defense from the table. By definition, the DH has no defensive component -- he literally cannot cost (or add) runs to his value via defense. Therefore, to me, it seems to be a completely reasonable question to ask what about playing catcher adds 2.5 wins in immediate value.

While I can see where you're coming from - my interpretation is that because you "wanted" catcher's to be worth as much as other positions, that you therefore were compelled to come up with a method (without much data to back it up), to reach your pre-determined conclusion. That's not the kind of methodology I personally find very alluring.

You note the reality that catchers take a beating in their offensive production - almost certainly a byproduct of the catching itself. This seems to lead to a cart-horse problem. Their offensive contribution is clearly less than that of other positions, therefore I will *make* their defensive contribution adjust for this.

============

That said -- I'll follow up with -- why the difference in Wins between 2B and SS?

Both are middle infield positions. The 2B position makes more actual plays. The SS position has just a few more errors, (due to the longer throw to first), but we're talking ranges of 6-18 for 2Bs and 7-22 for SSs in 2008, (starters) -- with the typical numbers being low double-digit errors for 2Bs and mid double-digit errors for SS. When the 'typical' difference for the two positions is 3-5 errors per season, I don't get where one could conclude that the position demands a 1/2 win difference between the two, (especially when SS is the one where the position is actually MORE prone to errors).

[TangoTiger]

Sandy - Raleigh, on Jan 26 2009, 12:39 PM, said:

While I can see where you're coming from - my interpretation is that because you "wanted" catcher's to be worth as much as other positions, that you therefore were compelled to come up with a method (without much data to back it up), to reach your pre-determined conclusion. That's not the kind of methodology I personally find very alluring.

You note the reality that catchers take a beating in their offensive production - almost certainly a byproduct of the catching itself. This seems to lead to a cart-horse problem. Their offensive contribution is clearly less than that of other positions, therefore I will *make* their defensive contribution adjust for this.

You are right that I am compelled to treat the catcher differently. And, if a guy as a pinch hitter takes a beating on his numbers, and if a pitcher as a starter takes a beating on his numbers, and a player as a catcher takes a beating on his numbers, and it is because you are tying those guys hands a bit, then I have to use that specific context to appropriately value them.

The numbers do show that Piazza is 10 runs worse than the league average catcher at stopping the running game, and he's the worst catcher of our generation at it. He was not a "natural" catcher. There's other work in the book I cited that showed that I do have some numbers to support me here.

Quote

why the difference in Wins between 2B and SS?

Matt answered it well: "In other words, defensive value is not just plays made but plays made * difficulty of plays. "

Look at 1B. He does make a play: he catches throws from the infielders. He ends up with being involved in alot more outs than the other fielders. But, the difficulty of catching a throw from the SS is nothing compared to the throw from the SS to the 1B.

Basically, the variance among players in making the play is what you are interested in.

And also, when guys move from SS to 2B, they perform better than when 2B move to SS (even though most 2B were SS at some point in their careers).

[TangoTiger]

SABR Matt, on Jan 26 2009, 01:04 PM, said:

Tom...I think you're completely off base to attempt to equate scarcity with value unless you're talking only about monetary value when deciding how much to pay per unit real value for a commodity. Value - in a baseball sense - is not the same as value in an economic sense. Real value to a team...the kind you can use to project future performance of said team...is expressed relative to the zero-value margin, not to the replacement player.

Matt, I disagree with your opinion as much as you disagree with mine (on this issue). I'm happy to leave it at that if you are.

[Taro]

TangoTiger, on Jan 26 2009, 12:54 PM, said:

You are right that I am compelled to treat the catcher differently. And, if a guy as a pinch hitter takes a beating on his numbers, and if a pitcher as a starter takes a beating on his numbers, and a player as a catcher takes a beating on his numbers, and it is because you are tying those guys hands a bit, then I have to use that specific context to appropriately value them.

The numbers do show that Piazza is 10 runs worse than the league average catcher at stopping the running game, and he's the worst catcher of our generation at it. He was not a "natural" catcher. There's other work in the book I cited that showed that I do have some numbers to support me here.




Matt answered it well: "In other words, defensive value is not just plays made but plays made * difficulty of plays. "

Look at 1B. He does make a play: he catches throws from the infielders. He ends up with being involved in alot more outs than the other fielders. But, the difficulty of catching a throw from the SS is nothing compared to the throw from the SS to the 1B.

Basically, the variance among players in making the play is what you are interested in.

And also, when guys move from SS to 2B, they perform better than when 2B move to SS (even though most 2B were SS at some point in their careers).

Hi Tango,

Is there a reason that there is only a 5 run positional adjustment difference between SS and 2B? Its seems that the difficultly between the two positions would be larger than 5 runs...
Using our team as an example Betancourt seems to be universely considered a -10, -15 run SS whereas Jose Lopez is a 0 to -5 run 2B. If Betancourt were to actually play 2B though I think he'd be a plus defender (just based on scouting him).

[Taro]

SABR Matt, on Jan 26 2009, 01:04 PM, said:

Tom...I think you're completely off base to attempt to equate scarcity with value unless you're talking only about monetary value when deciding how much to pay per unit real value for a commodity. Value - in a baseball sense - is not the same as value in an economic sense. Real value to a team...the kind you can use to project future performance of said team...is expressed relative to the zero-value margin, not to the replacement player.

I see what you're saying Matt, but teams just don't pay for performance between the zero-margin and replacement level (which is what Tango is trying to measure)... I think you guys are just aiming for different things here..

[Taro]

Quote

It follows with absolute logic that each position on the diamond is going to get paid roughly the same amount of money because they're all required to field a team without forfeiting games. It follows with absolute logic that there is an equal demand for talent at each position. So let's stay in the economic realm for a minute and register a complain of mine re: WAR even if used for economic purposes.

Fielders can move down the defensive pyramid, but they can't move up. I think it comes back to scarcity. Nearly any SS is 'capable' of playing any position either than C, but there isn't a 1B out there that can play in the middle of the diamond.

Quote

I think you have to make an unwarranted assumption to conclude that the replacement level player should be an average fielder at all positions, though. I don't think sabermetricians have spent enough time carefully calculating the average defensive value of your non-starter fielders at each defensive position, because I think what you're going to find is that if you sum up the 9 positions, the bench is about average, but at the skill positions they're better than average and at the slugging positions they're worse than average.

Here I agree...and it seems to be the one thing I really question about with WAR (the RL for defense). Defense at the extremes seems to change a player's value more than it should.

[Taro]

SABR Matt, on Jan 26 2009, 02:01 PM, said:

I'm not aiming for setting a budget...I'm aiming for winning baseball games (crazy...I know)...and predicting how changes in the roster will change the real results on the field (again...nutball crackers...what a concept!)

The problem with WAR...it can lead you to think that Endy Chavez will have the same impact on winning baseball games as Raul Ibanez. Clearly this is not true...while it may be that Chavez is the same distance above what a typical bench player would do...he's not the same distance above the zero-value margin and won't create as many wins.

I don't think the problem with WAR is in its structure, but I think that there 'may' be an issue in how replacement level for defense is set.

[Lonnie]

MtGrizzly, on Jan 26 2009, 01:23 PM, said:

Matt, Sandy, Tango - awesome replies. Thanks. This is one of the things that I absolutely love about this site. I asks a question, I gets my answers.

all around.

Love the debate, even though some of it is whizzing above my head.

Wow, and just think that it's free!





Lonnie

[Lakay]

Only thing I hate about thing is I have to spend :cpoints: from the bank already. BUT I will gladly spend them seeing the great quality of discussion on this thread. Kudos to all contributors especially to Matt, Sandy, and Tango!