At what point do we take a good long look at WAR and realize it needs fixin'

[DaveW]

Starling Marte: 1.7 WAR

Bryce Harper: 1.7 WAR

Gerrardo Parra: 1.7 WAR

Drew Butera: 0.0 WAR

 

:banghead:

[Riverbrian]

Let's see... It's April 29th, 2013... My Watch says 3:28PM Central Time.

 

I'd say April 29th, 2013... 3:29PM Central time.

[gunnarthor]

Kevin Correia 1.6 WAR.

[cmathewson]

Three years ago. My big thing is how it is a different stat depending on whether you subscribe to the FanGraphs or BBRef definitions. At least let's have a standardized way of calculating it.

 

Beyond that, my sense is it is overly weighted towards defense. The definitive case of this, for me, is Nick Punto vs. Delmon Young 2010.

 

Delmon (FG WAR 0f 1.6), 153 games, OPS+ of 124 and 281 TBs.

Punto (FG WAR of 1.3), 88 games, OPS+ of 71 with 76 TBs.

[Brock Beauchamp]

Kevin Correia 1.6 WAR.

 

Miguel Cabrera 1.1 WAR.

 

Kevin Correia >>>>>>>>> Miguel Cabrera.

 

Suck on that, Tigers fans.

[DaveW]

Kevin Correia 1.6 WAR.

That is fine, since pitchers seem to be helped by WAR what not anyways, and discussed in another thread, WAR is rather pointless in general for pitchers since you already know if they are getting effective results or not.

[DaveW]

Miguel Cabrera 1.1 WAR.

 

Kevin Correia >>>>>>>>> Miguel Cabrera.

 

Suck on that, Tigers fans.

 

Again, go actually look at Harper's stats and compare them with Marte (a LF) and Parra.

 

Tell me how they are equal in WAR, has Harper suddenly become Manny Ramierez in the field???

[DaveW]

Three years ago. My big thing is how it is a different stat depending on whether you subscribe to the FanGraphs or BBRef definitions. At least let's have a standardized way of calculating it.

 

Beyond that, my sense is it is overly weighted towards defense. The definitive case of this, for me, is Nick Punto vs. Delmon Young 2010.

 

Delmon (FG WAR 0f 1.6), 153 games, OPS+ of 124 and 281 TBs.

Punto (FG WAR of 1.3), 88 games, OPS+ of 71 with 76 TBs.

 

Yeah, exactly I think it takes defense (whose advanced stats are already flawed) into factor wayyyy to much.

 

Also Nick Punto 2013 OPS: .933

[jm3319]

Three years ago. My big thing is how it is a different stat depending on whether you subscribe to the FanGraphs or BBRef definitions. At least let's have a standardized way of calculating it.

 

Beyond that, my sense is it is overly weighted towards defense. The definitive case of this, for me, is Nick Punto vs. Delmon Young 2010.

 

Delmon (FG WAR 0f 1.6), 153 games, OPS+ of 124 and 281 TBs.

Punto (FG WAR of 1.3), 88 games, OPS+ of 71 with 76 TBs.

 

Have you seen Delmon play defense? Or should I say, have you seen him TRY to play defense? Scoring runs on offense gets negated if you flop around like a fish out of water in the outfield, hence Delmon not being that much higher than Punto in terms of WAR.

[DaveW]

Have you seen Delmon play defense? Or should I say, have you seen him TRY to play defense? Scoring runs on offense gets negated if you flop around like a fish out of water in the outfield, hence Delmon not being that much higher than Punto in terms of WAR.

Yeah Delmon was really awful, but this is getting off topic. How is Harper the same as Parra and Marte? Harper is considered very "great" in the field...no?

 

FYI: This topic was started by Heyman, so it's not like I am just pulling crap out of my proverbial ass.

[Brock Beauchamp]

Again, go actually look at Harper's stats and compare them with Marte (a LF) and Parra.

 

Tell me how they are equal in WAR, has Harper suddenly become Manny Ramierez in the field???

 

WAR relies on defensive metrics. Therefore, by association it's mostly useless for anything less than a full season.

 

Hell, I don't even like using WAR to compare position players to pitchers. The metric definitely has its flaws.

[DaveW]

WAR relies on defensive metrics. Therefore, by association it's mostly useless for anything less than a full season.

 

Or it shows just how ridiculous UZR can be. As a CF last year he had a UZR/150 of 19.1 this year as a LF he has one of -10.5

 

I know its only a month into the season...but still... these "advanced" stats shouldn't be showing such ridiculous discrepancies.

[gunnarthor]

Joe Posnanski linked to an article where the Oakland A's internal WAR had Cabrera ahead of Trout last year. The take away from that seems to be that the A's and WAR probably differed greatly on how they measured/valued defense and games played.

[Guest USAFChief]

Are you just trolling me now, Dave?

[Brock Beauchamp]

I know its only a month into the season...but still... these "advanced" stats shouldn't be showing such ridiculous discrepancies.

 

It's only natural for something to show a huge discrepancy when you're dealing with small variation numbers (most OFers get to the same balls) and small overall chances (partial season). Two miscues and Harper looks like Delmon Young if you're only using one month as a sample.

 

There's simply no way to create a defensive metric that will be accurate over short periods of time. The data isn't there to draw an accurate representation.

[cmathewson]

Have you seen Delmon play defense? Or should I say, have you seen him TRY to play defense? Scoring runs on offense gets negated if you flop around like a fish out of water in the outfield, hence Delmon not being that much higher than Punto in terms of WAR.

 

There is no way Delmon let 205 total bases happen in one season. He has generally been awful in his career. But he actually was not that bad that year. He was certainly better than Willingham last year, by my scouting eyes.

 

Also, as Brock mentioned, it calculates Punto's WAR based on UZR, which, over 88 games, way overestimates the value of his defense on winning. To be more accurate, it should measure the number of outs recorded rather than the rate of outmaking.

[Boom Boom]

What I'm hearing is that it's not WAR that needs fixing so much as the defensive metrics that go into calculating WAR.

 

Some consistency between Fangraphs and B-R would also help.

 

I think that most of the time you get a better picture by looking at the individual components of WAR than by smashing them all together in an amalgamated statistic.

[jay]

Re-post from a previous thread...

 

In my mind, we can't quite use it as the end-all-be-all just yet because it has a ways to go in a few areas --

 

- defensive metrics: it's crucial to include this in player value, but we just aren't incredibly accurate at putting values on this today (especially for C, 1B)

- positional values: this is accounted for and a large factor, but we miss team specific situations, ie - when a player may not be at his best value position due to roster composition.

- situational hitting/pitching: certain players are more skilled in this than others and we can't properly value the player capable of grounding to the right to score/advance a runner compared to those who swing away.

 

So, as long as we keep a margin of error in mind and don't try to use it to proclaim that my guy with 3.6 WAR this year is better than your guy with 3.4 WAR... then let's get our WAR on.

[Nick Nelson]

The problems with WAR are simple: 1) people try to use it as a comprehensive indicator of player value rather than what it is -- another interesting yet imperfect metric; 2) it is heavily influenced by defense and we still have no great way to measure that.

 

Even proponents of WAR would likely tell you that the statistic is pretty worthless in a one-month sample, so the examples you've pointed out aren't very compelling even though I generally agree with what you're getting at.

[Thrylos]

Context and using the right metric at the right time at the right place.

 

Every metric is broken if you use it inappropriately.

 

WAR is good in summarizing someone's performance during a season and comparing 2 players (position players vs position players; and to a lesser extend pitchers vs. pitchers) for things like post season awards

 

WAR is awesome in summarizing someone's career and comparing him with his peers for things like HOF induction.

 

For looking at weekly or monthly variations, esp in the beginning of a season WAR is not good. But it was never made to do that.

 

Steamrollers are made to pave roads and they are great for it; if you try to mow your lawn with one, you will be disappointed....

[diehardtwinsfan]

To be more accurate, it should measure the number of outs recorded rather than the rate of outmaking.

 

I'd bet over the course of a season, the number of outs divided by the number of innings played would present a far better metric than UZR... just my guess. I'm highly skeptical of any defensive metric that needs 3 seasons of full data in order to be considered accurate.

[snepp]

I'd bet over the course of a season, the number of outs divided by the number of innings played would present a far better metric than UZR... just my guess. I'm highly skeptical of any defensive metric that needs 3 seasons of full data in order to be considered accurate.

 

What you're referring is the "range factor" stat. Which not only has its own problems with sample size, but it compounds them by including the ginormous problem of pitching staff composition.

 

Low-k staff = more balls in play = more outs for the fielders = better fielders?

 

 

Just my guess, but that would be dramatically worse. I'd even bet on it.

[DaveW]

What you're referring is the "range factor" stat. Which not only has its own problems with sample size, but it compounds them by including the ginormous problem of pitching staff composition.

 

Low-k staff = more balls in play = more outs for the fielders = better fielders?

 

 

Just my guess, but that would be dramatically worse. I'd even bet on it.

I know this is going to sound really dumb, and prob will lose my stats credentials, but I have yet to find an advanced defensive metric that I think is worth much of a damn. I mean, when it comes to defense I think the "eye test" and scouts grades are truly the best way to measure.

 

Though I am a big fan of advanced metrics for hitting and pitching, nobody has done a good job on defense and thus the more I look at WAR the more and more I find myself ignoring it. If I haven't seen a guy play defense much live/on TV then I will just go off a scouting report or ask someone else who watches enough games.

 

It doesn't take a genius or huge sample size of games (50?) to realize if a guy is Amazing, Great, Very Good, Above Average, Average, Below Average, Mediocre, Bad, Poor, Delmon Young in the field.

 

Perhaps maybe we should start figuring a way to grade a guy defensively on an F to A+ scale and figure out a way to assign "value" that way? Arm+Range+Speed+IQ all included, then perhaps you weight certain positions differently then others: i.e. SS, CF weighted more heavy then LF/RF/1B.

[gunnarthor]

What you're referring is the "range factor" stat. Which not only has its own problems with sample size, but it compounds them by including the ginormous problem of pitching staff composition.

 

Low-k staff = more balls in play = more outs for the fielders = better fielders?

 

 

Just my guess, but that would be dramatically worse. I'd even bet on it.

 

I honestly don't know but I had thought that Dewin (or what ever his name is) +/- system did use range factor to a pretty big degree.

[snepp]

I was under the impression that the +/- system was put together in cooperation with BIS's video scouting and their defensive tracking data.

[Guest USAFChief]

I was under the impression that the +/- system was put together in cooperation with BIS's video scouting and their defensive tracking data.

I think it's morphed into "Defensive Runs Saved (DRS)". And yes, it's a product of Dewan's BIS.

[kab21]

Context and using the right metric at the right time at the right place.

 

Every metric is broken if you use it inappropriately.

 

WAR is good in summarizing someone's performance during a season and comparing 2 players (position players vs position players; and to a lesser extend pitchers vs. pitchers) for things like post season awards

 

WAR is awesome in summarizing someone's career and comparing him with his peers for things like HOF induction.

 

For looking at weekly or monthly variations, esp in the beginning of a season WAR is not good. But it was never made to do that.

 

Steamrollers are made to pave roads and they are great for it; if you try to mow your lawn with one, you will be disappointed....

 

For the first time ever I agree with something Thrylos typed. Between this and Nick's post there is a pretty good summary of WAR. Currently it's probably the best metric for valuing players (min of one season required). The hitting valuations are much better than defense and pitching though. At best someone should use multiple seasons to roughly determine player's value though. Looking at one month (like Dave) or even a single season can be misleading. I also discount players that only have a decent WAR because of position played and good defense. If a player can't hit then he's not very good even if he manages a 2-3 WAR.

 

I find it a bit ironic that they strip luck out of the pitching stats and use FIP but they don't take luck out of the hitting stats.

[Brock Beauchamp]

Context and using the right metric at the right time at the right place.

 

Every metric is broken if you use it inappropriately.

 

WAR is good in summarizing someone's performance during a season and comparing 2 players (position players vs position players; and to a lesser extend pitchers vs. pitchers) for things like post season awards

 

WAR is awesome in summarizing someone's career and comparing him with his peers for things like HOF induction.

 

For looking at weekly or monthly variations, esp in the beginning of a season WAR is not good. But it was never made to do that.

 

Steamrollers are made to pave roads and they are great for it; if you try to mow your lawn with one, you will be disappointed....

 

I agree with all of this 100%.

[drivlikejehu]

Perhaps maybe we should start figuring a way to grade a guy defensively on an F to A+ scale and figure out a way to assign "value" that way? Arm+Range+Speed+IQ all included, then perhaps you weight certain positions differently then others: i.e. SS, CF weighted more heavy then LF/RF/1B.

 

Fangraphs already has pretty much exactly that. There is a "fan scouting report," and the positional adjustment is fairly straightforward (though not 100% precise due to variation over time).

[jay]

Fangraphs also happened to take a look at the exact impetus of this thread -- WAR for Marte vs Harper season-to-date.

 

I agree in full that the defensive metrics need to continue improving, but I think there's also some validity to the point that it's much easier for us to quantify offensive contributions in our minds than it is to assess defense or baserunning (which is how Marte is so close to Harper in WAR).

 

Here's what Dave Cameron published yesterday looking at removing various defensive components to see if it made WAR more accurate (it didn't): WAR: Imperfect but Useful Even in Small Samples | FanGraphs Baseball