Question about WAR calculation

[big dog]

I have a question, and while the obvious answer to some is "because WAR sucks", that's not really what I'm looking for here...    If the moderators want to just scrub this whole thing, I understand. 

 

I was looking at some stats this morning and stumbled across this stat line when sorting by WAR.

 

Fielder, 1.7 vs Hosmer, 1.8  (but those are rounded numbers and they are listed next to each other in a ranking of all players, so it's probably very close to a tie).

 

When I actually look at the stats, here are the direct comparisons, Fielder vs. Hosmer

AB, 272 vs 260

R, 35 vs 39

H, 93 vs 76

2B, 16 vs 14

3B, 0 vs 2

HR, 11 vs 8

RBI, 47 vs 39

SB, 0 vs 3

CS, 0 vs 2

BB, 27 vs 27

SO, 37 vs 52

AVG, .342 vs .292

OBP, .412 vs .359

SLG, .522 vs .454

OPS, .934 vs .812

 

for fielding stats:

E, 4 vs 1

FPCT, .949 vs .998

RF, 7.26 vs 9.14

 

It seems to me that the gap in hitting is pretty strong in favor of Fielder.  My question to those who understand the WAR calculation- what creates the advantage for Hosmer that actually makes his WAR at least a little bit higher?  I assume it is some combination of fielding value and park effect, but I'd like to know more about it.  I would guess that 3 more errors and less range at first base doesn't really make up for 122 points of OPS, so what else puts the calculation slightly in Hosmer's favor?

 

Thank you.

 

 

[drjim]

Fielder has 59 games at DH, 11 at 1B, which would explain the discrepancy.

 

Also, in my opinion (and surely to derail the debate), WAR really overstates the value of defense relative to offense.

[Mike Sixel]

Just glancing at FG.......there is also a delta in base running, but I think that is a small component.

 

the delta on offense is dramatic (as it is on defense). If you just want to think about offense, and ignore defense and positional value (which is a big part of the defensive adjustment)......Fielder is way up there.

 

if you think playing a position where you also can save runs has value, and if you think playing a harder position has even more value, then you have to weight in defense some.....how much, I agree, is debatable.

 

 

[big dog]

Thanks very much, I haven't watched a Rangers game all year and I didn't realize Fielder was mainly a DH.

 

Agreed about the need for a defense/positional adjustment, too, no question, however they want to try and measure it.

 

 4 errors in 11 games at first base...yikes.  He wasn't too bad in the field previously, or at least I didn't think he was.

 

I appreciate the feedback.

[markos]

Several adjustments are made that end up pulling Hosmer up to Fielders overall value:

 

1) Position adjustment: Hosmer has played the field, while Fielder is primarily a DH. Players are given a credit for playing the field (the value of the credit depends on the position).

2) Stadium adjustment: The park in Texas is one of the best hitters parks in all of baseball. The park in KC is roughly league average

3) Baserunning: Hosmer has been the better baserunner (taking extra bases, etc).

4) Defense: When playing 1B, Fielder is (and has been for several years now) pretty terrible, while Hosmer is (and has been for several years now) solidly above average.

 

After adjusting for stadium differences, Fangraphs has Fielder ~10 runs better than Hosmer from their respective batting lines. However, Hosmer's superior defense (~5), baserunning (~3) and positional adjustment (~2) basically make up the difference. 

 

By the time I pressed "Post" this is basically just a copying what others have already said.

[HitInAPinch]

WAR is one of the Sabermetrics that I feel has more use in the "wrap up" paragraph on a team's scouting report than it does in real life. 

 

WHIP is the one where I ask:  why the hell haven't we been doing this before?  It makes a lot of sense in judging  pitcher.

[jimmer]

 

WAR is one of the Sabermetrics that I feel has more use in the "wrap up" paragraph on a team's scouting report than it does in real life. 

 

WHIP is the one where I ask:  why the hell haven't we been doing this before?  It makes a lot of sense in judging  pitcher.

WHIP has been around a long time.

[gunnarthor]

 

 

Also, in my opinion (and surely to derail the debate), WAR really overstates the value of defense relative to offense.

I think this is true.  A few years ago during the first Miggy v. Trout debate, Joe Posnanski revealed that the A's internal WAR calculation actually rated Miggy ahead of Trout but fWAR had something like a 3 WAR advantage for Trout.  And he also quoted Bill James who was pretty critical of WAR.

 

WAR is good in that it tries to go beyond mere rate stats (that used to be the real argument) but it has problems.  About 6 or 7 years ago, b-r recalculated their WAR.  For the most part, it didn't have a great impact on most players but one of the top 20 recalculated players was Torii Hunter, who gained, on average, over 1 WAR per season.  When he was going into FA, a lot of sabretypes ripped on the Twins and Angels for over valuing him.  Gleeman was pretty bad.  We read a lot about how Hunter had had only 1 3 WAR season in MN and teams were dumb for over paying for him and fans were clouded by his personality.  And then, lo and behold, a few years later, WAR is readjusted and those teams were right.

 

BP had a recent article on Jack Morris' HOF candidacy and suggested that the narrative pushed by stat-heads the last decade was wrong.

[jimmer]

 

 

BP had a recent article on Jack Morris' HOF candidacy and suggested that the narrative pushed by stat-heads the last decade was wrong.

ERA and ERA+ were two huge shots against Morris and those aren't exactly advanced stats.  They are used even by the most stat-adverse fans.

 

Ironically, it was a stat-head using a pretty advanced (and very AMBITIOUS) stat that now suggests he may have been HOF worthy.  If it was just advanced metrics saying no during his ccandidacy, he likely would have made it years ago when voters pretty much ignored those.

[nicksaviking]

WAR is one of the Sabermetrics that I feel has more use in the "wrap up" paragraph on a team's scouting report than it does in real life. 

 

WHIP is the one where I ask:  why the hell haven't we been doing this before?  It makes a lot of sense in judging  pitcher.

And personally I don't much care for WHiP as it gives a measly walk the same weight as a HR.

[REPETE]

 

And personally I don't much care for WHiP as it gives a measly walk the same weight as a HR.

That's because "Walks Will Haunt"!!!

[HitInAPinch]

 

And personally I don't much care for WHiP as it gives a measly walk the same weight as a HR.

I see what you mean.  I'm on the other side:  I think too much statistical analysis ruins what is, by definition, a game.  If I were big into Fantasy Baseball, it would probably be a completely different story

[drjim]

 

I see what you mean.  I'm on the other side:  I think too much statistical analysis ruins what is, by definition, a game.  If I were big into Fantasy Baseball, it would probably be a completely different story

 

There is a pretty easy way to avoid this. And I think it should be obvious that for many people the statistical nature of baseball and development of this analysis makes them a much bigger fan than they would be otherwise.

 

My personal complaint, as far as new trends are concerned, is that analysis/discussion of the game has slanted too much towards roster construction/transactions/managerial decisions and away from the actual players and how they actually perform. There is too much emphasis on everything except the actual games that are played.

[jimmer]

 

I see what you mean.  I'm on the other side:  I think too much statistical analysis ruins what is, by definition, a game.  If I were big into Fantasy Baseball, it would probably be a completely different story

Everyone is entitled to enjoy the game in their own way. I'm glad your way works for you.

[Otto von Ballpark]

 

Also, in my opinion (and surely to derail the debate), WAR really overstates the value of defense relative to offense.

Really?

 

Take Fielder for example.  At B-Ref, he has +19 batting runs above average (Rbat) this year.  Add in the +11 runs that average is above replacement level (Rrep), subtract -3 for baserunning (Rbaser and Rdp), and he's basically got a +27 over replacement for raw offense, without consideration for fielding or defensive position.  Which dwarfs his fielding runs (Rfield, -4) and/or positional adjustment (Rpos, -6).

 

What kind of ratio would be better, in your opinion?  If you skew it any more towards offense, I think you'd be reducing the value of fielding and defensive position too close to zero for my taste.  We know those things matter, and they can take some of the bite out of an otherwise good hitter (or boost the value of a lesser hitter), so it's good that they are meaningfully reflected in WAR.

[drjim]

 

Really?

 

Take Fielder for example.  At B-Ref, he has +19 batting runs above average (Rbat) this year.  Add in the +11 runs that average is above replacement level (Rrep), subtract -3 for baserunning (Rbaser and Rdp), and he's basically got a +27 over replacement for raw offense, without consideration for fielding or defensive position.  Which dwarfs his fielding runs (Rfield, -4) and/or positional adjustment (Rpos, -6).

 

What kind of ratio would be better, in your opinion?  If you skew it any more towards offense, I think you'd be reducing the value of fielding and defensive position too close to zero for my taste.  We know those things matter, and they can take some of the bite out of an otherwise good hitter (or boost the value of a lesser hitter), so it's good that they are meaningfully reflected in WAR.

 

I wouldn't try to convert defensive metrics into runs and I certainly wouldn't try to equate said conversion with offensive production and create one number. Debating the correct ratio is a distraction.

 

I'm skeptical of defensive metrics but I can accept them as a quick and dirty directional list within a position subgroup (with several caveats). But calculating them into a precise number and merging it with offensive numbers (and a lesser extent baserunning) is a few steps too far for me.

[Otto von Ballpark]

 

My personal complaint, as far as new trends are concerned, is that analysis/discussion of the game has slanted too much towards roster construction/transactions/managerial decisions and away from the actual players and how they actually perform. There is too much emphasis on everything except the actual games that are played.

Well, there's more discussion now, period (thanks, Twins Daily!).  But I don't think it breaks down the way you suggest.

 

Obviously, before the 1970s, there wasn't a whole lot to discuss in transactional terms, thanks to the reserve clause.  But by the same token, there was so little detailed data on player performances that I don't think there was really any greater discussion going on about that either -- how much can you really analyze/discuss a basic box score or season/career stat line?

 

First with the advent of Retrosheet, and now with video and PitchF/X, there is more discussion about player splits, swings, pitches, etc. than ever before.  (Thanks, Parker!)

 

And I would suspect that managerial decisions and roster management have been at the forefront of baseball discussion since time immemorial.

[Otto von Ballpark]

 

I wouldn't try to convert defensive metrics into runs and I certainly wouldn't try to equate said conversion with offensive production and create one number. Debating the correct ratio is a distraction.

The sum and interplay of offensive and defensive contributions are pretty darn important in this sport, and you're going to have to do it some way.  I welcome quantitative efforts in that direction, as well as keeping qualitative ones.

[Mike Sixel]

 

I wouldn't try to convert defensive metrics into runs and I certainly wouldn't try to equate said conversion with offensive production and create one number. Debating the correct ratio is a distraction.

 

I'm skeptical of defensive metrics but I can accept them as a quick and dirty directional list within a position subgroup (with several caveats). But calculating them into a precise number and merging it with offensive numbers (and a lesser extent baserunning) is a few steps too far for me.

 

Given that runs are all that matters, how would you measure defense?

[drjim]

 

Given that runs are all that matters, how would you measure defense?

 

Separately and with a different unit of measure than offense.

[drjim]

 

The sum and interplay of offensive and defensive contributions are pretty darn important in this sport, and you're going to have to do it some way.  I welcome quantitative efforts in that direction, as well as keeping qualitative ones.

 

They are important and I welcome all efforts as well. Just don't think WAR is it.

[Otto von Ballpark]

Separately and with a different unit of measure than offense.

Separate but equal, yeah that always works.

 

Seriously, any separate measure of defense that you devise or endorse is going to get converted to a scale similar to offense, by the teams themselves as well as us fans. Otherwise, how do you know if, say, Aaron Hicks' defense makes up for his lack of offense?

 

Edit to add: that doesn't mean WAR itself is the best or definitive tool for this type of evaluation, but any tool or evaluation one uses will lead to the same comprehensive form as WAR.

[Otto von Ballpark]

Just don't think WAR is it.

One thing that is important to understand about WAR is that it is not really one true number that gets dictated to us, no matter how it is presented in articles or comparisons like this.

 

Both major variants of WAR publish all the separate components of the measure, and for good reason. I think even they understand that the final total WAR is an imperfect measure, and there are many valid cases where the components are more useful individually or weighted/regressed differently.

 

I don't agree with a lot of total WAR-based premises either, but I think the components of WAR are still very useful and possibly the best comprehensive "layman stats" we may ever have in the sport (the next step above "layman's stats" would be numbers derived from PitchFX and Statcast data, based on physics).

[big dog]

 

One thing that is important to understand about WAR is that it is not really one true number that gets dictated to us, no matter how it is presented in articles or comparisons like this.

Both major variants of WAR publish all the separate components of the measure, and for good reason. I think even they understand that the final total WAR is an imperfect measure, and there are many valid cases where the components are more useful individually or weighted/regressed differently.

I don't agree with a lot of total WAR-based premises either, but I think the components of WAR are still very useful and possibly the best comprehensive "layman stats" we may ever have in the sport (the next step above "layman's stats" would be numbers derived from PitchFX and Statcast data, based on physics).

Thanks, everyone, for the thoughts.  This is the context in which I take WAR- one way of looking for the answer, not THE answer.  I was just wondering, as a novice, why two players with such dissimilar stats could have the same ranking in this particular system.  I appreciate all the insights.

[drjim]

 

One thing that is important to understand about WAR is that it is not really one true number that gets dictated to us, no matter how it is presented in articles or comparisons like this.

Both major variants of WAR publish all the separate components of the measure, and for good reason. I think even they understand that the final total WAR is an imperfect measure, and there are many valid cases where the components are more useful individually or weighted/regressed differently.

I don't agree with a lot of total WAR-based premises either, but I think the components of WAR are still very useful and possibly the best comprehensive "layman stats" we may ever have in the sport (the next step above "layman's stats" would be numbers derived from PitchFX and Statcast data, based on physics).

 

This is a good way to describe it and I would endorse this usage. For quick and dirty ways to compare players across positions and stadiums (and eras) it certainly has its usages and is probably even the best tool for a fan. But I wouldn't use it as much of an authority or the primary stat that I would cite.

 

And in response to your other quote, I would be quite surprised if teams gave it much thought. They have much more sophisticated systems of evaluation.

[Jham]

DH's don't cost their teams any defensive runs.  They don't even play defense.  A DH costs their teams less run than half of the starting defensive players in baseball, and more than the other half with the guy in the middle being the definition of average starter at his position.  Replacement defensive players aren't really of much use since many non-starters are great defensive players who can't hit or horrible defensive players with no position.  I don't think you should knock a guy who doesn't cost his teams runs on defense.  Nor do I necessarily think you should praise a guy who costs his teams runs at shortstop (on avg) over a right fielder who saves runs.  Is every AL pitcher less valuable to their team because they don't hit?  

[Otto von Ballpark]

in response to your other quote, I would be quite surprised if teams gave it much thought. They have much more sophisticated systems of evaluation.

True they have more sophisticated systems, but the end result for a lot of their MLB transactional analysis is going to be a WAR framework. What are our alternatives to player X, what are the components of player Y's contributions, that sort of thing.

 

Obviously scouting, coaching, and development are important too, especially in other contexts. But it seems pretty obvious that teams would use a WAR framework for things, even before they realized it or someone invented WAR as we fans know it.

[jorgenswest]

Sometimes it seems like not believing in defensive metrics leads to not believing defense matters. The measurement of defense is imperfect. Some believe that since we know it is imperfect, we shouldn't consider it.

 

Defense matters. Measuring it is essential. Those measures will evolve just as measures of offense have evolved beyond batting average.

[USAFChief]

Sometimes it seems like not believing in defensive metrics leads to not believing defense matters. The measurement of defense is imperfect. Some believe that since we know it is imperfect, we shouldn't consider it.

Defense matters. Measuring it is essential. Those measures will evolve just as measures of offense have evolved beyond batting average.

I doubt many people, inside the game or out, believe defense doesn't matter.

 

But I think it's fair to believe current saber metric ideas about how to measure it, and the amount of difference it actually makes (particularly at some positions relative to others), is flawed. Therefore some of us pay little attention to those metrics, which isn't the same as thinking defense can't make a difference.

[dbminn]

WAR is fine if it is viewed as a rough summary. It's like reading the Cliff Notes version of "Macbeth". The full text has a richer meaning than the summary. Seeing the performance is the greatest experience.

 

For baseball, detailed statistical data brings out more about each particular player than WAR. Seeing the actual players is what makes it all worth the debate.

 

The problems are bigger for defensive and pitching than for offensive WAR:

 

Pitching WAR varies quite a bit  based on underlying assumptions. FIP WAR (fielding-independent pitching) is a popular model. Historically, stats people have been unable to separate talent from variability (luck) on balls in play. FIP WAR simply ignores BIP. RA9 WAR is based on what has actually happened on the field but doesn't account for their fielding support (or lack of). For now, it's better to look at both calculations.

 

Defensive WAR is pretty much irrelevant over the short term. Even the developers of dWAR suggest it takes three years of data before there's much confidence in the specific value. The reason? Most plays are routine. Only a few plays in the field are difference makers in a given year.  IMO, dWAR is better viewed separately and in clumps: High dWAR players are the best fielders but differences of 10 or 20 percent are not significant. dWAR is going to get much better with more quantified data (route efficiency, jump, ground covered) once Statcast data is fully available.

 

I suspect managers and coaches use underlying performance data to make decisions, especially with prospects and young players. Contact rates for batters, pitch values (velocity differences between pitches, e.g.) for pitchers. Outcome data like WAR are for fans.