Does batter WAR in OOTP include his fielding?

[kidd_05_u2]

Originally Posted by uruguru

About a year ago I did a deep dive into the underlying assumptions and calculations of WAR. There was a section about determining the relative contribution of pitching and hitting, and the underlying ratio was driven in part by the amounts teams spent on hitters and pitchers. That's how I recall it, but my memory may be fuzzy.

Money spent on batter/pitchers was used because it seemed a reasonable assumption. Do you know why it is still used? Because the data showed that it is a very good assumption.
Go calculate WAR under a 50-50 split for batter-pitchers and check if it still correlates equally well with wins.

It's funny that you criticize WAR for not having a fully developed structural model linking how every variable affects value, and yet you have no issue about evaluating pitcher performance by ERA.

You can criticize the assumptions all you want. We have several years of data showing that the assumptions are more than reasonable for what the stat is trying to capture. And we have a lot of data showing that WAR is a very good statistic overall for what it is trying to do.

[uruguru]

Originally Posted by kidd_05_u2

You can criticize the assumptions all you want. We have several years of data showing that the assumptions are more than reasonable for what the stat is trying to capture. And we have a lot of data showing that WAR is a very good statistic overall for what it is trying to do.

ok.

The error bars for WAR are very high. BBref will tell you up front that differences of up to 2 WAR between players are not decisive. The average player is about 2 WAR. Think about that.

And they report WAR values in tenths of a WAR.

Random chance says that 4% of all MLB'ers (1 in 25) will have their WAR overestimated (or underestimated) relative to their actual value for every single season of their career..... with those error bars.

Look, there's no mathematical rigor to the calculation. The highest-leverage value, the replacement pct of .294, is crazily assumed to be a fixed number even though this concept screams out seasonal variability, yet the factors for ballparks, literally the most static thing in the sport, are calculated in a way that they fluctuate every season.

There is not a single person in this forum who, without a couple of hours and access to google, could even tell you what data is needed to calculate WAR let alone be able to calculate it from data or even be able to validate a value coming from OOTP, bbref or w/e. It's all trust and no verify at this point.

I was a sabermetric geek who stopped following baseball in 2005 (steroid scandal) and didn't come back until around 2017 and suddenly everyone is using this cool stat. So naturally, I wanted to understand it. It's pretty fuzzy as for as baseball metrics go.

edit: btw here is how they divided up wins between pitchers and non-pitchers. It's comical and statistically indefensible. Even if you thought this was a valid approach, this ratio would change every season but somehow the first time they did it they got it just right.

From BBRef:

The wins, and therefore the runs, are further divided between pitchers and position players. We assign 41% to the pitchers and 59% to the position players. This corresponds to the salaries of free agent pitchers vs. hitters over the last four seasons.

I repeat: You need to know the salaries of free agent signings as part of the process to calculating WAR. It is a great concept for a metric, but it's a garbage stat.

[MathBandit]

Originally Posted by uruguru

Have you ever attempted to calculate WAR for a player?

Yes. As has everyone who has ever made a comment that X player is overpaid, or that Y player's contract is a steal. Which is my point.

[uruguru]

Originally Posted by MathBandit

Yes. As has everyone who has ever made a comment that X player is overpaid, or that Y player's contract is a steal. Which is my point.

You really want me to believe that you did all of this because you thought some guy was overpaid:

https://www.baseball-reference.com/a...position.shtml

[MathBandit]

Originally Posted by uruguru

You really want me to believe that you did all of this because you thought some guy was overpaid:

https://www.baseball-reference.com/a...position.shtml

No, I don't want you to believe that, because I have never done that. Which is why I didn't say that I did that.

What I said is that everyone (in the history of baseball) who has ever said that Player X is overpaid, or Player Y is on a great contract, is using WAR.

[uruguru]

Originally Posted by MathBandit

No, I don't want you to believe that, because I have never done that. Which is why I didn't say that I did that.

What I said is that everyone (in the history of baseball) who has ever said that Player X is overpaid, or Player Y is on a great contract, is using WAR.

Fair enough. There was a miscommunication. I asked if you had ever calculated WAR, and you said yes. Which of course would be silly. I appreciate you clarifying it.


Yes, I agree that people use WAR all of the time as if it is authoritative. As someone who is both a baseball geek and a math geek, I see it as sort of a mass delusion or maybe (more accurately) a misguided and uncritical acceptance of expert authority. Those guys must know what they are doing!

[kidd_05_u2]

The WAR discussion was settled a long time ago. The theory behind it is sound and the stat correlates with wins like no other of its kind. Yes, there are assumptions that hopefully one day we can replace for more robust explanations and hard data, but if the assumptions are reasonable and the stats output proves it works, then it is a good stat. Have you ever looked at how well WAR correlates with team wins? You probably never did, because otherwise you wouldn't be saying it is a garbage stat.

You being unable to understand the concepts or replicate the stat doesn't make it a garbage stat. I'm sure you can't replicate the complex math behind your car insurance premium and probably the concepts behind that math escape you too. And yet, you probably trust that "those guys must know what they are doing!"

If you want to keep pretending you are a stats geek and a math geek, even though it is very obvious that you are not, go on dude. Sorry for baiting you into another discussion. Won't do it again because I'm tired of rehashing discussions that were settled a decade ago.

[Rain King]

Originally Posted by uruguru

Fair enough. There was a miscommunication. I asked if you had ever calculated WAR, and you said yes. Which of course would be silly. I appreciate you clarifying it.


Yes, I agree that people use WAR all of the time as if it is authoritative. As someone who is both a baseball geek and a math geek, I see it as sort of a mass delusion or maybe (more accurately) a misguided and uncritical acceptance of expert authority. Those guys must know what they are doing!

What statistics do you prefer?

[MathBandit]

Originally Posted by uruguru

Fair enough. There was a miscommunication. I asked if you had ever calculated WAR, and you said yes. Which of course would be silly. I appreciate you clarifying it.


Yes, I agree that people use WAR all of the time as if it is authoritative. As someone who is both a baseball geek and a math geek, I see it as sort of a mass delusion or maybe (more accurately) a misguided and uncritical acceptance of expert authority. Those guys must know what they are doing!

No, you don't understand me (still). I do calculate WAR on most players in baseball. As do you. As does anyone who has an opinion on which baseball players are good, bad, underpaid, overpayed, underrated, or overrated.

When Wally Pipp lost his starting job to Lou Gherig? That was because of the manager calculating WAR. When Miguel Cabrera stole Mike Trout's MVP in 2012? That was because of the WAR calculation by the sportswriters. When Jimmy at an NYC bar moans that Donaldson sucks and Cashman doesn't know what he's doing? It's because of Jimmy's WAR calculations.

[Matt Arnold]

Some people get too caught up on WAR, especially small differences. Just because 1 guy put up 4.5 WAR, and another guy 4.2, those are certainly close enough to think of them as roughly equivalent. And I don't think the translation from player WAR to team wins is as close as people truly believe. Like, I feel that if you replaced Springer (1.5 WAR) with Acuna (6.3 WAR) on the Jays, I would think they would be more than 5 games ahead of where they are now.

Where WAR gets tricky when you see it on sites is not mentally factoring in the positional adjustment. So you see Freddie Freeman with a big minus on the defensive side of his WAR calculation despite him being a good fielder at 1B, but that's heavily influences by the position adjustment. Or overreacting to defensive metrics. Everyone is pretty good at agreeing on how to value offense - you can have some arguments about how much to value a HR or a BB, sure. But defense still has massive error bars. Like Bobby Witt, who all the metrics agreed was terrible defensively last year:
UZR ranks him as basically an average SS this year. DRS has him as slightly above average (+2 DRS. Statcast Outs above Average ranks him as fantastic (+13). Which defensive metric you choose to believe will have a major impact on his WAR on the season.

[Daniel_09]

Originally Posted by Matt Arnold

Some people get too caught up on WAR, especially small differences. Just because 1 guy put up 4.5 WAR, and another guy 4.2, those are certainly close enough to think of them as roughly equivalent. And I don't think the translation from player WAR to team wins is as close as people truly believe. Like, I feel that if you replaced Springer (1.5 WAR) with Acuna (6.3 WAR) on the Jays, I would think they would be more than 5 games ahead of where they are now.

Where WAR gets tricky when you see it on sites is not mentally factoring in the positional adjustment. So you see Freddie Freeman with a big minus on the defensive side of his WAR calculation despite him being a good fielder at 1B, but that's heavily influences by the position adjustment. Or overreacting to defensive metrics. Everyone is pretty good at agreeing on how to value offense - you can have some arguments about how much to value a HR or a BB, sure. But defense still has massive error bars. Like Bobby Witt, who all the metrics agreed was terrible defensively last year:
UZR ranks him as basically an average SS this year. DRS has him as slightly above average (+2 DRS. Statcast Outs above Average ranks him as fantastic (+13). Which defensive metric you choose to believe will have a major impact on his WAR on the season.

This is true, many of us go by advanced stats or statcast on the defensive side, thinking that the input is automated as with hitters and pitchers, but not on the defensive side the input is mostly decided or entered by several people depending on their perception of what is an above average, average and below average fielder's play or route, hence the difference of the same player in various advanced defensive stats that affect the WAR calculation.

[uruguru]

Originally Posted by kidd_05_u2

The WAR discussion was settled a long time ago. The theory behind it is sound and the stat correlates with wins like no other of its kind. Yes, there are assumptions that hopefully one day we can replace for more robust explanations and hard data, but if the assumptions are reasonable and the stats output proves it works, then it is a good stat. Have you ever looked at how well WAR correlates with team wins? You probably never did, because otherwise you wouldn't be saying it is a garbage stat.

You being unable to understand the concepts or replicate the stat doesn't make it a garbage stat. I'm sure you can't replicate the complex math behind your car insurance premium and probably the concepts behind that math escape you too. And yet, you probably trust that "those guys must know what they are doing!"

If you want to keep pretending you are a stats geek and a math geek, even though it is very obvious that you are not, go on dude. Sorry for baiting you into another discussion. Won't do it again because I'm tired of rehashing discussions that were settled a decade ago.

ok, whatever. Your response to the objections is

a) settled long ago (i.e. don't address them)

b) I don't understand (I do)

c) I'm not a stat geek (I am)

You never actually address the objections, and just start with the ad hominems!

Like I said, I stopped following baseball in 2005 and was away for 10+ years and came back to find WAR. This stat was created and refined by some of the same people who had used statistical software that I wrote to parse Retrosheet game logs and generate statistical reports back in the early 2000s.

Look I made my comment about WAR and got off my soapbox, but then some guy implied that I didn't like WAR because I didn't understand it. Then when I start bringing up some of the crazy garbage in the stat that 99% of people who use it don't even know about, then suddenly it's all hand-waving about how I'm an idiot.

There's other stuff in that stat that is not rigorous. WAR was clearly created with a noble objective in mind, but a lot of shortcuts were taken to get there that don't really hold up mathematically.

[uruguru]

Originally Posted by MathBandit

No, you don't understand me (still). I do calculate WAR on most players in baseball. As do you. As does anyone who has an opinion on which baseball players are good, bad, underpaid, overpayed, underrated, or overrated.

When Wally Pipp lost his starting job to Lou Gherig? That was because of the manager calculating WAR. When Miguel Cabrera stole Mike Trout's MVP in 2012? That was because of the WAR calculation by the sportswriters. When Jimmy at an NYC bar moans that Donaldson sucks and Cashman doesn't know what he's doing? It's because of Jimmy's WAR calculations.

ok, thank you again for that clarification! When I said "WAR", I was talking about the stat. When you said "WAR", you were talking about the concept. Yes, I agree completely with that.

Like I said, Wins Above Replacement is absolutely a great concept because it represents how players are actually valued in the sport. But the WAR stat is a really bad implementation of that concept. When people use "WAR" to evaluate players, they are using an very imprecise metric that they believe to be more authoritative and accurate than it is.

If we created a stat to award batting championships and said, "It's like batting average, but for any player it's plus or minus 40 points" we would be laughed out of the room. That's what the WAR stat is, but for the MVP.

[uruguru]

Originally Posted by Rain King

What statistics do you prefer?

Generally position-dependent statistics. We don't have metrics to properly compare the contributions of pitchers vs position players, so I don't pretend to try. In my OOTP leagues, for example, Pitchers don't win the MVP -- that's an award for position players. (also, relievers don't win the Cy, they have their own award but that's really just b/c I play historical where relievers never compete for the Cy)

With regards to pitcher evaluation, I really do appreciate the value of FIP. I remember the real excitement I felt when Voros made his initial DIPS post on rec.sport.baseball, and I immediately modified my software to incorporate that concept.

The DIPS vs ERA argument is not related to my objections about using WAR. To me, there are two things that we measure in the sport: what we think will happen, and what actually did happen. FIP is a great measure of what we think will happen because it isolates pitcher performance from defense. Not perfectly, but pretty good. ERA is a great measure of what actually happened because there is always a significant difference between what we predict will happen vs what did happen, and it's not just the effects of defense. It's mostly real-world RNG, whatever you want to call that.

So when it comes to rewarding pitchers for performance, I use stats like ERA or ERA+ combined with innings pitched. Baseball is still a team sport and no one pitches in a vacuum. Is it biased towards pitchers with great defenses? Yes. But that's preferable (to me) than giving awards instead based on what should have happened but didn't.

Pitchers who do well with ERA & ERA+ get awards. Pitchers who do well with FIP get bigger contracts. That's the bottom line for me.

[Brad K]

When I read the first sabermetrics book there was chapter after chapter about how managers and GMs don't understand the game. How their approach to stolen bases was mostly running them into losses, that sacrifice bunts are usually a loser, that undervalued walks, etc etc etc.

But when it came time to evaluate how much offense to give up for a good defensive player, they relied upon the anecdotal observations and shoot from the hip decisions of a group of people they had already proven didn't know what they were doing. Because that's exactly what applying a percentage to the offensive production of a group of players at a position to determine replacement level is.

Anyway, I have some confidence using it to compare players at the same position but a lot less when at different positions. And that's the problem it's supposed to solve.

[MathBandit]

Originally Posted by uruguru

If we created a stat to award batting championships and said, "It's like batting average, but for any player it's plus or minus 40 points"

I know this isn't the point, but to be fair you basically described exactly Batting Average and ERA right there.

edit:

For the curious, the point at which various metrics have more signal than noise:

* AVG- 910 ABs
* BABIP - 820 Balls in Play
* XBH - 1610 PA

* XBH against - 1450 Batters Faced
* HR rate - 1320 BF
* BABIP against - 2000 BIP

The all-time record for AB in a season is 716. No pitcher has faced 1300+ batters in a seaon in 45 years.

[kidd_05_u2]

Originally Posted by Matt Arnold

I don't think the translation from player WAR to team wins is as close as people truly believe. Like, I feel that if you replaced Springer (1.5 WAR) with Acuna (6.3 WAR) on the Jays, I would think they would be more than 5 games ahead of where they are now.

Are you kidding? The translation from player WAR to team wins is amazingly close and this is easily verifiable. You are nearly always within 3-4 wins of the pythag record [for this calculation you would use Bref WAR by the way].

Do we have evidence that WAR systematically overvalues/undervalues certain kinds of players? Because if not, I would certainly trust WAR over your hunch Matt.
You would think a team with Trout and Ohtani would get close to a .500 record...

[Brad K]

Originally Posted by kidd_05_u2

.

Do we have evidence that WAR systematically overvalues/undervalues certain kinds of players?

Players from certain positions dominate the top of the WAR charts. It seems that if WAR were accurate and mangers put players in appropriate positions then WAR distribution would be equal.

And if managers aren't putting players in appropriate positions then the replacement value established for each position is wrong. Error feeding error.

[MathBandit]

Originally Posted by Brad K

Players from certain positions dominate the top of the WAR charts. It seems that if WAR were accurate and mangers put players in appropriate positions then WAR distribution would be equal.

That premise doesn't hold.

In (American) Football, do you think the 5 best QBs in the NFL help their team win more games than the 5 best Right Guards?

[Matt Arnold]

Originally Posted by kidd_05_u2

Are you kidding? The translation from player WAR to team wins is amazingly close and this is easily verifiable. You are nearly always within 3-4 wins of the pythag record [for this calculation you would use Bref WAR by the way].

Do we have evidence that WAR systematically overvalues/undervalues certain kinds of players? Because if not, I would certainly trust WAR over your hunch Matt.
You would think a team with Trout and Ohtani would get close to a .500 record...

Yeah, it's actually pretty good overall. This year comparing wins above average and team winning pct I get an R^2 value of 0.8. So I guess even taking that unto account would probably handle all the variance of swapping guys.