Why do people value WAR/Arbitrary formulas.

[Lucky Jim]

First repeatedly criticizing me and claiming I'm a troll becuase you don't understand the meaning of the word "subjective" isn't my fault chief. Get a dictionary; maybe it will help.

I conceded that I shouldn't have called you a troll. That was wrong of me. Not sure what more you want.

I still think your argument was disingenuous, at best. And, given your attempt at rehabilitating "subjectivity," it remains so.

[Lucky Jim]

First of all, it's my understanding there are 3 WAR models. It doesn't really matter one way or another. Just saying.

What do you mean by "decent?" I mean, what do you think the range of error could possibly be, and are you comfortable saying it falls within "acceptable" range?

I'm not. And a lot of other people aren't either.

You mean...

Either you know how to use a statistic or you don't. There are people of both kinds who embrace and/or reject WAR, for largely relevant reasons. Absent some evidence that you're in the former camp, this thread should probably just die. Aneurysm-like. Not slowly.

[glenn__davis]

Sure it does. Batting average is defined as the percentage of times a player get a hit per official time at bat. Both the number of hits and the number of official at bats is subjectively determined by an official scorer.

Certainly you are correct, but I doubt that the number of times that such a subjective determination would have to made would have much of an impact on the actual statistic. The majority of the time the hit/error ruling is very clear.

[Larry18]

I wish we could could have a discussion about advanced metrics without it becoming a holy war. In my experiences, people who use advanced metrics to say "this proves I'm right" would do the same using conventional statistics. And while it is annoying, it shouldn't be used to condemn an entire system. I tend to look at baseball statistics like a photograph, it can be very good and capture almost all of the scenery, but I'd rather visit the Grand Canyon that see a picture.

[Lucky Jim]

This conversation has gotten pretty esoteric at times. Speaking for myself, I find the offensive component of WAR to be pretty reliable. The decision on how much to weight a single, double, triple, HR, walk, etc. is being based on objective information about the value of each component, and in general, if you were to show me a list of the top 25 guys in oWAR for any particular year I doubt I would have any issue with it. The defensive component has two problems: (1) lack of transparency, (2) just a lot of instances where the results don't seem to agree with the "eye test." Those two problems are interrelated, and the fact that there are still large differences between the results of the top defensive stats like UZR, +/- and Rtot just prevents me from getting nearly as comfortable with the defensive component than the offensive component. But, I still use rWAR and fWAR because there is nothing better available.

This seems reasonable. Your discussion reminds me of the BA --> OPS --> wOBA evolution, itself. My understanding is that the first two, under the argument of a few in this thread, would not be 'subjective,' while the latter, in trying to capture what the former two did not and/or correct limitations/imbalance in how they represent their components, would.

I'll add that your stat/eye binary highlights something I think about frequently - which is that I tend not to trust absolutely what I see, and tend not to trust absolutely what some statistics/models tell me. My confidence is at its highest (though not perfect) when my eyes and the stats agree. It's at its lowest when my eyes and the stats disagree (though I tend to favor the stats). And when the stats disagree with themselves, I think you simply have to hedge your bets. (I tend to look at all defensive metrics before even considering defensive value. And I tend to internally discount fWAR due to its UZR components.) Life is a probability function, as Drungo once said. WAR is just another of framing some small portion that we fit into it. Or don't.

[DrungoHazewood]

And again, this has nothing to do with the subjectivity of the stat, only the subjectivity of scorer.

Ok, you got me. It's an utterly objective stat which relies totally on subjective component metrics.

[Lucky Jim]

Boring thread is boring.

The good news is we don't need any statistical modeling to capture your contribution to it.

[Lucky Jim]

Ok, you got me. It's an utterly objective stat which relies totally on subjective component metrics.

Right, this "purist" approach is really a semantic argument dressed up as a theoretical one.

[Lucky Jim]

See, now this was funny!

And posting Archer videos isn't? This thread has everything. Everything, I tell you.

[DrungoHazewood]

Certainly you are correct, but I doubt that the number of times that such a subjective determination would have to made would have much of an impact on the actual statistic. The majority of the time the hit/error ruling is very clear.

And I'd counter that, over time, the definition of an error (and therefore a hit) has shifted. In the earliest days of baseball a line drive hit two steps to one side of the third baseman was probably unfieldable with the primitive gloves of the era. To some degree this persisted well into the 20th century, and with it were lowered expectations of official scorers. In Cy Young's career errors influenced at least 1/3 of the runs scored against him.

So while we look at .973 fielding percentages across time, and .330 batting averages, and see them as representing the same things, they're really just points picked out of of ever-changing distribution curves.

[CA-ORIOLE]

First of all, it's my understanding there are 3 WAR models.

Well, I only know of rWAR and fWAR as the two primary (accepted) models. What is the third?

It doesn't really matter one way or another. Just saying.

What do you mean by "decent?" I mean, what do you think the range of error could possibly be, and are you comfortable saying it falls within "acceptable" range? I'm not. And a lot of other people aren't either

I understand. You would rather have no statistic/modeling of player value.

[glenn__davis]

And I'd counter that, over time, the definition of an error (and therefore a hit) has shifted. In the earliest days of baseball a line drive hit two steps to one side of the third baseman was probably unfieldable with the primitive gloves of the era. To some degree this persisted well into the 20th century, and with it were lowered expectations of official scorers. In Cy Young's career errors influenced at least 1/3 of the runs scored against him.

So while we look at .973 fielding percentages across time, and .330 batting averages, and see them as representing the same things, they're really just points picked out of of ever-changing distribution curves.

Well sure, I don't disagree with this, but I don't really see a lot of people making comparisons between a player's batting average now vs. the earliest days of baseball. I think most people understand that there are plenty of reasons why those comparisons aren't really valid.

But we DO see many people misusing the WAR value as a standalone way to evaluate a player. I think it sort of goes back to what Frobby said - he understands WAR is flawed, but uses it because it's the best we have. But if the best we have isn't very good, then it shouldn't be used as a standalone statistic. And I know folks like Frobby, yourself, LJ and many others understand this, but many people don't.

[Nigel Tufnel]

I have nothing to add, except that I was just reading a thread on a Bruce Springsteen message board about people's top 5 shows on his current European tour. An argument just started there about how people were being too subjective, and it made me laugh.

[Lucky Jim]

But if the best we have isn't very good, then it shouldn't be used as a standalone statistic.

I'm not sure it's wise to use "very good" as a threshold for using any of these statistics (or whatever you want to call them). But it shouldn't be a "standalone" either. Trea got destroyed for doing this, repeatedly (he did it repeatedly and was thusly destroyed repeatedly). I don't really think it happens much anymore. For most of us, we use it as a kind of heuristic, or snapshot, or just "best evidence" stand-in when we try to analyze value.

I mean, I'm defending WAR, and I wrote this in response to Gordo's invocation of Arrieta 1.6 fWAR just the other day:

Based entirely on his FIP - which doesn't take into account his xBABIP or the fact that his K-rates are inflated relative to his below average SwSt%. Ugh. The absurd use of statistics just continues, unabated.

[CA-ORIOLE]

But we DO see many people misusing the WAR value as a standalone way to evaluate a player. I think it sort of goes back to what Frobby said - he understands WAR is flawed, but uses it because it's the best we have. But if the best we have isn't very good, then it shouldn't be used as a standalone statistic. And I know folks like Frobby, yourself, LJ and many others understand this, but many people don't.

When people misuse it on here they are generally called out. The WAR components and valuations are routinely debated on here. I can understand that there may be issues on other sites and on TV etc, but I really don't see the issue here. I basically see the oppostion here as having a general lack of knowledge and applying a series of logical fallacies, quite frankly.