The hypothetical has nothing to do with WAR.

(but WAR does make sense and people do take it seriously)

Then what was the point of introducing the hypothetical in a discussion about the legitimacy of WAR? Do you even remember what you're arguing?

Posted by Jtpsops on 6/14/2016 5:29:00 PM (view original):
That's not what I'm arguing. I didn't say "NO ONE KNOWS WHAT THE FORMULA IS!!!" I said everyone is not in agreement on what formula is best. Therefore it's hard to put much stock in it as a reliable number.

It would be like if Nerd A says "Smith had 55 homers last year." and Nerd B says "Every park is different. I don't count balls that go over the wall. I count any ball hit over 400 feet as a homer. Smith only had 42 homers".

Both calculations may have Smith as the league leader in homers, but you're not going to get consistency between the two approaches. And if both became accepted by half of fans, it would be very difficult to rely on any statistical homerun data.

And yes, I'm aware that's a really simplistic example, but it's the point that once people start arguing over what the definition of something is, how can you trust the values you're getting?

I was responding to this post with the hypothetical. It's perfectly reasonable for two sites to calculate a stat differently. As long as you know, it doesn't hurt anything.

It's like how wRC+ on Fangraphs is essentially the same as OPS+ on BR. They don't share a name, so it isn't exactly analogous, but they basically do the same thing with a slightly different calculation.

Posted by Jtpsops on 6/14/2016 10:00:00 PM (view original):
That's quite the hypothetical. It supposes that WAR makes sense and everyone takes it seriously.

But it doesn't, and they don't.

You keep repeating this, but it doesn't make it true. A LOT of people look at WAR and "take it seriously." Most of the baseball writing world is included in that group.

Posted by Jtpsops on 6/14/2016 5:08:00 PM (view original):
I never said each method isn't applied equally to all players. But if different methods are used, players are not always going to fall at the same ranking in each method. And if each method produces different numbers for each player or doesn't produce results ranking players in the exact same order, then clearly it's not something that is 100% reliable and must be taken with a grain of salt.

There are concrete, measurable stats (BA, OBP, homers, runs, etc.) that are finite - you can't calculate them differently and come up with a different result. Metrics like WAR leave room for discrepancy. By and large, they give you a good approximation of who the best players are, but there are flaws since it is not a concrete stat.

How is it true that if something is not 100% reliable it must be taken with a grain of salt? My car is not 100% reliable, but I take it seriously. My experiment is not 100% reliable, but I'm getting a PhD from it, and I think most people will take that seriously. Almost nobody's household scale is 100% accurate, but they still step on them at a ridiculous frequency. Grading systems everywhere just straight-up suck for lots of things. But people take grades seriously - both the students and potential employers, graduate schools, etc.

Ultimately, almost no numerical measures of anything are really 100% accurate in the real world. So what? They're still useful approximations to things.

OK, since about 50% of this is blocked, can anyone give me an example of differing WAR values for a handful of players?

Is it a case of Player A being 5.7 on one site and 5.6 on another? Does the top 20 have some huge variance where Player B is 12th on one list and 23rd on another?

I love the smell of a retard fight in the morning.

Top 10:

Fangraphs:

1. Kershaw - 4.6
2. Bogaerts - 3.9
3. Trout - 3.7
4. Machado - 3.5
5. Altuve - 3.4
6. Arenado - 3.1
7. Donaldson - 3.1
8. Syndergaard - 3.1
9. Fowler - 3.0
10. Fernandez - 3.0

Baseball-Reference:

1. Kershaw - 4.3
2. Marte - 3.5
3. Altuve - 3.5
4. Trout - 3.5
5. Bogaerts - 3.5
6. Arrieta - 3.4
7. Machado - 3.4
8. Arenado - 3.4
9. Syndergaard - 3.3
10. Donaldson - 3.3

Each site has 2 players in the Top 10 that don't even appear in the other's Top 10. None have the same WAR value on each site, and only Kershaw has a consensus ranking. And this is just two sites. But please, explain to us how that's "reliable".

I guarantee no matter which site I go to, Kershaw will have the same ERA, same WHIP, same number of K's, etc.

Unblocking jtpops usually makes me dumber but, in this case, he's sort of proving his point.
Marte is 3.5 on BR less than 3 on FG. Let's say he's 2.9. That's a 17% difference. If we were comparing homers, it's 30 to 25. That's the most glaring example. We know he's hitting .335 with 26 xbh, 19 SB and 7 walks. Why the big difference between the two sites? Is it dWAR? I hope so because I love me some dWAR bullshit.

Unlike jtpops, I don't need the order to be exact. I don't mind that one guy is 5th and 12th. But, if the numbers are 17% different, there's a problem somewhere. Use 17% in record(wins). If this is Wins Above Replacement, 17% is a lot.

There's the subjectivity.

Think Marte and Bogaerts basically have the same value? Use the BR numbers, because BR says they're the same.

Like Bogaerts more than Marte? Use the FG numbers, because FG says that Bogaerts is 1.0 better than Marte.

dahs and BL don't like the word subjectivity. It confuses them.

Yeah, BBR sees Marte as being basically a win better defensively than FanGraphs. Even so, 3.5 and 2.9 are not all that different when you're dealing with a small sample size. If you make the sample size bigger, things get better:

FanGraphs		Baseball-Reference
Babe Ruth	168.4	Babe Ruth+ (22)	183.6
Barry Bonds	164.4	Barry Bonds (22)	162.4
Willie Mays	149.9	Willie Mays+ (22)	156.2
Ty Cobb	149.3	Ty Cobb+ (24)	151.1
Honus Wagner	138.1	Hank Aaron+ (23)	142.6
Hank Aaron	136.3	Tris Speaker+ (22)	133.7
Tris Speaker	130.6	Honus Wagner+ (21)	131
Ted Williams	130.4	Stan Musial+ (22)	128.1
Rogers Hornsby	130.3	Rogers Hornsby+ (23)	127
Stan Musial	126.8	Eddie Collins+ (25)	123.9
Eddie Collins	120.5	Ted Williams+ (19)	123.2
Roger Clemens	133.7	Cy Young+ (22)	170.3
Cy Young	131.5	Walter Johnson+ (21)	152.3
Walter Johnson	117.1	Roger Clemens (24)	139.4
Greg Maddux	116.7	Pete Alexander+ (20)	117
Randy Johnson	110.6	Kid Nichols+ (15)	116.6
Nolan Ryan	106.7	Lefty Grove+ (17)	109.9
Bert Blyleven	102.9	Tom Seaver+ (20)	106.3
Gaylord Perry	100.1	Greg Maddux+ (23)	104.6
Pete Alexander	96.5	Randy Johnson+ (22)	104.3
Steve Carlton	96.5	Phil Niekro+ (24)	97.4

They're closer on batters because the way WAR is calculated for batters by the 2 sites is fundamentally very similar, whereas the way it's calculated for pitchers is fundamentally quite different - BBR uses ERA (really ERA+) and FanGraphs utilizes expected run allowance values. I like the BBR method better because it is a truer representation of what really happened on the field, but most statnerds tend to prefer the FanGraphs version because they think it reduces the impact of luck. But for batters, the top 11 are identical with a similar order.

So in statistical analysis, a 9% difference (Ruth's WAR numbers) isn't that significant to you?

And the difference in the pitching lists/calculations further emphasizes how subjective WAR is. You say yourself that each site values different stats that they input into their formulas. I'm not sure how you can argue against the fact that WAR is inconsistent.