WAR What is it good for?

dallasactuary

Absolutely nothing.

That probably overstates it a bit, but I've always found WAR (Wins Above Replacement) to be very unreliable. And a little while ago one possible reason hit me and I did a quick check to confirm it. For my quick check I used 1975, for no particular reason. I listed each team, the number of games they won, and the total WAR for their entire roster.

The average team won 80.5 games with WAR of 33.2. What that means is that a team of replacements, according to the method, would win 80.5 - 33.2 = 47.3 games. And that's one problem: a team of nothing but replacement players wouldn't win 47 games. The closest team to that standard in modern times is the 1962 Mets; they only won 40 games and they had a handful of decent players. But leaving that problem to the side, there's a bigger problem. If the method says a team of replacement players wins 47.3 games, then each team's WAR should equal their actual wins minus 47.3. On average, of course, they do, but the variation from team to team is significant. The Reds won 108 games, with WAR of 52.0; a team of replacement players now wins 56 games. Looking at a bad team, the Cubs won 75 games, with WAR of 17.3; a team of replacement players wins 57.7 games.

At the other end of the spectrum there's the Yankees. They won 83 games, with WAR of 44.9; now a team of replacement players only wins 38.1 games. And that is an enormous problem: each team, and each player on each team, is being evaluated against "replacement level", but "replacement level" moves all over the place. It's not a fixed standard. If true replacement level really is the average of 47.3, for the sake of argument, then the Cubs players got cheated out of 10.4 WAR, and the Yankees got bonus WAR of 9.2.

I think, but don't know how to prove, that this problem is entirely, or almost entirely, due to the fielding measures, including the positional adjustment. I've used WAR, and it's cousin JAWS, myself from time to time. I'm not going to anymore. The system doesn't work.

1951WheatiesPremium

WAR has always been inherently flawed. It’s one tool in the player evaluation tool kit.

In many cases? It seems like the ‘advanced metrics’ are useful but produce odd outliers.

Have you measured the WAR against the Expected Wins for each franchise? That could potentially work better as both try to isolate variables and are mostly predictive.

One thing I consider is the ‘replacement player’ in WAR is average but that’s not always how it plays out in reality. First off, the replacement player for Wally Pipp was Lou Gehrig. Sometimes WAR value doesn’t know what’s coming next. 😂. But seriously, there is no such thing as the ‘average player’ because the skill cage from player to player. Let’s take a guy like Luke Voit, mashing 1B of NYY. He’s there for his bat. Others may come in for glove. Need and score will dictate, not WAR.

WAR makes the assumption that hitting is always most valuable skill; it is in ties and games you are losing but less so in games where you lead.

perkdog

Only the dead have seen the end of war. Plato

thisistheshow

A lot of good stuff on Sports Talk early this Wednesday morning.

I'm not usually here this early............or am I @doubledragon ?

keets

correct me if I'm wrong, but the WAR calculation is only based on what a specific player does in a chance appearance in a game, maybe for only an inning or two. how can that kind of statistic be extrapolated for an entire nine player Team for an entire game, let alone an entire season?? it's a POP TART stat that gets used because baseball freaks just love stats.

here's a weird thing to consider --- with the supposed WAR players playing even an entire game, what would the stat do to accommodate for someone replacing one of them?? what would that be, a WAR WAR?? for the sake of the statistic, would the WAR players have to play every inning of every game for an entire season? what about the pitchers?? it's not realistic to assume a statistic for an entire Team of replacements that would still use the top starting pitchers.

I think it's a bogus stat that helps distort how everyone looks at players only because fans like the OP who are hip-deep in trying to compare players have nothing else to work with. MLB scouts over the years have traveled thousands of miles, spent hundreds of hours and kept meticulous notes trying to evaluate players and find talent. I wonder if they use a stat like this??

POP TART stat, JMHO.

DrBuster

I had a glimmer of hope this was about War the band. One of the best shows I've been to - War played at Music Midtown in Atlanta in the 90s, was a great show among others.

JoeBanzai

We generally use these stats when comparing two very good to great players.

Using a number that lumps them in with every other player in the league has never made much sense to me.

Justacommeman

I find WAR is only used when it validates ones own side of a debate

m

bronco2078

Dead sports dont need stats anymore.

Tabe

WAR is a decent stat in the aggregate. It generally gets things right but...it creates all kinds of outliers and results that make no sense. You get stuff like Joe Morgan's WAR going down from 1975 to 1976 despite his numbers being far better and other results that just make no sense.

dallasactuary

Addressing a couple of the comments here, a "replacement player" is just a statistical construct, a measure of the level of play of someone that would get called up from the minors to replace an injured player. And yes, there are problems inherent in the definition of a "replacement player" on its own terms, and different stat nerds define it differently. But, for purposes of a stat like WAR those problems shouldn't matter. As long as every player is getting measured against the same yardstick, the stat still "works". If Player A has WAR of 6.0 and Player B has WAR of 1.5 against one definition of a replacement player, then those same players will have WAR of 5.5 and 1.0 against a slightly better "replacement player", and you could still use WAR to say that Player A beat Player B by 4.5 WAR.

The problem I pointed out is far more serious. The definition of the replacement player stays constant, but if Player A plays for the Reds his WAR might be 6.0, while if he plays for the Yankees it might be 8.0. At least that looks like it could have happened in 1975. In another year where the two players had seasons identical to 1975, Player A's WAR might change to 5.0 and Player B's might change to 3.0, since the goalposts can shift. A moving standard isn't a standard, and measuring players against a moving standard is useless. Say it again.

@1951WheatiesPremium said:
Have you measured the WAR against the Expected Wins for each franchise? That could potentially work better as both try to isolate variables and are mostly predictive.

As I was doing my analysis this very question popped into my head. And I suspect that the results would probably be closer to correct if I did it that way. But, that's still a problem. WAR purports to define how many wins a player was worth, not how many he was theoretically supposed to be worth. If WAR says a team should have won 100 games and they only won 90, that's an awfully big discrepancy to just shrug off. This, among countless other reasons, is why I trust Win Shares more than any other stat; the problem I've described here with WAR simply can't happen with Win Shares. Step one of the Win Shares process is to assign Win Shares to each team in direct proportion (that proportion is 3, if anyone cares) to how many games they won. Win 100 games, the team gets 300 Win Shares; win 50 games, get 150 Win Shares. With WAR, there is an underlying assumption that luck can be responsible for up to a 10 game swing in the standings. I'm open to Bill James perspective that luck counts for nothing over a long season, and I'm open to the idea that really lucky or unlucky teams might win a net game or two over the course of a season. But 10 game swings? That's not luck, that's something that WAR, and some other stats, are simply not measuring (or are measuring incorrectly).

bronco2078

Keep talkin you are just hammering nails in the lid of baseballs coffin

JoeBanzai

War
I despise
It means destruction to innocent lives
War means tears
To thousands of mothers' eyes
When their sons go out to fight
And lose their lives
War, huh, yeah
What is it good for?
Nothing

1948_Swell_Robinson

WAR messes up everything when they give equal weight to the defensive portion as they do the offensive. 95% of the plays for the outfield are of the routine variety and can be replaced by nearly any college outfielder. Infield, not quite as high as a percentage as 95%, but still quite high as minor league players replace that quite easily.

Also, for any play where a routine ground ball or fly ball is produced, then 95% of that out is a credit to the pitcher, not the fielder, because the commodity is finding the pitcher who can produce easily fielded balls, not the other way around.

WAR offensive measure is better than the defensive measure, but is still not even as good as something as simple as OPS+(with proper account for playing time).

IF you want to use WAR, then the more accurate way would be to assign the defensive portion and positional adjustment portion at only 25% of their values and then you get a little closer to reality.

bronco2078

most important stats ,

average time between pitches headed for a minute , average age of baseball fan (soon to hit 80)

average attendance (trending to zero_)

Darin

I'm going to try to add some dramatic effect to Broncos' posts.

most important stats,

average time between pitches headed for a minute, average age of baseball fan (soon to hit 80)

average attendance (trending to zero)

Darin

keep talkin you are just hammering nails in the lid of baseballs coffin.

Darin

Dead sports don't need stats anymore.

bronco2078

baseball fans may be rarer than rachel dratch fans

Brick

I saw a list of the top players and only recognized a couple of names.