WAR

If Keuchel has 0 for dWAR, I would not consider that a credible metric.

 

dWAR...huh...what is it good for?

 

I don't know all the details, but a few things I can contribute:

1) Things like "defense behind the pitcher" matter. I'd imagine the Twins have had a better D this season, so that wouldn't explain what you're seeing...but that's like an unknown underneath and unknown so I thought I'd mention.
2) You end up multiplying by innings pitched in the calc, so the 2 innings matter to some degree.
3) WAR is calculated differently by different sites, and Fangraphs has Keuchel at 1.9, Ervin Santana at 1.1 thus far in 2017

Were you looking at 2016? Fangraphs has Santana at 3.2 last year and Keuchel at 2.7...very close to your #'s.

 

I prefer fWAR for pitching and don't understand bWAR as well. bWAR favors ERA heavily. My guess is that it comes down to opponents (ERA fluccuates with opponent quality more than FIP) or leverage when runs were given up.

 

Pitching WAR is a notoriously sketchy stat, to the point where I literally pay no attention to it.

I like WAR for players because there are 4 elements (hitting,fielding,running,position) and it does a good job (not perfect) of quantifying value combining them so a players value beyond simple OPS shows up.

Pitchers pitch..that's all. There are peripherals, and luck stats projecting future results, but ultimately the simple pitching stats do a pretty good job of telling you how well a guy performed.

 

crazy how far off they are. Looks like they pull from baseball reference (or at least use the same source)

I'm going with sealclubber and saying pitching WAR must be garbage. Either that, or I trust fangraphs even more than I already did

 

I've never heard this before. There is more variance in pitching WAR between sites as people disagree on the importance of FIP vs. ERA, but I would say that this disagreement will be the same if just looking at simpler stats like FIP and ERA. I will say pitching WAR is less useful when comparing starting pitcher to starting pitcher as ERA/FIP do a great job at that. Leverage, opponents, and innings typically make a bigger difference in relievers.

Not knowing bWAR pitching (or ESPN) too well, I expect Santana v. Keuchel is one of the bigger differences for similar innings and similar ERAs using ERA based WAR.

 

ESPN uses bWAR. I have a strong preference for fWAR, but that is mostly due to it more closely matching my expectations, and I don't have an understanding of how they are calculated. I do think fWAR normalizes for team defense.

WAR is far from a perfect stat, but it is a single number that reasonably estimates a player's overall value. I'm sure in the near future there will be a more accurate stat that does a better job.

 

WAR

I've started using it for simplicity because everyone else does as a simple all-encompassing stat, but I fundamentally don't like it for a number of reasons:

1. Everyone calculates it differently, which makes it really subjective
2. It's not remotely intuitive
3. I don't think it actually measures "wins above replacement", because I think the actual wins above replacement a player contributes depends on many external factors.

 

Those WAR numbers don't include the game yesterday, so Keuchel's WAR will be higher than that once updated. Expect Dallas will get about another half win.

Pitching WAR preference depends on what you prefer. Actual results (ERA) or predictive stats (xFIP). I prefer the actual results when looking at WAR, because I'm usually not looking at pitching WAR for future performance. I'm looking at FIP, xFIP, & SIERA.

 

It's certainly not perfect, but it does a good job of providing quantitative value for a good fielding shortstop versus a butcher at first base who may have better offensive numbers.

When there is a significant gap in WAR it is a very important number to me. When 2 guys have a WAR in the same ballpark (1 point or so) I personally don't put too much stock into it, because there is a lot of room for error in the stat.

 

That is my main issue with it. The WAR number on one site doesn't just have some different calculations, it's an inherently different stat. Placing large value in FIP versus little to no value, they are basically night and day.

I can look at their peripherals and get a better indicator of how good a guy has been, and how good he should be moving forward.

 

It can definitely be confusing. I've had the same problems with it.

Having googled it many times, the consensus I saw is that fWAR is based more on peripherals and independent stats (like FIP, xFIP) and is more representative of dominance and future performance whereas bWAR is weighted more on actual run prevention (e.g. ERA) and past performance.

bWAR is going to favor a guy like Keuchel who maintains a low ERA and WHIP whereas fWAR is going to favor a guy like Chris Sale who has dominant peripherals (12.7K/9, 1.85 BB/9) and the best FIP in the league but an ERA an entire run higher than Dallas.

As for why Santana is higher than Keuchel in bWAR, not sure.

 

Is Chris Sale a better pitcher than Dallas Keuchel moving forward? Sale has significantly better peripherals but imho its a toss-up who'd win head to head.

 

On your issues,

1. I expect over time, the different WAR models will get closer. I find WAR much more objective than saying player X is better than Y because X is better at "random stat here".
2. I will say it is a complicated equation and not knowing the ins and outs don't make it wrong. Pitching WAR is a tougher to track than position player WAR as leverage, FIP, ERA, and a few different things go into it. For hitting, calculating runs isn't that tough and runs are typically worth about 0.1 WAR. Base running stats and defensive stats are black boxes to me, but generally follow with what I expect by watching the game.
3. WAR does a better job at taking in as many factors as possible than just relying on raw stats that don't take into account leverage, opponent, etc. WAR models are usually updated every year to get closer to the value a player actually provided.

For me the problems with fWAR (my favored WAR model) are the following:
1. Catcher defensive measures suck compared to Baseball Prospectus.
2. Reliever leverage seems to be under-rated. MLB teams pay a lot more for elite relievers than public models indicate. I expect MLB teams have WAR models that estimate performance with leverage better.
3. For pitchers that demonstrate an ability to suppress strong contact over time aren't accurately measured. This is about 5% of pitchers so I still find fWAR generally more accurate than ERA-based WARs.
4. Pitchers WAR does not add in offensive and defensive contributions, though these are usually within error of WAR model (+/- 0.5 WAR).
5. Replacement level values likely need to be updated as game is changing with athleticism. I suspect with increasing athleticism, position values for up the middle may need to be reduced as it is easier to find good hitters that can play those positions.
6. Defensive WAR is still very weak and needs a large sample size similar to ERA. Statcast data may help beef up defensive WAR numbers more quickly.

 

I agree with this - but I also feel that if there's a big gap in WAR, you can tell from the eyeball test or other stats too. And if WAR isn't that precise a measurement where you can't rely on it +/- 1 WAR or so, then I don't know how much value it really adds.

I also think measuring over a replacement-level player is an odd baseline, because the depth of the position matters. If there's 2 amazing shortstops and everyone else is crap, that puts a different premium on those two players than if there are 2 amazing ones, 20 pretty good ones, and then the rest are crap. It almost seems like "average MLB" player would be a better standard.

I've pointed to the Correa-was-5th-in-WAR stat last year a few times. But do I really think Correa was the 5th most valuable offensive player in baseball last year? Not really. (ignoring age, future potential, etc)

 

Agree with all of what you say - and I'm not suggesting WAR is inherently bad. Just wanted to give some reasons I personally have a hard time with it, especially when trying to debate a player's value, simply because I can't wrap my head around how it was calculated (like you said - that doesn't make it wrong/bad).

The above part I quoted is an example of an issue I have with it though. Unlike other more traditional stats, this one is a manipulation of someone's opinion - for example, adjusting replacement level values because the game is changing seems fairly arbitrary. Necessary, I agree - but how to do that is sort of a function of the opinions of the people doing it. Unless basic stats like AVG/OPS/ERA/etc, or even newer evaluations like ISO or bat-speed or whatever, the inputs really matter for WAR. The other ones just feel more "objective" to me, even if they don't measure things perfectly. WAR outputs seem so much more dependent on arbitrary inputs. You can't really manipulate SLG or AVG or ERA on the backend by changing what you input into it.

 

Going off this years raw numbers alone, yes. His K/BB numbers are outrageous. I'm not sure if it will stay at that level, but a player that is keeping the ball out of play at that rate would likely be better.

But numbers aside, having watched the two I would easily chose Keuchel if I had to pick a guy to anchor my rotation in October.

Positional value is probably the most important part of a WAR stat. Base running is a negligible factor, and offensive numbers are easy to compare. But even to this day there aren't people who realize how much value comes from playing a premium position, or playing defense at all

The Adam Everett's of the world have a chance to look good with the WAR stat. While the decent hitters but otherwise useless players are adequately punished for it. The players who do everything well rise to the top. As for the WAR difference over 1 being noticeable, that's not always the case. Often you can look at 2 guys and think they would value out pretty similarly, but the WAR number will make one look noticeably better.