rjazzguy
In the third-to-last paragraph of your sixth “section,” (assuming “sections are separated by asterisks) the correct grammar would be, “Those weren’t the conditions…”

Those that object to this correction: To hell with you; the dumbing down of America is due to such objections.
7:02 PM Aug 11th

docfordock
Dave:

The basic components of bWAR are RA/9 and IP (with adjustments for team fielding, park, opposition quality, etc.). The basic components of fWAR are FIP and IP, again with adjustments. Both formulas already take account of IP - it's a core part of both of them.

Your objection is that for the purpose of cross-era comparisons, the formula isn't properly taking into account the different contexts of performance. That's true in a sense. The WAR framework can tell you that player X in 2018 contributed approximately 9 wins as against a construct of a "replacement player" in 2018 and that player Y in 1966 contributed 9 wins as against a 1966 replacement level construct. It can't tell you tthat the respective replacement level constructs for the two years are equivalent or that the "wins" it is measuring in 1966 or 2018 are of equal worth. There is an incommensurability across time.

So far, so good. But your argument goes further to say that (in the case of Koufax and deGrom) that Koufax's 1966 performance is more valuable because "pitching WAR is giving modern starting pitchers all of the credit for the contexts that they’re pitching in, and none of the demerits for those contexts.":

In the modern game, a starting pitcher like Jacob deGrom can come in and throw 100 miles-an-hour, because there is no expectation that he will go nine innings. He can pitch, in essence, like a closer: go hard, as long as you can, and leave the last two-thirds of the game to the bullpen.

That wasn’t the conditions Koufax operated within . . .

Let's assume that's true. In the WAR framework, that means that deGrom will have a relative advantage in terms of RA/9 or FIP but the disadvantage in terms of IP. The 6 inning start giveth and it taketh away. To the extent there is a RA or FIP advantage, it is not an advantage unique to deGrom- it is an advantage for anyone pitching under today's conditions. I.e. deGrom is being compared against other pitchers that turn the dial to 11 on every pitch, while Koufax is being compared to other pitchers that (supposedly) paced themselves in an effort to get complete games.

The difference in conditions doesn't give deGrom in advantage over Koufax, unless deGrom has some special personal ability to take advantage of those conditions as compared to Koufax's ability to take advantage of his. But if true - and I don't know what the basis would be for reaching that conclusion - that seems to be working by design. A player's ability to take advantage of their context to deliver outperformance that helps his team win games is one of things the WAR framework is attempting to capture.
9:56 AM Jul 8th

OBS2.0
It occurs to me that the same disdain we have for pitchers "Wins" has to be scrupulously scrubbed from all thoughts and metrics about pitching.
12:17 AM Jul 7th

docfordock
[i]But who was more valuable in reality? Who won more actual games for their team:the pitcher who was making 38 starts each season and completing most of them, or the pitcher who was averaging 30 starts a year and completing fewer than half? Who won more games: not games against replacement peers, but actual baseball games?]/i]

I don't really understand the question as framed. Individual pitchers don't win baseball games, teams do. Sandy Koufax didn't win any games in reality; the Dodgers did. All one can do is try to measure the individual's contribution to the teams ability to win. The WAR framework in one method for answering that question but it necessarily does so by examining player performance in the context in which he played. You can't evaluate what deGrom's contribution would be to the 1966 Dodgers because deGrom didn't play for the Dodgers in 1966.

The concept of replacement is abstract but to make it more concrete, in 1966 Koufax had the 27-9 record, 1.73 ERA in 323 IP. The next year he was out of baseball. He was replaced. His place in the rotation was taken by Bill Singer 12-8 2.64 ERA in 204 IP and Jim Brewer 5-4 2.68 ERA in 100 IP (neither had significant IP the prior years). That's close to a point of ERA difference; or about 36 ER over 323 IP.

deGrom pitched in 2019 so one can't do the same comparison but one can note that the 5-6 mets starters in 2018 had ERAs in the high-5s. I think it is reasonable to conclude that there was no likely easily available replacement for deGrom that would be posting a 2.70 ERA or anything close. So it isn't crazy to conclude his performance had equivalent value to his team.

We can make these kinds of comparisons all we want, but what we can't do is move Koufax to the 2018 NL and move deGrom to the 1966 NL. We can't resolve the inherent incommensurability of the different contexts, either because it simply isn't possible or the tools to do it haven't been developed yet.

I think the more interesting questions are:
1) In the present day - are the prevailing usage patterns really optimal or would there be some benefit from trying to extend out certain starters even it means pacing effort or bucking the 3rd time through the order effect?
2) Looking at past eras - can one argue that the usage patterns were suboptimal back then? Did managers/GM underrate the potential of relief pitchers at the time and leave some low hanging performance fruit unpicked?
11:32 PM Jul 5th

shinsplint
Regarding Marichal, Carlton, Nola---Nola only looked so high up on bWAR because of the apparently lousy defense he had to overcome that year. His RA9def for 2018 is -.61, which is pretty low. Carlton's in 1980 was +.02. Marichal's in 1965 was -.02. So basically bWAR figures he was much better than his ERA+ indicated. If so, then Nola was much better than Carlton or Marichal on a per-inning basis in regards to the years in question, which makes up for his much fewer innings. As 110Phil says, that RA9def for Nola is questionable, but that's a different issue than suggesting that bWAR is context adjusting for innings pitched.

I decided to look at deGrom/Koufax/Gooden for the same years as Dave, but this time in regards to bWAR instead of fWAR. As shown below deGrom IS lower than Koufax or Gooden for bWAR. deGrom's bWAR is roughly
proportionately less than Koufax and Gooden when considering his fewer innings. But bWAR thinks deGRom is better than his ERA+ indicates, given his (again apparently) worse fielding. Just eyeballing it, if anything it looks like deGrom is being shortchanged here.


year/inning/ERA+/RA9def/bWAR/pitcher

2018 217.0 218 -.38 08.0 deGrom
1966 323.0 190 -.07 10.3 Koufax
1985 276.2 229 +.01 12.2 Gooden

Bottom line, I don't see any evidence that bWAR is somehow context-adjusting in such a way that a pitcher gets value easier in these days of fewer innings pitched. Could be that fWAR is doing that, though.

I do, however, see a greater truth that Dave alludes to here that is exhibited by the Pedro vs. Sandy comparison. 2 great pitchers, but one throws 100 more innings a year. Pedro is solidly better in terms of runs allowed per inning, but there must be some level of diminishing returns for pitching a great game. Maybe Sandy won some games 4-2 where Pedro won 4-1. A win is a win, so pitching slightly less excellently may be penalized too much relative to the benefit received for pitching more innings.



2:27 PM Jul 3rd

110phil
Dave:

But is that what WAR does? I was pretty sure it included raw playing time, not playing time relative to a player's peers. As I understand it, WAR has no idea how many innings other starters pitch, for either Koufax or deGrom.

I think WAR thinks deGrom was a much better pitcher *per inning*. Perhaps that's because of a higher run environment today, and/or Koufax pitching in Dodger Stadium?

Or I could be wrong about how WAR works. Perhaps someone else will chime in if so.
12:24 PM Jul 3rd

laferrierelouis
Nice piece M.Fleming. Koufax was running marathons while deGroom is running 1500 meters. And we try to compare them: I agree with you, there a is a problem here.
9:19 AM Jul 3rd

110phil
My understanding is that WAR measures how many more wins than a replacement player FOR THAT AMOUNT OF PLAYING TIME.

If DeGrom beat Koufax while pitching only 2/3 as many innings, it's because he was 50 percent better than Koufax (compared to replacement) in the innings he did pitch.

If that's the case, then what's the problem?

Yes, you'd think that deGrom would have a lower WAR if he was asked to pitch more innings, which is why they don't ask him to do that. And, maybe Koufax would have a lower WAR if he were able to pitch full-out in fewer innings like deGrom, which is why they didn't ask *him* to do that.

Or, it could be that they didn't ask either of them to do that because it wasn't normal for the times in which they pitched. But in that case, why come down on deGrom and not Koufax? It's just as likely that deGrom could pull a Koufax (if he has been asked to) than that Koufax could pull a deGrom (if he had been asked to). Your argument makes it sound like Koufax could have matched deGrom (pitching harder but less), but deGrom couldn't have matched Koufax.

----

(BTW, Aaron Nola's huge bWAR for 2018 in a middling number of innings is, I think, because he's given way too much credit for having a bad fielding team behind him. I wrote about that here: blog.philbirnbaum.com/2020/12/splitting-defensive-credit-between_29.html )
2:28 AM Jul 3rd

Gibbo1224
Great job Dave, this is something I have been going over in mind the last few years and just so happens I made comment to Dan Marks on Starting Pitchers ranked 21 thru 30. My comment was comparing Seaver to DeGrom and how you can win Cy Young awards with 10 and 11 wins. Seaver pitched for the Mets when they were a losing team but still won 20 or more games 5 times. The guys like Seaver control the outcome of the game by pitching 9 innings +/- while SP that pitch 6 innings are leaving the game in control of pitchers not as good as them. The most important thing here is to win the game. Your article really gives clarity here and much appreciated!!
9:52 AM Jul 2nd

MichaelPat
Here's another 9.3 WAR pitching season season (BRef war):
Rube Waddell 1905
27-10 1.48 ERA 181ERA+ 328 2/3 IP

Then, of course, we have Cy Young, with nine seasons of 9 WAR or more (four of them after 1900), maxing out at 13.9 in 1892 and 12.5 in 1901.
Obviously his value was far greater than anything accomplished by Koufax, Seaver or Clemens.
Koufax never pitched more than 335 innings in a season, a total Cy surpassed fourteen times. 14!!

Was it a 'mistake' to move away from using pitchers the way Young and Waddell were used?

No, the game changed.

If a clone of Waddell, Young, Koufax, Seaver, or Clemens were to appear today, do you really think there would be any possible chance they would be used the way they were when they pitched?

The game has changed, and the only fair way to evaluate pitchers is to compare them with their peers.

Yes, pitching more innings means you will contribute more wins (and more losses) to your team's total. So does that necessarily then mean that all the really great pitchers took to the mound before 1905?

In today's game, where a team will use thirty pitchers in a season (9 was the average in 1900, 15 in 1960, 19 in 1990) the expectations for how much any one pitcher should work are far different.

It used to be that guys who could throw 95 mph were few and far between.... Now nearly every team has twenty or more guys who can do that.

I don't have any problem with the idea that deGrom in 160 to 180 innings today could be the equal of Koufax throwing 300 in 1965 or Cty Young throwing 400 in 1894.



12:51 AM Jul 2nd

DaveFleming
Sorry...Elizabeth Warren?

Those numbers are Pedro's stats from 1997-2000, not 1999-2003. It is a four-year stretch (Pedro's best), but I called them the wrong years. The numbers are correct...just ignore the dates.
6:14 PM Jul 1st

elwarren
For your Pedro Martinez vs. Koufax comparison, '99-'03 represents a five year span, while '63-'66 represents a four year span. Did you mistype? Wanted to compare vs. Win Shares, then realized there may be an issue.
5:08 PM Jul 1st