This guy must be someone who just discovered War about a week ago.

Anyway, at the time of Dontrelle/Carpenter award, nobody was arguing that Dontrelle got snubbed, they argued Roger Clemens got snubbed. This writer is focusing too much on war though.

Carpenter 21-5, 2.83era, 7cg, 4 sh, 241 2/3 ip, 82runs 76er 1.055 whip, 150 era+, 2.90 fip. 5.8 bWar 6.3 fWar
Dontrelle 22-10 2.63 era, 7cg, 5sh, 236 1/3 ip, 79runs, 69er, 1.134 whip, 152 era+, 2.99fip. 7.3 bWar 6.5fWar If you can find a clear difference between the two that gives a snub award to Dontrelle good for you.
Roger 13-8, 1.87 era, 1cg, 0sh, 211 1/3 ip, 32 runs, 32er, 1.008 whip, 226 era+, 2.87 fip, 7.8bWar, 6.0fWar

Sudden Sam had 8.3 bWAR in 1970 and finished 3rd behind Jim Perry and Dave McNally, both of whom had 3.8. The also each had 24 wins to McDowell's 20 which is what voters cared about

Stieb had 7.6 bWAR in 1982 and finished 4th behind Vuvkovich(2.8), Palmer(4.8) and Quiz (3.3)

Anyway, at the time of Dontrelle/Carpenter award, nobody was arguing that Dontrelle got snubbed, they argued Roger Clemens got snubbed.

Poor Pedro Martinez: in 2005, he had 7.0 bWAR (3rd among pitchers), led the NL in WHIP, was 3rd in Ks, and 4th in ERA ... and didn't even receive a single Cy Vote, unlike Roy Oswalt, Andy Pettitte, Chad Cordero, and the three others mentioned above.

That was still the days of only 3 slots on the ballot. There were only 4 non-Carpenter, Willis, Clemens votes across the 32 ballots. Oswalt picked up 2 3rd-place votes because he got to 20 wins. Cordero presumably got one out of closer love. The vote for Pettitte is a bit weird -- maybe the two Hou voters each decided to toss a vote to a local and they decided to split the bounty.

Wow, the Nats have been around for nearly 20 years -- time flies.

Pedro also got a bit of short shrift in 2002. Zito got the AL Cy Young for having gotten 23 wins. He had a little better WAR but Pedro was miles ahead in ERA+ and, come on. He was Pedro.

Willis was a better player than Carpenter in 2005, but he created a lot of the difference with his hitting, which I always had the impression didn't count in CYA voting.

I am biased, but it seems clear that Clemens losing to Welch in 1990 was the most egregious error in CY voting. The whole thing hinged on exactly one stat: 27 wins. Even by the standards of the time, it was clear that Clemens was far better that year--his eye-popping 1.93 ERA was more than a run better than Welch's. In addition, Clemens had better H/9, BB/9, K/9, and HR/9 (just 7 HR in 228 IP!). But those 27 wins were simply too much to ignore. Weirdly, voters somehow found Clemens to be more valuable than Welch, placing him 3rd in the MVP race (with 3 first place votes) and Welch 9th.

Advanced stats of today further illustrate the massive gap between the two:

Clemens: 10.4 bWAR. 8.2 fWAR
Welch: 2.9 bWAR, 1.8 fWAR

Steib should get some type of lifetime snub achievement award, having led the league in WAR in 1982, 1983, and 1984 with zero Cy Youngs to show for it. This included getting zero votes in 1983 despite doubling Lamar Hoyt's WAR! His 1983 even got snubbed in this list of snubs.

I've looked at Stieb's 1982-85 closely.

1982 - The voters couldn't look past "17-14 on a 78-win team" but who else? It was a weird year in the AL with no dominant starting pitching performances. Stieb is the obvious winner on bulk alone. The second-best pitcher that year was Floyd Bannister, who had a losing record.

The NL race was also messed up. Carlton's 23 WINZ! won it, though both Steve Rogers and Mario Soto were clearly superior.

1983/84 - LaMarr Hoyt was a joke, but Quisenberry's 1983 and Hernandez's 1984 are both in my top ten relief seasons of all time, and those would've been my choices.

1985 - Saberhagen was a worthy winner. Stieb won the ERA title, but 16 of the 89 runs Stieb allowed that year were unearned. In RA/9 both Saberhagen and Blyleven have stronger cases than Stieb.

If Stieb has two Cy's is he already a hofer?

Willis was a better player than Carpenter in 2005, but he created a lot of the difference with his hitting, which I always had the impression didn't count in CYA voting.

FWIW, the 7.3-5.8 WAR difference listed above between Willis and Carpenter is pitching only; Willis had 8.3 bWAR if you count hitting.

If Stieb has two Cy's is he already a hofer?

This game is also called "if Johan Santana had four more years of decline phase, is he a hall of famer?"

My first thought, re: an answer, was Roy Halladay. He had almost as many IP as Stieb, but was a better pitcher (136 ERA+ to 122). Then there's Saberhagen - slightly shorter career, slightly better ERA+. Sabes isn't in, despite two CYYs. So, where is the line between Hallady and Saberhagen? It's probably somewhere around "was widely thought of as the best pitcher in baseball," which would keep Stieb-with-hardware out, but might put Santana-with-more-innings in.

Then there's Saberhagen - slightly shorter career, slightly better ERA+. Sabes isn't in, despite two CYYs. So, where is the line between Hallady and Saberhagen? It's probably somewhere around "was widely thought of as the best pitcher in baseball," which would keep Stieb-with-hardware out, but might put Santana-with-more-innings in.

I think Sabes is hurt by being considered a bit of a mercenary and by the bizzare even-year odd-year pattern he had--starting from 1985, his ERA+ were: 145,102, 138,106,180, 118, 135, 101. This gave him the patina of "inconsistency"

the patina of "inconsistency"

No patina about it, the man was inconsistent. You throw in few 3WAR years surrounded by 8-9WAR years, it seemed odd then and seems odd now. When he was rolling, he was the best in the business.

This game is also called "if Johan Santana had four more years of decline phase, is he a hall of famer?"

You are very correct, if Santana has four more years that probably puts him around 180 wins 2700 innings pitched and 58-60 war. Pretty much where Stieb ended up.

FWIW, the 7.3-5.8 WAR difference listed above between Willis and Carpenter is pitching only; Willis had 8.3 bWAR if you count hitting.

I always make this point when these discussions come up, but bWAR is often a terrible metric for determining awards when it comes to pitchers.

Willis and Carpenter had similar IP, similar ERA+, similar UER allowed. So why does Dontrelle have a 1.5 advantage (26%) in pitching WAR?

Well, it looks like Willis gets credited with 0.43 runs per 9 IP because he pitched in front of a bad defense, while Carpenter gets dinged 0.35 runs per 9 IP because he pitched in front of a good defense. That's a huge shift, equivalent to ~20 runs over the course of the season.

Did Carpenter actually get 0.78 runs more defensive support per 9 IP relative to Willis? That seems unlikely given that Willis' ERA outperformed his FIP by 0.36 while Carpenter only outperformed his by 0.07.

Fangraphs, which does not make this kind of defensive adjustment, has them basically equal, with Willis ahead 6.5 to 6.3. I suspect the difference is due to Willis' offense.

It's probably somewhere around "was widely thought of as the best pitcher in baseball," which would keep Stieb-with-hardware out, but might put Santana-with-more-innings in.

Would Stieb-with-hardware not have been thought of as the best pitcher in baseball, at least for a couple of years? If he'd won Cy Young awards in '83 and '84, there wouldn't have been much competition at that time; the early '80s were pretty barren once Carlton started slipping.

I always make this point when these discussions come up, but bWAR is often a terrible metric for determining awards when it comes to pitchers.

Willis and Carpenter had similar IP, similar ERA+, similar UER allowed. So why does Dontrelle have a 1.5 advantage (26%) in pitching WAR?

Well, it looks like Willis gets credited with 0.43 runs per 9 IP because he pitched in front of a bad defense, while Carpenter gets dinged 0.35 runs per 9 IP because he pitched in front of a good defense. That's a huge shift, equivalent to ~20 runs over the course of the season.

Did Carpenter actually get 0.78 runs more defensive support per 9 IP relative to Willis? That seems unlikely given that Willis' ERA outperformed his FIP by 0.36 while Carpenter only outperformed his by 0.07.

Fangraphs, which does not make this kind of defensive adjustment, has them basically equal, with Willis ahead 6.5 to 6.3. I suspect the difference is due to Willis' offense.

I'm usually a fan of bWar over fWar for pitchers, but you do bring up good points, most notably the general reason why I prefer bWar is it's result based instead of theory based, but then you pointed to it also is basing it upon perceived defense which negates some of it's result base advantage. If you have a good defense and you pitch to it, you shouldn't be penalized for getting the result you were trying for. Why go five pitches in an at bat, when you can get a ground out in three pitches?


Ultimately, fully acknowledging I'm a homer on this, I can not see a real reason to give it to Willis ahead of Carpenter, more innings pitched with a better rate result gives it to Carpenter. The extra win and extra shutout is nice and all, but Carpenter ultimately had the better result over the course of the season.

Well, it looks like Willis gets credited with 0.43 runs per 9 IP because he pitched in front of a bad defense, while Carpenter gets dinged 0.35 runs per 9 IP because he pitched in front of a good defense. That's a huge shift, equivalent to ~20 runs over the course of the season.

Did Carpenter actually get 0.78 runs more defensive support per 9 IP relative to Willis? That seems unlikely given that Willis' ERA outperformed his FIP by 0.36 while Carpenter only outperformed his by 0.07.

Carpenter allowed a BABIP of .282 that year vs. .288 for Willis. WAR probably is overstating the difference between the defensive support that each man got, but Willis outperforming his FIP probably had more to do with how he pitched with guys on base. He allowed an OPS of .684 with the bases empty, .581 with runners on, and .628 with RISP. Carpenter was at his best with the bases empty, giving up a .585 OPS with no one on, .699 with runners on base, and .725 with RISP.

I'm usually a fan of bWar over fWar for pitchers, but you do bring up good points, most notably the general reason why I prefer bWar is it's result based instead of theory based, but then you pointed to it also is basing it upon perceived defense which negates some of it's result base advantage.

Yes, fWAR is based on FIP (or xFIP, I don't remember) and bWAR rewards or penalizes players for overall team defense even when there is no solid evidence they benefited from it. If I had to choose a metric I'd choose RA9-WAR from Fangraphs, which is WAR just using the pitcher's actual runs allowed and not making any adjustments for FIP or defense. By that metric, Carpenter had a 7.2-7.0 lead over Willis in 2005.

With respect to bWAR, I wouldn't have a problem with a defensive adjustment if it was sensibly done. The way it's done right now, assuming that every pitcher on the team benefitted equally from the team defense. At this point we are calculating defensive runs saved by fielders on a play-by-play basis -- so we should be able to deduct those run saved from whichever pitcher was in the game at the time.

Carpenter allowed a BABIP of .282 that year vs. .288 for Willis. WAR probably is overstating the difference between the defensive support that each man got, but Willis outperforming his FIP probably had more to do with how he pitched with guys on base. He allowed an OPS of .684 with the bases empty, .581 with runners on, and .628 with RISP. Carpenter was at his best with the bases empty, giving up a .585 OPS with no one on, .699 with runners on base, and .725 with RISP.

A good point, but how much of that OPS difference was due to BABIP/defense? Carpenter had a .270 BABIP with bases empty, .303 with men on base, .304 with RISP. Willis was .310 with bases empty, .252 with men on base, .291 with RISP.

I think they were roughly equal in value that year. Of course, Clemens (9.1 RA9-WAR) probably should have won the CYA over both of them.

Fangraphs, which does not make this kind of defensive adjustment, has them basically equal,

I wouldn't phrase it this way. Fangraphs makes an even more extreme defensive adjustment -- assuming that the entirety of the difference between a pitcher's BABIP and league BABIP is the result of his defense (and luck). FG assigns essentially all of an individual's ERA-FIP difference to the defense. (Note, what we probably want here is RA9 - RA9fip but FIP is on a ERA scale so I'll stick with ERA-FIP below since we can easily get that.)

This means the level of defensive support received (as measured by FG) varies wildly from pitcher to pitcher within the team. Justin Steele (119 IP) of the Cubs had an ERA-FIP difference of just 0.02; Keegan Thompson (115 IP) of 0.64. It may seem unlikely that Carpenter received 0.78 R/9 more defensive support than Willis but that seems much more likely than Keegan Thompson receiving 0.62 more runs defensive support than Steele while playing (mostly) in front of the same defenders and in the same ballparks. (Interestingly, by ERA and FIP, Steele is well ahead of Thompson; by RA9, the gap is just 0.22 ... Thompson's RA9 was also lower than his FIP (by 0.17) while Steele's was 0.81 higher.)

By fielding runs at b-r, the 2005 Cards were #2 in the league at 79; the Marlins were last at -73. That's a gap of nearly 1 run per game on defense. That may not be accurate (i.e. issues with TZ/DRS/UZR, etc.) but it makes the Carpenter-Willis difference seem "likely." Obviously there is within-team variation so some Marlins pitchers got worse support than others -- but fg adjusts Willis's numbers as if he was receiving good defensive support and adjusts Carpenter's numbers as if he was receiving good (but not that good) defensive support. In comparing these two WAR measures you are given two options -- assume Carpenter received 0.78 more runs of defensive support than Willis (based on differences in team defensive performance) or assume Willis received 0.71 more runs of defensive support than the average Marlins pitcher (based on his FIP-ERA vs team FIP - team ERA) AND Carpenter 0.53 runs worse than the average Cardinals pitcher.

If the first option is unrealistic, what term do we assign to the second? Which is more likely: (A) The Cards' excellent fielders played excellent defense behind Carpenter while the Marlins' lousy defenders played lousy defense behind Willis or (B) The Cards' excellent fielders played just a bit above-average defense behind Carpenter (so therefore even better defense behind others) while the Marlin's lousy defenders played a bit aboveaverage defense behind Willis (and really, really badly behind others)? The first assumes the expected thing happened twice; the second assumes the very un-expected thing happened twice.

The b-r assumption is of course a simplification (less so than the FG one) but trying to tease out individual differences in defensive support would be like adjusting individual park factors for the specific game time and weather conditions of each game they started. Trust me, if 2/3 of one pitcher's starts at Wrigley have the wind blowing out while somebody else gets 2/3 with the wind blowing in, they did not pitch in the same "park." (Plus PF is averaged over 3 seasons anyway.)

Statistical nerd aside: Essentially any difference can be split into a "within" and "between" component (the issue is whether it's meaningful/useful to do). For example the difference between your house value (or income or weekly grocery spending) and the national average can be broken down into the difference between your house value and the average house value in your neighborhood plus the difference between your neighborhood average and the national average. If we define Z as the individual value, R as the regional average and N as the national average, it's simply:

(I - N) = (I - R) + (R - N)

We can then break down your house's value into the value of your house relative to the neighborhood and the value of the neighborhood. This is how you start to get at questions like "what is the impact of school quality on house prices" (which goes into the R-N component) or what is the value of adding an extra bedroom (which goes into the I-R component). Note, in may applications, N is a baseline value that might as well be zero ... or when comparing say two pitchers in the same league in the same year then the Ns cancel. If you wanted to include a component for, say, changes in N over time, replace N with N_t above with t indexing time and add (N_t - N_0) to the equaiton. When these componenents are (roughly) independent then we can easily break the total variance down into within and between componenets.

In our case this gets us to: a pitcher's total defensive support (vs. a league average of zero so N=0) equals that pitcher's defensive support relative to team average support plus that team average support relative to league average. FG assumes a pitcher's individual performance-independent environment is (I - N) and therefore subtracts all of that from their individual overall performance. B-R assums each pitcher's performance-independent environment is (R - N) which is roughly constant within team and subtracts that from individual performance, leaving (I - R).

Is the pitcher responsible for zero (FG) or I - R (B-R) or somewhere in between? Any sensible person will say "somewhere in between" but how do we get at that estimate? An average of the two individual values is an easy way; regression towards some population-level value would be another -- in a practical sense, that ends up being a weighted average. In equations, let PR be "pitcher responsibility for BIP outcomes":

PR = c*(I - R) where c is a constant between 0 and 1

FG assumes c=0; BR assumes c=1; an average assumes c=0.5; a "regression" would come up with some other estimate for c. I have no idea what we'd regress to so the last really isn't an option. If the true value of c <0.25, we'd pick FG; if it's >0.75, go with BR; if it's in the middle, the average is best.

Using DRS, Carpenter's (R - N) was 79 runs per 162 games; for Willis it was -73 ... so give or take, call that +0.45/9 vs -0.45/9 (a gap of 0.9/9 of course). Willis's FIP-ERA was 0.35 meaning his (I - R) was around 0.8 for our purposes. Carpenter's FIP - ERA was 0.07 so his (I - R) was around -0.38.

FG: Through whatever combination of defense and luck, Carpenter pitched in a BIP environment that lowered his ERA by 0.07, we will remove this from his overall performance. Willis's BIP environment lowered his ERA by about 0.35, we will remove this.

BR: Carpenter pitched in a BIP environment that lowered his ERA by 0.07 (not explicitly part of WAR) and we have estimated that 0.45 of that was his defensive environment, leaving him resposible for 0.38 below-average performance. Willis's BIP environment was about 0.35 while his defensive environment was about -0.45 defensive environment of +0.45 level, we will remove this, leaving him responsible for -0.38 in poor BIP performance. Willis pitched in a defensive environment of -0.45 level, we will remove this, leaving him responsible for +0.80 in outstanding BIP performance.

I understand the skepticism that Willis could have out-pitched his defense by that much ... but his results were that much better than his team context. So I also understand the skepticism that was all actual defensive over-performance (vs. other Marlins pitchers) and luck.

I read the whole article, which makes me a little proud and a lot sad. What's a less-positive term for "word salad"? (Salad is good and doesn't deserve this comparison.)

It may seem unlikely that Carpenter received 0.78 R/9 more defensive support than Willis but that seems much more likely than Keegan Thompson receiving 0.62 more runs defensive support than Steele while playing (mostly) in front of the same defenders and in the same ballparks.

I see no reason to think that defensive support would always be evenly distributed across a pitching staff. Offensive support certainly is not.

Consider another example: Aaron Nola had a 3.01 FIP and a 2.37 ERA in 2018. bWAR also credits him with 0.41 for pitching in front of a bad defense -- basically saying that in front of a neutral defense he would have had a 1.96 ERA. This is a guy who had never had a FIP or an ERA below 3 in any other season before 2022. Meanwhile Nick Pivetta and Vince Velazquez both had ERAs about a run above their FIPs for that same Phillies team. Maybe they got less defensive support and Nola got more.

I don't know which is more likely, that Nola was as good as peak Jake degrom for 33 games, or that the Phillies defense put together 33 games of above average performance/luck for him. It's a question we should have the data to answer, so it seems bad to simply assume one or the other. That's my point.

I read the whole article, which makes me a little proud and a lot sad. What's a less-positive term for "word salad"? (Salad is good and doesn't deserve this comparison.)

I read the whole article, it felt more like a guy just reciting numbers as if it was a class assignment to me. I'm not sure the writer understands the word snub to be honest. Basically his entire premise was "these are guys who might have deserved more votes for the cy than they got based upon my poor understanding of a new stat that I discovered called war, while ignoring what was actually going on during the voting process."

And as pointed out in this thread already, he didn't even mention Pedro Martinez.

Hmm. Off the top of my head, Clemens in 1990 is probably the very worst, followed by pretty much every great starter who lost out in any Cy Young season posted by a reliever (with the possible exception of Marshall's 1974 season).