there would be a lot less confusion if the pitcher wins stat had been renamed SPWNL (Stopped Pitching With Non-overcome Lead)...


. . . and on a completely different note: do any of the newer defense-independent stats include balls-in-play that have an out/not-out result that is independent of the defense (such as a line drive that hits 25 feet up on a fence)?

I'll be the contrarian and say that: - 25-3 means more to me than 1.78;

- 19-2 always impresses me;

- 24-4 makes me think of what could have been if not for drugs;

- 31-6 is difficult for me grasp;

- 27-10 for a 59 win team is damned impressive;

- 20-3 somehow always takes me back to 9/11 (and for that matter, 24-4, particularly that year's WS reminds me of my grandfather dying of lung cancer in a hospital);

- 16-1 reminds me of coming home from school and watching an improbable playoff run;

- 23-4 and 18-6 were amongst the most impressive exhibitions in baseball I will ever see; and,

- 20-9 and bye-bye is a lovely way to leave the game.

(Or said differently, I realize wins are not the most indicative stat out there, but they mean something to me. 16-8 will never look the same to me, after this year, you know....).

If Pitcher A has a 3.50 DIPS and a Pitcher B has a 3.75 DIPS, that does not mean Pitcher A was more valuable than Pitcher B. It means that Pitcher A is likely to be more valuable than Pitcher B in similar conditions going forward.

Voros McCracken showed that using strikeouts, walks, and homers (Defense Independent Pitching Statistics, or DIPS) to evaluate pitchers without their fielders does a better job of evaluating pitcher performance than ERA.

I think he means "projecting" rather than "evaluating."

I think Doc's problems were more an issue of arm problems than drug problems.

If Pitcher A has a 3.50 DIPS and a Pitcher B has a 3.75 DIPS, that does not mean Pitcher A was more valuable than Pitcher B. It means that Pitcher A is likely to be more valuable than Pitcher B in similar conditions going forward.

Exactly.

I saw that The Huffington Post has added Wayne Winston to their sports section. They certainly have a type.

- 31-6 is difficult for me grasp;

How about 59-12?

Man, what an awful article title.

For example, if Pitcher A gave up two doubles in the same inning, thus scoring a run, while Pitcher B gave up 2 doubles in separate innings, scoring no runs, and they did everything else exactly the same, then Pitcher B was more valuable than Pitcher A (and from the batter's side of things, the double with the guy on second was more valuable than the first double or the two doubles that lead to no runs). But projecting forward, you can expect them to be about the same.

Value and skill are not the same thing. DIPS is more of a skill stat, useful for evaluating players to acquire. ERA is more of a value stat, useful for Cy Young and Hall of Fame discussions.

18-3 means the luckiest pitcher in history.

Wins are correlated with... team wins.

No way!

Well, at its heart DIPS was created to adjust for the value of defense. So it's possible that someone with a better DIPS ERA outperformed someone else with a lower actual ERA but higher DIPS ERA. However, DIPS doesn't just measure quality of defense, but also a large amount of luck and even some small amount of pitcher skill. In general, I think adjusted ERA (or adjusted RA) is a better measure of in-season value, and DIPS is of more predictive value.

If Pitcher A has a 3.50 DIPS and a Pitcher B has a 3.75 DIPS, that does not mean Pitcher A was more valuable than Pitcher B.

You keep asserting this. I would say that if we had no more information, Pitcher A was also more valuable. But that's silly, we always have more information in real life.

You keep asserting this. I would say that if we had no more information, Pitcher A was also more valuable.

Yes, but if Pitcher A had a 4.00 ERA and Pitcher B had a 3.00 ERA, the DIPS would not make Pitcher A more valuable. Just more likely to be better going forward, if they pitched in similar circumstances.

I would say that if we had no more information, Pitcher A was also more valuable. But that's silly, we always have more information in real life.

If people would actually USE the additional information, we wouldn't be having this discussion.

-- MWE

Starting pitchers should be evaluated by games in which they left ahead or behind, which strips out everything that happens after they throw their last pitch. That takes a significant amount of luck out of the equation and evaluates in-game performance of "pitching to the scoreboard." No fancy calculations, just look at the game log.

By A-B instead of W-L:
Lincecum 19-7 (includes one game in which he left ahead before the fifth inning)
Carpenter 17-7
Wainwright 19-10
Vazquez 20-10
Haren 14-12

Would anyone think it was nuts that Vasquez received a Cy Young vote if the five games he left ahead in the score had all been converted into "pitcher wins?" Haren, OTOH, doesn't fare so well in this metric.

AL A-B records:
Greinke 19-9
Hernandez 18-13
Verlander 16-12
Sabathia 19-10
Halladay 17-11

So it turns out the pitcher with the best FIP in each league also has the best A-B% compared to the other Cy contenders, which is sadly not reflected in W-L% because of all the variables that happen after the starter leaves the game.

maybe that's better than W-L but that still leaves in run support which is a huge problem with W-L.

Not to mention that a game in which the pitcher leaves after 4 innings in a 6-6 tie counts the same as someone pulled after 9 in a 0-0 tie.

Not to mention that a game in which the pitcher leaves after 4 innings in a 6-6 tie counts the same as someone pulled after 9 in a 0-0 tie.

And does it take into account leaving with the bases loaded up 2 runs with nobody out?

I might find this A-B stat more useful if it included some context and was averaged out to the usual ERA scale (1.xx is phenomenal, 2.xx is extremely good, 3.xx is very good, 4.xx in the average range, 5.xx and above not very good to sucks balls).

Yes, but if Pitcher A had a 4.00 ERA and Pitcher B had a 3.00 ERA, the DIPS would not make Pitcher A more valuable. Just more likely to be better going forward, if they pitched in similar circumstances.

I don't think this is true at all. If the DIPS or FIP or whatever you use illustrates that those extra singles and doubles fell in because Pitcher A was surrounded by an inferior defense (and I'm not saying those stats do a perfect job of showing that sort of thing, but if they did), then Pitcher A would still be more valuable, because DIPS/FIP/whatever shows that he he did better than Pitcher B did at the things pitchers can control. It's not the pitcher's fault that an inferior defense pushes his ERA up, any more than it's to the pitcher's credit that he happens to get sufficient run support to allow him pile up a whole bunch of "wins."

re #20, but your post assumes that "If the DIPS or FIP or whatever you use illustrates that those extra singles and doubles fell in because Pitcher A was surrounded by an inferior defense" even though you qualify it somewhat. I think the post that you are disagreeing with does NOT believe that DIPS/FIP illustrate those things.

No mention of 18-1?

Amazing year, even if he was a vulture.

I think he means "projecting" rather than "evaluating."

And what was DIPS projecting? Future ERA. If ERA isn't the appropriate measure of pitcher quality, what is the purpose of predicting it?

It is a massive statistical assumption to go from "I am going to use X to predict Y" to "the predicted value from my model is the 'true' Y and everything else in the observed Y is measurement error." By my memory, Voros never made any such claim in his DIPS work nor should he have. And neither should current proponents of FIP/DIPS although this appears to be what fangraphs do in calculating WAR for pitchers and what Law did in his CYA analysis.

If people would actually USE the additional information, we wouldn't be having this discussion.

Seriously? Is there anybody who only uses FIP?

The last few days, I've seen a few people railing against FIP. I haven't seen anybody claim that it's the be-all, end-all, or using it as a primary metric to rank Cy candidates. I think the pendulum has swung pretty far in the backlash direction.

The only people I've noticed ignoring other information are the ones setting up silly FIP strawmen.

Yes, but if Pitcher A had a 4.00 ERA and Pitcher B had a 3.00 ERA, the DIPS would not make Pitcher A more valuable. Just more likely to be better going forward, if they pitched in similar circumstances.

Well, first of all you're tilting the example by suggesting a .25 difference in FIP and 1.00 in ERA.

If pitcher A has a 4.25 ERA and pitcher B has a 4.00 ERA, I don't know which one is more valuable. More likely, it's pitcher B, but it's kinda silly to guess without looking at the rest of the information.

I don't see this magic line between ERA and FIP where ERA is on one side where we're allowed to use it to measure value, and FIP is on the other side.

I haven't seen anybody claim that it's the be-all, end-all, or using it as a primary metric to rank Cy candidates.

I have. Maybe not here, but it's definitely out there.

OK, I'm listening to the Keith Law audio now, and I take back most of what I said in #24.

I don't think this is true at all. If the DIPS or FIP or whatever you use illustrates that those extra singles and doubles fell in because Pitcher A was surrounded by an inferior defense (and I'm not saying those stats do a perfect job of showing that sort of thing, but if they did), then Pitcher A would still be more valuable, because DIPS/FIP/whatever shows that he he did better than Pitcher B did at the things pitchers can control. It's not the pitcher's fault that an inferior defense pushes his ERA up, any more than it's to the pitcher's credit that he happens to get sufficient run support to allow him pile up a whole bunch of "wins."

But DIPS doesn't show that. DIPS, despite its name, does not indicate how a pitcher would have done with a neutral defense, it shows how he would have done with neutral luck and defense. And the luck is the big one.

DIPS, despite its name, does not indicate how a pitcher would have done with a neutral defense, it shows how he would have done with neutral luck and defense.

and that the pitcher has no impact on the fielders (well, unless he throws knuckleballs...)

-- MWE

3) How does it fit with our intuition?

Would you intuitively put Vasquez or Haren over either Wainright or Carpenter?

-- MWE

- 24-4 makes me think of what could have been if not for drugs;

I think Doc's problems were more an issue of arm problems than drug problems.

I assumed the 24-4 was in reference to Clemens and "what could have been" was a Hall of Fame legacy!

And does it take into account leaving with the bases loaded up 2 runs with nobody out?

Runners left on base is the main thing that's not accounted for by A-B, but it's still more performance-related than W-L in assessing what the starter actually did. The bequeathed runners that end up scoring still count against the pitcher's runs allowed, but their A-B record would determined solely by the score at the time of exit. It's not as accurate as WPA, obviously, but it's a straightforward approximation requiring no advanced math, since that apparently hurts the heads of the general populace.

Also I'd use RA instead of ERA. Seems unfair that pitchers playing in front of plus-range defense not only get the benefit of hits turning into outs, but hits turning into errors that changes what would be earned runs into unearned. It boggles my mind that Felix Hernandez finished second in the Cy and Roy Halladay finished fifth thanks to the archaic scorekeeping of W-L and ERA. Halladay pitched 1/3 of an inning more and allowed one more run, and their WHIPs and tRAs were virtually identical. By A-B and RA, the closeness of their performances is obvious: Felix 18-13, 3.05; Doc 17-11, 3.09. And Greinke's superlative 19-9, 2.51 would be that much more striking. Again, no sophisticated thinking required, just a modification of scorekeeping.

there would be a lot less confusion if the pitcher wins stat had been renamed SPWNL (Stopped Pitching With Non-overcome Lead)...

Yeah, because I don't have a hard enough time remembering SNWLP. The introduction of SPWNL would totally reduce confusion. ;)

The purpose of DIPS initially was simple, and no it was not intended as a predictive stat (though obviously it's ramifications in that regard were interesting). As the name of the stat suggests, the idea was to represent a pitcher's performance using only those stats his fielders could not affect. That doesn't mean that things like hits and runs and earned runs and wins were all about fielding, but since fielding can affect those things, they cannot be used in DIPS by definition. It's not DEPS (Defense Eliminated Picthing Stats), the idea is independence. Using hits or runs or double plays or whatever can not be used because they are not independent of defense like walks, home runs (almost always) and strikeouts (almost always) are.

What was interesting about it was what a huge amount what we think of as pitching is encapsulated by just walks, strikeouts and home runs. Further refinements then allowed us to try and predict defense dependent stats like hits per balls in play, using defense independent metrics.

That was the whole idea.

I see no problem with using it to argue who was the better pitcher, though it's valid to not do so too. The argument is that if you go to 10 Cubs games this year and the Cubs go 9-1 and hit .330/.420/.610 in those games, there's no reason to give you a Most Valuable Fan award or hand out the Ronnie Woo-Woo Award based on that. The argument is separating out bare-ass luck and other things completely beyond the pitcher's control is certainly a valid approach. You don't really have to use DIPS to do that, but to argue that a run scored is a run scored regardless isn't necessarily the only way to approach it.

FIP shows you ONE component of pitching, just like OBP and SLG each show you ONE component of hitting.

Treat it like that.

the idea was to represent a pitcher's performance using only those stats his fielders could not affect. That doesn't mean that things like hits and runs and earned runs and wins were all about fielding, but since fielding can affect those things, they cannot be used in DIPS by definition. It's not DEPS (Defense Eliminated Picthing Stats), the idea is independence. Using hits or runs or double plays or whatever can not be used because they are not independent of defense like walks, home runs (almost always) and strikeouts (almost always) are.

Oh, who are you to say what was and wasn't intended with DIPS?

18-3 means the luckiest pitcher in history.

Dice K is the only pitcher to ever go 18-3. There have been four 18-4, but only one 18-3.

So you're saying he's Lucky Dice?

I'll go.

Some pitchers, through their pitching style or reputation, can cause the umpire to call a bigger strike zone. That carries over to the other pitcher and can reduce run support. That is definitely what happened when the Diamondbacks were shut out in four consecutive Randy Johnson starts. The strike zone was so big that even Jose Jiminez could pitch a no hitter. That's what happened with Roger Clemens in 2005 - 2006. I've never seen Dice-K get a big strike zone.

Some pitchers, through their pitching style or reputation, can cause the umpire to call a bigger strike zone. That carries over to the other pitcher and can reduce run support.

This is an assertion that would seem to be ripe for study. Makes sense on face, definitely, but I would not be at all surprised to find this effect is non-existent/very minimal/not repeatable or any combination of the three.

The purpose of DIPS initially was simple, and no it was not intended as a predictive stat (though obviously it's ramifications in that regard were interesting).

In the original article, wasn't one of the big examples of DIPS that DIPS ERA in year X predicted the year X+1 ERA better than the year X ERA?

FIP shows you ONE component of pitching, just like OBP and SLG each show you ONE component of hitting.

Treat it like that.

If it had been treated like that, we wouldn't be having this discussion.

-- MWE

For example, if Pitcher A gave up two doubles in the same inning, thus scoring a run, while Pitcher B gave up 2 doubles in separate innings, scoring no runs, and they did everything else exactly the same, then Pitcher B was more valuable than Pitcher A (and from the batter's side of things, the double with the guy on second was more valuable than the first double or the two doubles that lead to no runs). But projecting forward, you can expect them to be about the same.

Value and skill are not the same thing. DIPS is more of a skill stat, useful for evaluating players to acquire. ERA is more of a value stat, useful for Cy Young and Hall of Fame discussions.

A quibble I have with this is how value is defined and what is being awarded. If Pitcher A gave up two doubles in the 4th inning (scoring one run), but no other baserunners over 7 innings, while Pitcher B gave up two singles in the first, a double in the second followed by a single in which the runner was thrown out at home, two more singles and a walk in the third, a double in the 4th, a double and a walk in the 5th, nothing in the 6th, and two more singles in the 7th, but gave up no runs, well then who provided more value? It has to be Pitcher B. But who pitched better? I'd say Pitcher A did.

So what are you awarding for Cy Young? The pitcher who provided the most value or the pitcher who pitched better? I see that both are legitimate answers and I don't think that you can wave away the FIP argument by saying that it only is good for predicting. It does measure how well you controlled the things that you can control*, and therefore is a better measure of how well you pitched. If you want to give the award to the pitcher who pitched the best this year, FIP is a reasonable measure to use. If you want to give the award to the pitcher who was the most valuable, well, then FIP is not suited for your purpose.

*Personally, at least for this thread, I'm not interested in whether FIP shows what it purports to show. Only, given that it does, whether it should be used for current year's value.

Greg, this is the exact same argument we have every year for MVP. The seeming majority are dismissive of WPA, I think it's only real use is exactly for MVP discussions. You're looking at value and performance and asking which do you vote for, and that's the central question when deciding what stats to use, isn't it?