shunned

Maddux was a better pitcher for a period as long as Koufax's entire career. But in your view they were equally effective at preventing runs, because Maddux's ability to pitch for another 11 seasons reduced his career rate to match Koufax? Seriously?

I was not really intending to go into that specific example as it does not match what I said well I guess but I do stand by the response in general, for cases where it does. If your peak was on a performance level equal to another person's entire career including pre-prime and decline then they performed better than you, in general, on average.

OK, good so far. But can I also point out that from age 25-36, another 12-season stretch, Maddux put up an ERA+ of around 170? That is, would it be OK to mention in a discussion of these pitchers that for a period as long as Koufax's entire career, Maddux was MUCH better? Or must I then be shunned as methodologically impure?

I don't think it's very useful. It's interesting, but I'm skeptical of what the point is. If we're talking about value or some handwavey "greatness", it's worth looking at various peak stretches. This is basically just taking the absolute worst peak of any length (which happens to be his best 12 year peak) for Koufax and comparing it to Maddux's best 12-year peak; it's fine as far as it goes, but it's probably not really a great argument for many purposes.

The point is that comparing one player's entire career to another player's peak is just not really that sensible, unless we believe a 12-year peak is a good measurement to judge things by.

I feel like this is getting a bit esoteric though. Traditionally, players are typically compared by either total career numbers or shorter defined peaks. Why not stick to that?

What do you mean by "standard" advantage? The general estimate is that a reliever's ERA would rise by about 1 run as a starter.

I looked at James and Tango on the subject. Tango came up with .60, James much less, but I forget specifically. What are you looking at?

All you can really do is rate Rivera against other relievers, and replacement level at his "position."

Right. But what is replacement level for a closer?

The bullpen is tricky b/c of leverage, and the ability of the manager to allocate the best pitchers to the high leverage IP.

jyjjy: Tango's estimate of reliever advantage is 1.00 (unless he's changed it recently).

Snapper: I agree it's tricky. A replacement reliever is probably about league-average ERA, give-or-take. For the closer you need to use chaining to set an even higher replacement level, but then he gets credit for the leverage of his innings.

I feel like this is getting a bit esoteric though. Traditionally, players are typically compared by either total career numbers or shorter defined peaks. Why not stick to that?

Because we're discussing starters vs. relievers, and so when people drool over Rivera's 200 ERA+ in 1200 IP, a performance everyone agrees is great, it absolutely must be analyzed just _how_ great that performance is, and whether the performance was so unique and off the charts that other pitchers weren't able to accomplish it. Clearly others were, and that is absolutely relevant, even if you don't understand the concept of cherrypicking and when that is flawed - and when it is not.

The Maddux/Koufax comparison is quite different, because it's explicitly not cherrypicked. You're comparing the same chunk of Maddux's career against Koufax's. Maddux then went to on to tack on a bunch of very effective work.

But Rivera's career cannot be viewed the same as Pedro's. He has been able to stay as good as he is because he only pitches 70 innings a year. If Pedro went balls out for eight games and never coasted his ERA+ would be a lot higher. If Rivera is used like a starting pitcher he would have a similar slow start and tail end to his career.

Seasons with 50 or more Games Finished and ERA+ of 200 or better:

Mariano Rivera, 10
Joe Nathan, 5
Billy Wagner, 4
Robb Nen, 3
John Wetteland, 3
five tied with 2
53 tied with 1

It's a fair point that there are currently 3 or 4 seasons like that every year, and it's been like that since the late '90s. But for quite a stretch, that meant Rivera every year plus 2-3 other guys, and for a few years, Rivera + Nathan + 1 or 2 other guys.

Hey. I'm a skeptic on the value of closers and their roles, myself. And it's also fair to point out that Rivera (like Nathan, Nen, and many others) was a washout as a starter, and likely wouldn't have had a good career as a starter. It's fair to point out that a lot of starters could be good relief pitchers, and that starters are generally much more valuable. It's fair to point out that earlier-model relievers like Face, Wilhelm, Abernathy, Lyle, Marshall, Gossage, and company had different roles and were darn good at them, too.

It's also fairly clear that Rivera as a closer has been a uniquely optimal match of ability to role.

That his entire career works as someone else's peak... Yeah, I'm going to say I think the opposite. If your whole career including pre-prime and decline is at the same performance level as someone else's "peak" then you were a "better" pitcher.

Not if your whole career also equals the same *quantity* as someone else's peak.

Ray is right about the quantity and situational issues that are inherent in evaluating any reliever, not just Rivera. He's wrong about other things, like the postseason, of course.

He has been able to stay as good as he is because he only pitches 70 innings a year

Since when do relievers have no decline phase?

Not if your whole career also equals the same *quantity* as someone else's peak.

Why? That was specifically what I meant. You played less but you were better on average. That is my whole point.

Snapper: I agree it's tricky. A replacement reliever is probably about league-average ERA, give-or-take. For the closer you need to use chaining to set an even higher replacement level, but then he gets credit for the leverage of his innings.

Replacement level for Rivera and other 9th inning "closers" is the next best reliever.

The tOPS+ of the American League in 2012 improved by 3 from the 8th inning to the 9th -- 86 to 83. Not much.

Put the 30 designated "closers" in a spaceship on April 1 and blast the spaceship to Jupiter for the season. Managers have to pick a "closer" and use him the same way. The impact is then the true value over replacement the closers generate. How many additional saves do we think would be blown? One per team is 30. I'll take the under.

How many additional saves do we think would be blown? One per team is 30. I'll take the under.

Do you mean blown by the closer, or the blown save stat which can be accumulated by guys in the 7th and 8th innings?

Because if it's only blown saves by the new anointed closer, then for "aces" you would have to compare them to the new "ace" that would be bumped up from #2 in a similar spaceship scenario involving the 30 opening day starters, and not to starting pitchers in general.

He has been able to stay as good as he is because he only pitches 70 innings a year.

Why hasn't any other reliever been able to do it then?

Blown by the closer in the 9th inning.

"If your peak was on a performance level equal to another person's entire career including pre-prime and decline then they performed better than you, in general, on average."

I must be misunderstanding something here. Koufax's numbers look so great precisely because he had no decline phase. He got hurt in his prime and then never pitched another game again. In order for the comparison to be fair to Maddux, it's only fair to cut off Maddux's post-age 30 seasons. Of course, it doesn't really matter as Maddux mops the floor with Koufax regardless. Maddux had a better ERA+ (132) than Koufax (131) although the former threw almost 2700 additional innings. If you just mirrored Koufax's career where his age-31 season looked like his age-30 season, his age-32 like his age-29, and so on so that his age-42 season mirrored his age-19 season, you would still have a pitcher that was inferior to Maddux on a rate basis (albeit only just) and still 350 innings behind Maddux. The difference would be someone with a career like Brian Wilson's (320 IP @ 129 ERA+).

Duplicate post. My apologies.

What this boils down to is: Rivera isn't in a class with the elite pitchers.

People know this, intuitively, even if they don't know that they know it. When making a list of the greatest pitchers ever people talk about WJohnson and Clemens and Maddux and Pedro, etc. Rivera's name never comes up.

I'd take over but the meaninglessness of the save stat of course makes the whole question a bit silly.

Replacement level for Rivera and other 9th inning "closers" is the next best reliever.

Then replacement level for "clean-up hitters" is the next best hitter in the lineup, and for "aces" is all of the #2 guys in rotations.

I must be misunderstanding something here. Koufax's numbers look so great precisely because he had no decline phase. He got hurt in his prime and then never pitched another game again. In order for the comparison to be fair to Maddux, it's only fair to cut off Maddux's post-age 30 seasons. Of course, it doesn't really matter as Maddux mops the floor with Koufax regardless. Maddux had a better ERA+ (132) than Koufax (131) although the former threw almost 2700 additional innings. If you just mirrored Koufax's career where his age-31 season looked like his age-30 season, his age-32 like his age-29, and so on so that his age-42 season mirrored his age-19 season, you would still have a pitcher that was inferior to Maddux on a rate basis (albeit only just) and still 350 innings behind Maddux. The difference would be someone with a career like Brian Wilson's (320 IP @ 129 ERA+).

First off why would you write all that in response to the sentence after I admit Maddux/Koufax is not an example that fits what I was getting at? Second, Koufax got hurt years before he retired, which he did at the end of a season, not all of a sudden and then he never pitched again.

I must have misread your #239 as you said: "If your whole career including pre-prime and decline is at the same performance level as someone else's "peak" then you were a "better" pitcher. More valuable? Probably not. Better? Yeah." I think I see what you were saying now.

Right, I should have worded that part better. I'm aware that Koufax was injured previously and battled arthritis during his last few years in the big leagues. My point was that Koufax abruptly quit due to an injury rather than take time off and attempt a comeback. I should have made it clear that I meant his retirement was abrupt and final in his prime. Koufax obviously faced a series of lingering injuries that made continued pitching impossible or dangerous. Regardless, he still retired before his decline phase, which was my point.

No you read that one right... but you quoted when I restated, which happened right after admitting those two pitchers are not actually an example of what I mean.

Koufax obviously faced a series of lingering injuries that made continued pitching impossible or dangerous. Regardless, he still retired before his decline phase

He did, but one of the interesting things to ponder about Koufax is to what degree the chronic pain and ghastly swelling he endured through his last few seasons inhibited his performance. Obviously it's unknowable, but it would be ludicrous to imagine that his physical agony was helping his performance, and, ya know, he went 97-27 with an ERA+ of 170-something from 1963-66. A healthy Koufax is kind of scary to contemplate.

I am fine with calling Rivera the greatest closer in history, and arguably the greatest reliever in history (I'd have to look again at Wilhelm), and a certain Hall of Famer (by the standards set by the voters).

That really should satisfy the pro-Rivera contingent in this thread.

Anything beyond that, which includes starting pitchers, and I absolutely want to stop and examine just what the claim being made is.

I am fine with calling Rivera the greatest closer in history, and arguably the greatest reliever in history (I'd have to look again at Wilhelm), and a certain Hall of Famer (by the standards set by the voters).

I don't disagree with any of that.

Edit: though I'm pretty sure I'd say Wilhelm was the greatest RP.

jyjjy: Tango's estimate of reliever advantage is 1.00 (unless he's changed it recently).

You are right that I was looking at an outdated number from him but here is the latest I can find on the subject from tango;

http://www.insidethebook.com/ee/index.php/site/comments/rule_of_17/

17% rather than the .6 runs just makes my case stronger.

17% rather than the .6 runs just makes my case stronger.

Do you really believe Rivera's RA/9 would only have increased by 17% if he started?

Because, if that's true, the Yankees made the worst player personnel decision since the Babe Ruth trade with him .

That really should satisfy the pro-Rivera contingent in this thread.

Best pitcher ever on a rate basis or bust yo. It simply is what all the math says, or our current understanding of maths at least, if you will.

Do you really believe Rivera's RA/9 would only have increased by 17% if he started?

I do not believe I nor you could ever know that so I'm sticking with actual statistical analysis of the subject. If you want to go off the reservation on this one feel free but recognize it is what you are doing.

Best pitcher ever on a rate basis or bust yo.

That would be the guy who faced one batter in his career and struck him out, which is probably Wade Boggs or someone.

I do not believe I nor you could ever know that so I'm sticking with actual statistical analysis of the subject. If you want to go off the reservation on this one feel free but recognize it is what you are doing.

I think I'm the one firmly on the reservation. The statistical analysis is biased b/c only RP who teams think have a good shot at being good get to start. The fact that the Yankees never considered it should be taking as evidence.

If Rivera could have put up a 2.50 ERA as a SP in his era, he would have been the best SP ever. Even Pedro gave up a 2.93 ERA. How likely is that?

Best pitcher ever on a rate basis or bust yo.

That would be the guy who faced one batter in his career and struck him out, which is probably Wade Boggs or someone.

Absolutely true, and if we raise the innings threshold a little, it is Kimbrel, and if we raise it further it is Rivera, and if we raise it higher it becomes Pedro, and then Lefty Grove, and then Walter Johnson, and then Cy Young.

17% rather than the .6 runs just makes my case stronger.

No, 17% makes it larger (.17 * 4.6 R/G = .8 runs). Read the thread at your link: Tango still has replacement level for relievers at 1.0 run better (lower) than for starters.

And if you believe Rivera could have posted a 2.60 ERA as a starter, well, I don't even know what to say.....

Best pitcher ever on a rate basis or bust yo. It simply is what all the math says, or our current understanding of maths at least, if you will.

No, best reliever ever on a rate basis (and perhaps career basis). We can all agree on that I hope.

Absolutely true, and if we raise the innings threshold a little, it is Kimbrel, and if we raise it further it is Rivera, and if we raise it higher it becomes Pedro, and then Lefty Grove, and then Walter Johnson, and then Cy Young.

Which only proves my point. There's no reason to set the threshold at exactly the point where Rivera is the beneficiary.

Sean could have gone with 1300 innings rather than 1000, which would have left Rivera out in the cold. No, wait, Rivera has 140 postseason innings - surely Sean includes those? Alas, no. Sean does not include innings pitched in exhibition games. So Rivera would still be out in the cold.

And if my complaint about lack of similar opportunities for others to pile up innings in the postseason is invalid, why doesn't Sean include postseason innings in the seasonal and career totals?

Surely we can agree that B-Ref's criteria for listing players is irrelevant. If we use that, we have to say that Jason Bay is the 7th greatest base stealer of all time on a rate basis. Who wants to sign up for that?

And honestly, I can't believe that even Mariano fans would have said (before this discussion) their minimum IP to be considered in a "greatest pitcher" category is just 1200 IP. I would say something like 2,000. And that's even before worrying about the reliever advantage (which is a huge additional complication).

I didn't bring him into the discussion for this reason, but Koufax (50) and Mo (53) are actually about the same in career rWAR. That sounds about right. 1200 IP @ 206 = 2300 IP @ 131. I think rWAR uses the same replacement level for starters and relievers -- is that right? -- in which case I'd bring Mo down to maybe 46 WAR. But they are in the same ballpark.

I think I'm the one firmly on the reservation. The statistical analysis is biased b/c only RP who teams think have a good shot at being good get to start. The fact that the Yankees never considered it should be taking as evidence.

If Rivera could have put up a 2.50 ERA as a SP in his era, he would have been the best SP ever. Even Pedro gave up a 2.93 ERA. How likely is that?

You clearly didn't take the Yankees paying big money for Soriano as proof of the worth of a closer and are often roundly critical of their choices(usually correctly) but all of a sudden management's decisions with regard to Rivera is proof of something? Sorry but hypocritical is the word here.

Again, I really really really do not particularly care if Rivera would have made a good starter. He was and is a pitcher. Anyone summarily dismissing a whole class of pitchers from the start when asking the question who was/is the best pitcher is... simply answering a different question. Innings add value surely and it would be idiotic to ignore that when figuring out who was the most valuable pitcher. In the end what I'm saying is most valuable and best are really not interchangeable. To reverse my previous example if a rookie next year hit a home run in every at bat then retired, best hitter ever or not? This is what the debate has turned into as I pointed out earlier, semantics, not statistics and I do believe I'm on the correct side from either angle. There is no math indicating Rivera's performance advantage should be dismissed because he was a reliever unless you simply throw out all relievers and I'm even more certain my interpretation of "best" is a much more common usage. Again we can divorce those from baseball. Band with 3 amazing albums vs one with 7 that while very good are notably inferior. I'll say the one that amazes me is better every time.

Which only proves my point. There's no reason to set the threshold at exactly the point where Rivera is the beneficiary.

"Exactly the point"? Rivera is number one from 160 innings to 1200 innings - that might be the largest spread of any of the names I listed. But yes, it is an arbitrary threshold, and the leader's accomplishment becomes more impressive at each higher level.

And if my complaint about lack of similar opportunities for others to pile up innings in the postseason is invalid, why doesn't Sean include postseason innings in the seasonal and career totals?

He doesn't want to taint the inferior regular season numbers! ha!

"Exactly the point"? Rivera is number one from 160 innings to 1200 innings - that might be the largest spread of any of the names I listed.

And what number is he from 1400 innings to 5000 innings?

And what number is he from 1400 innings to 5000 innings?

same place Pedro is for 3000 to 5000, or Lefty Grove is 4000 to 5000, or Walter Johnson is 6000 to 7000.

No, 17% makes it larger (.17 * 4.6 R/G = .8 runs). Read the thread at your link: Tango still has replacement level for relievers at 1.0 run better (lower) than for starters.

Just reread it, dunno what you are referring to. He says;

This becomes very useful for when we do research into comparing year-to-year changes in performance, and we see that pitchers move between roles. So, to be able to put everyone on the same scale, you can take a reliever’s RA9 and divide it by 0.83 (that is, 1 minus 17%), and that puts it onto the starter’s scale.

Seems pretty straightforward.

I probably need to check out til tomorrow but I'll leave with one point to put things in perspective. The gap between Rivera's ERA+ and the next player is coincidentally exactly the difference between that player and an average one(of course ignoring the preposterous postseason resume.) You need a better reason to dismiss such a thing than saying, "but he was a reliever!" Seriously. If you are really all about quantity over quality then well, meh, we will never agree about what is "best" because you ignore the best if there isn't enough of it to the point that a player can pitch until they are 43 at an established position for pitchers and it's not enough even for "best on a rate basis." I find your standards unreasonable and that is all there is to it.

You need a better reason to dismiss such a thing than saying, "but he was a reliever!"

The problem with your argument is that that is a damned good reason.

Seriously. If you are really all about quantity over quality

It's not quantity over quality but a mix of the two and not just that but the fact that relieving is so very obviously a hell of a lot easier.

Hell, even compared to other relievers Rivera has it easy: he starts a fresh inning with the bases empty the vast majority of the time. And mostly pitches in 1 inning bursts. And typically gets to work with a lead and thus margin for error. It's bizarre the trouble people have seeing this.

The math has shown you are wrong. You are just saying things you want to believe at this point. Your estimation of the relative difficulty of starting vs relieving is invented and your cut off line arbitrary. The latter is arbitrary by nature of course but that you set it to exclude relievers inherently means you are denying the question itself has meaning when applied to relievers rather than attempting to make any sort of qualitative comparison. You shouldn't even be involved in a discussion of the topic IMO.

jyjjy/293: You need to work a little harder at this. Tango says: "if you set the replacement level for a starting pitcher at say 130% of league average, then the replacement level for a relief pitcher is 83% of that, or 108% of league average." So if you have 4.5 R/G:
Starter RL = 1.3 * 4.5 = 5.85
Reliever RL = 1.08 * 4.5 = 4.86
5.85 - 4.86 = .99 runs. I hope my rounding that off to "1" is acceptable.

That was in the comments which I did not get to as the main text seemed pretty clear. In terms of ERA+, the stat we have been using, the baseline is the earned run average average of all pitchers, not RA/9. Seems silly to be figuring out replacement level for a different stat in this context. I don't feel like looking it up so let's just say on average a starter goes 6 innings with the other three for the pen. If it is 17% harder in general to be a starter as Tango says I believe that means to even things out you would compare a starters ERA+ to 90 and a relievers to 105. Rejiggering things so would bring Pedro up to 171and bring Rivera down to 196. Yes I know, the 17% may not strictly apply to ERA specifically and 6/3 is off some so... yeah, but the result isn't close.

The gap between Rivera's ERA+ and the next player is coincidentally exactly the difference between that player and an average one

You can't subtract points of ERA+ that way. Mariano's ERA is 48.5% of league average; Pedro's is 64.9%. The distance from Mo to Pedro is less than half the distance from Pedro to average. (Not that it isn't still impressive.)

WRT postseason stats giving some players more opportunities than others - you could correct for this pretty easily if you want to. Give players credit for only their regular season playing time, but include their postseason numbers when calculating their rate stats. Doesn't make a huge difference for most people - for instance, Pedro pitched about 3% of his career innings in the postseason. But for Rivera, who's pitched over 10% of his innings in the postseason (at an unbelievable level), it drops his ERA by 7%.

jyjjy: Look, if you are determined to compare relievers against starters in terms of total value (both quality and quantity), what matters is the baseline we are measuring these guys against. For Rivera, it's approximately 1.00 R/G lower (better) than for a starting pitcher. You should read up on this -- a lot of work has been done on it, which you don't appear to be familiar with.

The point of Tango's formula is not to predict what Rivera, or any other specific pitcher, would do if he changed roles. And you certainly can't use the 17% rule, derived from players who could perform in both roles, to project the starter performance of guys who were only relievers. Our assumption has to be that such relievers (Wagner, Rivera, Hoffman)were selected precisely because they were much better suited to relieving, and would likely have a much larger reliever/starter spread if forced to start. And honestly I doubt that the reliever advantage is proportional to runs allowed, which would suggest that lousy starters would reap huge dividends while good starters benefit far less (and also that great relievers would pay a very small price for starting). Maybe Tango has research to support that, but I doubt it.