Just curious, how do you have access to Tendu?

They asked me if I would like access to review the software and usefulness of the data -

me being a reputable researcher and analyst and all... ;-)

I checked with them before publishing this.

Wow. The "mound" as the one thing associated with Shea that might have been causing him problems at home? Funny, he seemed to pitch OK on that mound prior to 2003, when he was wearing that odd-looking uniform and wearing a (metaphoric) black hat.

No, you managed to sidestep it nicely, Chris, but Glavine's still an old-fashioned guy in a Questec world. The explanation for his problems is something like:

one part lost velocity;
two parts horrible defense;
one part actually NOT getting calls he used to because of a change in umpiring, due partially to Questec;
one part thinking he wasn't getting the calls and letting it bother him.

The last two were a bigger problem at Shea than on the road.

What I believe might be quite revealing would be a study of the outcomes of ABs against Glavine broken down by the count on the pitch of decison, compared to prior years. My hypothesis would be you'd find he was doing almost as well in 2003 on each count as he used to, but that he was facing a lot more unfavorable counts and so was getting into trouble.

Very interesting stuff, Chris. One other thing worth mentioning is just how badly his stats vs. the Braves skewed his overall numbers last year:

Glavine vs. Atlanta - 10.35 ERA (23 ER in 20 IP)
Glavine vs. everyone else - 3.80 ERA (69 ER in 160.3 IP)

So overall against every team against ATL he was pretty decent; he pitched 1/9th of his innings against Atlanta, but allowed 1/4 of his runs against them.

Of course, any pitcher is going to look better if you take out his outings against his most dominant opponent, but it has to be unusual that a single team accounts for a full 0.72 increase in seasonal ERA. And in this case, the team in question was one whose players knew his pitching style better than any other.

I'm not sure that's the right BABIP comparison. If we're trying to explain Glavine's decline, it's not a question of how his defense did relative to other Mets starters, but how his defense did in 2003 vs. 2002 and earlier.

Using Prospectus' defensive efficiency numbers, the 2002 Braves had a BABIP of 270, compared to the 2003 Mets BABIP of 294. (they have Glavine's hit deltas at -1 and 1 respectively, so he received nothing extra special from his defense either year) I don't know how many runs an extra 24 points of BABIP would be expected to translate into, but obviously it explains some of Glavine's "decline."

Or maybe I take that back. Good ol' MGL puts a run value of about .8 as the difference between an out and the average hit (HRs excluded). [note, I'm far from convinced that's right, but it's a place to start] By my count Glavine gave up 620 BIP in 2003. That .024 difference in BABIP comes out to about 15 hits in 2003 that would have been outs in 2002, which would be about 12 runs. That would drop his ERA from 4.52 to 3.93 and change his ERA+ from 94 to 108. Of course if I could just find myself a site with 2002 and 2003 DIPS ERAs, this would have been easier. :-)

That extra BABIP may have in part caused some of Glavine's other problems. If he couldn't rely on his defense (or if he thought the problems were his fault instead of the defense, though that doesn't sound like Glavine), he may well have felt the need to nibble more and stay away from the inside part of the plate even more than usual, possibly leading to fewer Ks and more BBs.

As one of those who've predicted doom for Glavine for some time now, I'm certainly accepting of the notion that this decline is real. But the improved Mets defense should be a big help to Glavine. Add in that his poor performance was largely due just to the Braves and his second half performance was much better (including much lower HR and BB rates), and an above-average year seems quite achievable to me. I'm not going to believe this guy is finally toast until he's been burnt to a crisp.

Now I will be stunned if he ever posts an ERA+ in the 120-130 range again, so I don't see how the Mets could ever get close to their money's worth out of him.

David,
interesting; I hadn't really considered the "should" factor of HBIP based on his K rate. I think collapse/rebound is about 50-50.

Colin,
I edited that out. In a post on STAT_LG, (some of which is included in this article), I made that observation. People whine about bias without understand the difference - the difference is, the Mets *do* control who Glavine pitches against. And skipping the Braves is a viable alternative. So it isn't really unreasonable to manage Glavine's starts around the Braves - almost every other pitcher would fare as well or better, so it doesn't really hurt you to move him. The goal is to optimize team wins, and *if* that means skipping Glavine vs. Atlanta, then so be it.

Sam M,
you don't like Glavine much, huh? He wasn't particularly victimized by his defense - that's pretty well borne out in his team average BABIP.

Glavine *still* threw the same percentage of pitches where he always does - they were just hit better.

As for previous trips to Shea - in 2001 and 2002 (because that's what ESPN has), Glavine had a decent ERA, but his walk rate was up, as was his HR rate (similar to what David Smyth posted). Plus he was facing the Mets hitters in those situations.

Maybe Questec squeezed him, but it isn't evident.

I looked at his data by count (you can do this at ESPN, and he had some problems:
By Count 2003 2002 2001
Count 0-0 16.86 12.22 15.61
Count 0-1 7.33 9.32 7.49
Count 0-2 6.01 6.42 5.96
Count 1-1 9.97 10.71 8.88
Count 1-2 10.41 12.22 13.71
Count 1-0 10.26 9.57 11.42
Count 2-0 4.40 2.90 4.82
Count 2-1 9.09 8.82 9.01
Count 2-2 14.66 14.99 13.32
Count 3-2 11.00 12.85 9.77

Those numbers are % of PARPLs.

He simply wasn't fooling anyone on his 1-2 or 0-1 pitches. Hitters simply didn't swing at them - his K rate in 03 was 32.3%, 02 was worse 30.8% and 01 was better 37.0%.

And worse, in bad counts, when the balls were put in play, they were hit a lot harder.

By Count 2003 2002 2001
Count 2-0 1.089 0.596 0.835
Count 2-1 0.984 0.895 0.742

Now this can be (and probably is) a function of lost velocity. It can also be a function of missing one's spot - leaving the pitch in the strike zone. The inability to come down and in really punches this up.

Sam M, you don't like Glavine much, huh? He wasn't particularly victimized by his defense - that's pretty well borne out in his team average BABIP.

Actually, I don't dislike Glavine, and think he has a chance for a solid rebound.

To say he wasn't "particularly victimized by his defense" is right, I think, in the sense that he wasn't hurt to a greater extent than other Mets' pitchers. But he was hurt by it relative to his prior performance. So if we're trying to explain why he was worse than Leiter, defense tells us nothing. But if we're trying to explain why he was worse than Glavine 2002, it tells us a lot. And that is why an improved defense might help him -- AND his fellow hurlers.

As for Questec, do you doubt that he thought he was getting squeezed by it? His public comments during the season sure led me to think it was getting to him, just as the poor defense was. If he was pitching behind a greater percentage of the time, that would cause a number of problems -- the data on ESPN show how much less effective he was behind in the count (nothing unusual there, of course); he had to take something off to get pitches in (perhaps explaining some part of the loss of velocity).

Walt beat me to it, but I also don't think that comparing Glavine to other Met pitchers is particularly relevant.

Question, Chris: do you specifically take foul outs out of BABIP? If you do what I usally do (hits minus home runs, outs minus strikeouts, etc.), then they'd be in there. Of course, I'm lazy.

Big Train,
imagine the strike zone. The location of pitches where plate appearances ended (a hit, a walk or K) was in the low and away area of teh strike zone. This was true, in 2002, 66% of the time. In 2003, his opponents PAs ended on a pitch low and away 67% of the time.

In 2002, his opponents PAs ended on pitches low and insdie 11.5% and only 7% in 2003. This (to me) indicates he was unsuccessful coming down and in.

I hope that helps.

In 2002, his opponents PAs ended on pitches low and insdie 11.5% and only 7% in 2003. This (to me) indicates he was unsuccessful coming down and in.

Chris, I get it a bit better now. But doesn't it depend on how the plate appearance ended? Ending with a 450 foot blast, after all, is a lot different from ending it with a fly ball to Roger Cedeno.

Beat . . . Wait for it . . . Beat . . . wait for it . . . and -- swing:

OK, maybe those aren't all that different . . . .


But a fly ball to Cameron -- now that's a better outcome! ;-)

Well, here I go again Chris (thanks for the Tendu answer, BTW - I wondereed if you worked for a team or for Tendu, as is is very, very expensive to access their data on your own)...

The proper way to analyze Glavin'e peforemnce in 2003, as compared to his previous performances, without all this silly speculation, although the "silly speculation" sometimes leads to more granular analysis, is:

Clearly as a pitcher gets older, and after a certain age, on the average, he will get worse. Unfortunately, there is no clear correlation between age and performance with pitchers as there is with batters. Probably the biggest forebearer of decline with age is average speed of fastball. If Glavine, like Maddux, has had a 1-2 mph decline in the speed of his fastball, as I have no doubt that he has (he used to be 87-89, with an occasional 90 or 91 - I think he is 86-88 now, but Chris can probably verify this from the Tendu data), then has HAS to be quite a bit worse! And yes, I would guess that as his fastball speed decreases, hw would be less and less likely to want to/b eable to come inside. IIRC, he used to come inside with the 4-seamer at 89-90 quite a bit to keep players "honest" with his outside 2-seamer and changeup. That is going to be more difficult now, as an inside 86-88 fastball is very vulnerable. Anyway, if he hasd already started a precipitous decline based at the very least on the decrease in his fastball speed, there is no reason not to expect that decline to continue. Pitching aging/experience patterns are very much unlike batting patterns. Pitchers bascially (again, as always, on the average) do not perform at peak levels when they first come up to the majors, regardless of their age. After several years, when they gain the requisite experience, knowledge of the batters, new pitches, etc., they then perforem at peak levels, again, regardless of their age, for many years, until eventualkly age and/or injury causes them to decline in performance. You can usually notice this, as the first thing to go is usually a pitcher's fastball speed (e.g., Maddux, Glavine, RJ). Of course, as we know, pitcher perforemcne is difficult to project from year to year. So the caveat to all of that is that pitcher talent level can fluctuate wildly all during those major league years, causing a lot of noise in that general, average pattern I described. That f;uctuation, again, much moreso for pitchers than for batters, is caused by small and large injuries, learning new pitches, mental and emotional maturity, learning new "aproaches" to pitching, different pitching coaches, metnors, etc. And many other unkown things. Some pitchers, for reasons which no one seems to know, suddenly become good or bad and it is NOT jusy random fluctuation. Anyway...

As far as Glavine and his defense, the only way to "handle" that is to simply take the "true" defense of the Mets fielders and "say" that is the amount that Glavine was hurt by his defense. It has nothing to do with his actual BABIP (that includes the defense of course, but it also inlcudes Glavine's "talent" and luck) and it has nothing to do with Glavine's BABIP as comapred to the Mets other pitchers. Bascially Sam M. and Walt were right ton the money with this. You adjust Glavine's stats for his defense in the same way that you adjust a player's home stats for their home park - you try and erstimate the true value of that defense (or park) relative to the average defense (or park) and then you apply the adjustment factor. As with parks, or any other sample data, a bad defensive team's true defense is not as bad its sample defense (same for a good defensive team), so you Don't want to adjust Glavine's BABIP by the Mets sample DER. If you did, it wouldn't be that bad, but technically you want to regress that DER first. Even more technically, since a team's DER is actually made up of 6 or 8 independent defensive entities, you want to regress each player's defensive stats separately and then add them all up, rather than regress the team's sample defense (DER) as a whole. SO the various ways to do the defense adjustment to Glavin'es BABIP (Or somponent stats), from least to most rigorous, would be:

1) Use DIPS (substitute league average BABIP for his sample BABIP)
2) Use the Mets' DER to adjust his BABIP, as Walt did (I think).
3) Use each Mets' player's last 3 years' regressed UZR's (or DRA's or ZR's, or whatever), prorate them for the number of defensive opps in 2003 with Glavine on the mound (or for all games in 2003), and add the weighted values up. That is the "true" defense behind Glavine, like a true multi-year regressed park factor. Now you can adjust Glvine's BABIP, or whatever you want to adjust, using this "number" (basically, true team UZR).

Here is the team defense UZR adjustments for the Mets in 2003:

IF -25 per 162 games
OF -10 per 162 games

That comes out to only around .22 runs per 9 innings in adjustment for a pitcher. Of course, the more BIP he has, the higher that is, and the more he is a GB pitcher, the higher that is, as the Met's IF was worse than their OF (it was also hard to compute what the "true" defensive value of the Mets' OF was last year, as lots of players played multiple positions).

As far as Glavine's home/road splits, to infer something about the "mound's" effect on Glavine (but not all other Mets' pitchers as a whole) or anything else about Shea, is absolutely insane! I thought I heard all kinds of "explanations" for extreme "splits"! I'll leave that one at that.

Finally, other than normal, random fluctuations around a players's (pitcher's) stats in any given sample period (1 year, 5 months, 1.2 years, whatever), what about Questec. Well, first of all, since Questec was in 10 or 12 parks out of 30, if a pitcher pitches in a Questec park, and a Questec park had a different strike zone (hence, a source of "adjustment" for a pitcher's and batter's stats), it can't be all that large of an adjustment (as opposed to if a player played in the only Questec park in the league).

First, you would have to compute a "Questec factor" if there is indeed one (and I don't know whether there is - it is on my "to do" list), by looking at stats at Questec prks vesus non-Questec parks (and adjusting for the home pitchers and batters and the parks themselves). Then, if you found one, you could apply that to Galvine's stats. As far as whether pitchers "like" Glavine, who rely significantly on that outside corner strike, are particularly affected by Questec, is another story, and seems like a real possibility to me (and I don't say that very often about things affecting particular players or types of players uniquely). How you would go about "adjusting" for that, I have no idea, other than to say that since he will be pitching in the same park next year, whatever real effect Questec had on him last year, it will have the same effect this year, so from a projection standpiont, there is no need to adjust (except for the fact that you would be combining Questec-affected stats from 2003 with noin-Questec affected stats from 2002 and 2001 to compute Glavine's 2004 projection, which could be problematic.

Anyway, when the smoke clears, I would think that Glavine's adjusted numbers for 2003 were not all that bad, and not all that different from the previous 2 or 3 years, although they probably reflect soem decline due to age and velocity problems. I would expect his numbers in 2004 to reflect his last 3 year's composite numbers with an adjustment for "decline" as evidenced by his velocity decrease and the potential unique problem for him of pitching in a Questec home park. Certainly an improved Met defense is going to raise his expected sample stats, not inclduing the fact that he is now one year older than last year.

Oh, and finally, finally, computing an ERC (compknetn ERA) also strips away some of the "luck" in a pticher's ERA, so that shoudl be done as well. IOW, a DIPS ERA or a regular ERC after adjusting for defense and the like (park) should give a better snapshot of how Glavine actualy piched last year, than his ERA or even ERA+.

IMO, the easiest way to calculate DIPS ERA is to take his FIP (which, in Glavine's case, was 1.68) and add 3.20. This yields a defense-neutral ERA of 4.88. This is based on Tangotiger's work a couple of years ago -- it almost always produces a DIPS-like number within 0.10 of the DIPS computation.

I've put a list of FIP and DER for all NL pitchers in the accompanying link.

Crack, out of a 77 line analysis, that's what you comment on? Is that the only sentecne you understood? If I had said, "all baseball players are communists," in the middle of my analysis, it would have made no difference. Whether that statement is right or wrong, I stand by everything else I said in those 77 lines (although some of it is speculation and opinion on my part. Is there a term for picking out one throwaway comment in a 77 line analysis and making a big deal about it as if that negates the entire analysis or somehow makes it any less (or more) valid? What I meant by "his numbers" was his "composite numbers," such as adjusted ERC, after all the "adjusting" is done. I don't know why I am bothering explaining that as I doubt you understand what I am talking about. And even if I was wrong, as I said, it means absolutely nothing to the rest of the analysis. In fact, I was neither dfeensing or criticicizing Glavin'es pitching perforemnce. I even pointed out the fact that there is evidence that age has taken a toll on him (the decreased velocity) such that he may not be nearly the pitcher he once was, in terms of true talent (which is not necessarily the same as his 2003 performance, although that certainly "informs" his true talent). So what exactly is your point? That that one sentence which inlcudes the words "my guess," BTW, is technically not accurate? Well, let's see. Even if we assume that I meant "every one of his component stats," when I said "his numbers," which I didn't - but let's assume that I did - then we have the problem of interpreting what is meant by "not that bad." I'll grant you that if something is terrible by some objective standard, that you cannot accurately say it is not that bad. Let's look at Glavine's K rate per PA, as per inning is just a way to approximate a per PA rate, which is the correct way to look at K rate, and compare it to all pitchers who threw to at least 100 batters (arbitrary minimum). I'll also compare it to all pitchers who threw to at least 300 batters, so we can compare him to mostly starters. Finally, I compareed to him to all pitchers who threw to at least one batter.

You say that he "has one of the worst K rates in baseball." Of course, that is an ambiguous standard as well ("one of the worst"), plus are you referring to ALL pitchers? At least we can see exactly where he stands among other pitchers in terms of K rate, and leave it at that (without using ambiguous words, which I hate to do anyway, unless I have to or it doesn't really matter, which in this case, is an example of the latter - I was not trying to make any particular point, I was just amkign a casual comment that had virtually nothing to do with my "thesis" in my post). To be true to exaclty what I was referring to, I am using all pitchers' park adjusted K rates. I adjust pitchers' home K stats according to the 10 year regressed K park factors of their home park. As I said, inthe comment in question, I was certainly referring to Glavine's park adjusted "numbers." I said that several times in the post and it should be obvious.

After doing the park adjustments, Glavine had 81.5 K's in 790 PA's. There were 35 pitchers who threw to at least 100 batters who had a worse K rate and 423 who had a better K rate. He is indeed "one of the 36 worst" among these pitchers.

Among pitchers who threw to at least 300 batters, there were 14 worse and 200 better. The worse ones were: Graves, Cook, Jimenez, Stark, Reuter, Sele, Mulholland, Westbrook, Cornejo, Lima, Mays, Bell, and Valdes. Here he is "one of the 15 worst."

Among ALL pitchers, he was 92nd out of 650 pitchers in K rate per PA...

Nice article Chris.

The only thing is what Sam and Walt have pointed out already. Before joing the Mets Tom Glavine's BABIP in 16 major league seasons was 278. While its certainly possible that Glavine's stuff is worse (as his k/9 indicates), i think is reasonable to say that defense had something to do with that BABIP going up.

Yes, Atlanta's regressed total UZR defense (all fielders' regressed 3-year UZR combined) in 2002 was +19. In 2001 it waa +27, and in 2000, +18 (all per 162 games). So there is a pretty big swing between the ATL defense and the Met defense. By virtue of that alone, we could expect a pitcher with an average BIP rate to have his ERA jump almost .4 runs just by switching teams in 2003...

MGL, Tom Glavine pitches differently than other pitchers - he has this goofy positioning wrt the rubber. He stands all the way off of it to the third base side with just the tip of his foot in contact. His extreme positioning would make the mound very different for him. Everything about the way he pitches from the tip of the rubber makes the mound different. *Plus* every mound in every park is different. There is nothing "insane" about it. But I appreciate the thought out critique.

I said the BABIP that Glavine had indicated he was no more lucky/unlucky compared to the other pitchers. Trachsel, Leiter and Seo all managed better ERAs than Glavine in front of the same defense. You can see that some pitchers got much worse BABIP numbers. Had Glavine's BABIP numbers come in at .335 (or something), that is a specific blame for his ERA.

Comparing Glavine's numbers to his teammates tells you a pitcher can pitch in front of the Mets defenders and post an ERA+ of 111. Since Glavine wasn't bashed around by his defense, and he had the same baseline as Trax/Leiter/Seo, something *else* must be the problem.

Even given the BABIP difference Walt calculates (which I agree with), that's still an increase of a full run on his ERA (2.97 to Walt's 3.96). (And MGL's subsequent estimations of 0.4 jump due to defense)

What we find is Glavine is to blame for his failures - is it all velocity - I doubt it. A few years ago, Glavine struggled and he came back throwing harder - closer to 92-93 for a season, throwing hard and inside more. Of course, I am old, so that could have been 1998.

There's no getting around it - Glavine's ERA jumped a lot - even accounting for his defense, which is also *clearly* not the cause for the majority of his problems - in fact, according to MGL and Walt, had Glavine pitched in front of the Braves-equivalent defense, Glavine would have posted an ERA of ~3.4 (ceteris parabis w/ 2002).

I think all Met fans would have been satisfied with that performance.

So back to the question, what else caused another full run of ERA on Glavine? Pitch location? Velocity? (Which can be explained by the mound...)

Yeah, but Chris -- do you have any affirmative reason to pin it on the mound? I mean, it seems like your argument for proposing the mound is that you've eliminated the other factors, so it's gotta be the mound. I certainly don't think it's insane, but I think it's awfully speculative, given that he hadn't pitched poorly at Shea in the past (not tremendously so, anyway).

If you were Rick Peterson, what would you tell Glavine? Would you go to Shea with him and a groundskeeper, and work off different mound configurations, trying to find a comfort level?

Sam M,
of course it is speculative - what else could it be -

The mound makes sense. That Huckabay article really stuck with me. I think it makes sense - it sure as heck isn't "insane".

Yes, I think Glavine may have started the season feeling squeezed. He is a pro. He'll adjust.

Yes, I think Rick Peterson will go over tape after tape after tape with Glavine. I think they will consider the mound.

I also think at his age, touch is tougher to have every time out. as a touch pitcher, that means 6 IP 3 H 0 R one day and 6 IP 8H, 5 R another.

When people hate Statheads, it's because they don't consider the human factor - the small things that make the numbers become the numbers - the things that cause fluctuations in ERA. (See my Mike Hampton predictions for 2003).

Sometimes, watching the game shows you things that explain why the numbers are what they are.

This is getting tiresome! Chris, you do good work, but you apparently don't recognoze or understand your limitations.

I'll address this post to Sam M., because I don't think that Chris is either able or willing to understand it.

When we find a large home/road split in a player's stats, particularly a one-year sample, but it applies to large samples as well, it is a collosal waste of time to try and "figure out" whether there is any reason for it (this is where I lose Chris who goes off in a tantrum). As I demonstarted in a Q&D;study on Primate a while ago, players with large home/raod splits (as compared to the "normal" splits expected from HFA and park facrtos of course) will regress back to normal splits in another time period. What this tells, as a mathematical certainty, is that abberations in home road splits are mostly noise. IOW, of we find a player, like Glavine, who has a weird home/road split, the chances of it being anything but a random fluc is very small (less than 1% maybe). To me, to go further than that is a waste of time.

Now, even if you say, "Even if there is a 1% chance of there being 'something else' going on, I want to know what it is," you are still going to be faced with the almost insurmountable problem of being able to figure out what that "something" is. Not to mention the fact that even if there is some significant "something" in that weird split, bsides a random fluc, you are never going to be able to find out whether that weird split was 40% luck and 60% something, 20% luck and 80% something, 80% luck and 20% something, etc. The only thing we know for sure, is that when we find a weird split, without knwoing anyhting else, there is a very small chance that it is due to anything significant other than "luck." If that weren't the case, we would not see near 100% regression in these studies of players who have weird splits. That really should be the end of the analysis, and if "insane" weren't an appropriate word for the "mound sepculation" thing, maybe "inane" is. It doesn't really matter what I or anyone else calls it, does it? The FACT of the matter is that the chance that ANYTHING other than luck is a significant factor in those splits is VERY SMALL (again, perhaps less than 1%). Now let's assume that it is 1%. IOW, the chances that Glavine's splits are due to anything other than luck (to more than a de minimus degree) is 1%. Considering that there must be other reasonable explanations vis-a-vis that 1% (I don't know, the baseballs at Shea, the humid air, the lighting), how much of that 1% should we assign to the "mound explanation?" It has to be less than 1% now. IOW, if the chance that "anything" other than luck is contributing signficiantly to those splits is 1% or so, the chance that it is "the mound" has to be less than 1%. How is that kind of speculation not "insane." Silly? Fruitless? Waste of time? Unproven? MERE speculation? Whatever you want to call it is fine by me, as long as you call it VERY UNLIKELY.

Now, speculation is the seed of many a good research project. Fortunately, or unfortunately, depending on your point of view, it doesn't nearly end there. You then use the scientific method to look at evidence supportint that speculation. I've done that above, and the evidence suggests that such a theory ("the mound is signficiantly affecting Glavine's true ERA at Shea and thus is a signficiant factor in explaining his unusual home/road splits in 2003") is unfounded, or if we want to be more gentle in oru semantics, "not very likely to have merit." One final piece of evidence to either put the nail in the proverbial coffin or perhaps to open up the "speculation box" again, is to look at Glavine's past performance at Shea. If it turns out that it is relatively normal, that would GREATLY reduce the already small chance that his splits are anything more than luck and that that something "more" would be the mound at Shea. If his preformance at Shea prior to 2003 is relatively normal, and someone comes back, with "Well, they could have changed something about the mound betwen 2002 and 2003, or something like that," then I give up, because, while that particular statement may be true, that person would have had absolutely no idea what I am talking about in this post. Let's see what his 99-02 numbers are at Shea. First, here are his raw unadjusted stats home and road for 2003:

Home (per 500 PA)

s=94, d=26, t=3.7, hr=13.4, NIbb=39, so=45

Road

s=73, d=21, t=9.2, hr=13.1, NIbb=34 so=59

To failry compare those, of course, they would have to be adjusted for HFA and for Shea's park factors. Regardless, it is clear that his h/r splits are indeed extreme for 2003.

Now let's look at his (road) stats at Shea only versus his road stats at all other parks in 9802. We actually "expect" the Shea stats to be better than at other parks, because it is an extreme pitcher's park, but if the "mound thing" has any merit, and it existed during those years, we would expect to see worse stats at Shea after doing the park adjustments.

Shea (199 PA)

s=75, d=15, t=0, hr=20, bb=33, so=78

All other parks but ATL (road stats) (1874 PA)

s=79, d=21, t=1.1, hr=10, bb=38, so=72

Just one thing though, MGL. What's fruitless or not worth the candle for one researcher may not be for another, because the latter may have some "clues" the former doesn't.

Let's say the chance of finding something is 10%, rather than 1%, in a given situation. For instance, Glavine's splits this season remain extreme. That would (I take it -- correct me if I'm wrong on this) dramatically increase the likelihood there truly is something genuine here and not just a random fluctuation.

Now, at that point, it would make sense for anyone interested in the Mets to want to look for that "something." OTOH, someone just generally interested in baseball might still not find the chance worth the time it would take, especially since they would still be looking for a needle in a haystack because the potential causes (lighting, the mound, Ques-Tec, family problems he gets away from on the road) are so varied and unquantifiable.

But what if you're Tom Glavine himself? Your interest is the strongest of all. This is your career, after all -- you want to identify it even if it's only a 1% chance, because Lord knows you don't want to repeat that season at your home ballpark. And you don't want to risk another season like it before you act to investigate. So you want to methodically consider every possibility. You want your pitching coach to spend the time to think through the possibilities, watch tape, comb the data for clues.

I'm guessing he one thing Glavine doesn't want Peterson, or the Mets, to tell him is, "Hey, Tom. It was a random fluctuation. Don't worry about it." For some guys, that might be just fine, especially if the "don't worry" comes from someone they trust, as Glavine trusts Mazzone, perhaps. But Glavine, with a new pitching coach? Not so sure. Plus, he strikes me as someone who believes in meticulous preparation and not leaving anything to, or believing in, "chance." This is where Chris's human element comes in: by trying to identify something that might have caused it, the Mets actually help reduce the (admittedly slim) chance that it will repeat, because it affects the player's mental approach to the game.

Now, at some point, the chance of finding anything is small enough that it's not worth the pitching coach's time to worry about it. He's got a lot of other guys to worry about, too. And the chances re. Glavine here are, based on your argument, probably that small. My point is just that the "tipping" point where it does become worth spending the time is a lot different for Glavine and the Mets than it is for you, me, or Chris. So if Peterson and Glavine want to spend an afternoon at Shea this winter playing with the mound -- just in case -- it'd be fine with me.

This is getting tiresome!

Then click on a different thread.

Chris, you do good work,

Thanks!

but you apparently don't recognoze or understand your limitations.

Oh, you didn't mean it.

I know my limitations - you simply don't know yours, or simply don't understand the exercise.

I'll address this post to Sam M., because I don't think that Chris is either able or willing to understand it.

You're so sweet.

When we find a large home/road split in a player's stats, particularly a one-year sample, but it applies to large samples as well, it is a collosal waste of time to try and "figure out" whether there is any reason for it (this is where I lose Chris who goes off in a tantrum).

A tantrum? I'm not throwing tantrums. Want to take a poll on that one?

And *no*, it is a waste of time *to you*.

Yes, the split could be random fluctuation. *Could* it be the mound? Could it? Yes? Then you are just wrong. It might not be of interest *to you*, but that doesn't make it wrong or an actual waste of time. Try not impose your beliefs on me.

Take this trip with me, MGL:
This offseason the Mets went to Atlanta and talked over the mound creation with the Braves groundskeepers (or Glavine did with Maddux/whatever). The Mets *exactly* reproduce it in Shea. Glavine posts a 2.5 ERA at Shea.

What does that mean? To someone solely absorbed in probability, that means it is just a different random fluctuation.

No, it *could* be. It could be a real effect. Can we determine if it necessarily is? Maybe, maybe not, because you cannot eliminate Glavine's belief in it.

I try to control as much as possible and then make a single change.

As I demonstarted in a Q&D;study on Primate a while ago, players with large home/raod splits (as compared to the "normal" splits expected from HFA and park facrtos of course) will regress back to normal splits in another time period. (blah blah blah)

See this, MGL - *you* didn't come up with that. That's been standard information for more than a decade. David Grabiner demonstrates as much with clutch hitting. It's been discussed ad nauseum between Nelson Lu and everyone else wrt one-year park factors. *YOU* didn't do anything except regurgitate. And that's fine, but to me, that is a waste of time. But pat yourself on the back.

And how about pulling a number like 1% out of your ass? Oh, you did, thanks.

Now, even if you say, "Even if there is a 1% chance of there being 'something else' going on, I want to know what it is," you are still going to be faced with the almost insurmountable problem of being able to figure out what that "something" is....That really should be the end of the analysis,

No, MGL, that's when *you* walk away. Good thing you aren't a research scientist.

You don't have to like it. And you can rant all you want.

and if "insane" weren't an appropriate word for the "mound sepculation" thing, maybe "inane" is. It doesn't really matter what I or anyone else calls it, does it? The FACT of the matter is that the chance that ANYTHING other than luck is a significant factor in those splits is VERY SMALL (again, perhaps less than 1%).

You are good with the ad hominem attacks. And again, you are simply making 1% up. Completely. That's a good research technique - make something up.

Is the mound "likely"? No. Is it possible? *You* acknowledge that it is.

You are pulling a Pete Rose. You say "that's nuts, what a stupid idea, but it could be right".

So you don't like it - move on. It sure as heck isn't wrong, whether you think it is fruitless or not.

Now, speculation is the seed of many a good research project.
Fortunately, or unfortunately, depending on your point of view, it doesn't nearly end there. You then use the scientific method to look at evidence supportint that speculation. I've done that above,

You did no such thing above.

and the evidence suggests that such a theory ("the mound is signficiantly affecting Glavine's true ERA at Shea and thus is a signficiant factor in explaining his unusual home/road splits in 2003") is unfounded, or if we want to be more gentle in oru semantics, "not very likely to have merit." One final piece of evidence to either put the nail in the proverbial coffin or perhaps to open up the "speculation box" again, is to look at Glavine's past performance at Shea. If it turns out that it is relatively normal, that would GREATLY reduce the already small chance that his splits are anything more than luck and that that something "more" would be the mound at Shea. If his preformance at Shea prior to 2003 is relatively normal, and someone comes back, with "Well, they could have changed something about the mound betwen 2002 and 2003, or something like that," then I give up, because, while that particular statement may be true, that person would have had absolutely no idea what I am talking about in this post.

What happens when the Mets and Glavine *say* they changed the mound? Would that in anyway change your mind? Is it possible to change your mind?

Let's see what his 99-02 numbers are at Shea. First, here are his raw unadjusted stats home and road for 2003:

Home (per 500 PA)

s=94, d=26, t=3.7, hr=13.4, NIbb=39, so=45

Road

s=73, d=21, t=9.2, hr=13.1, NIbb=34 so=59

To failry compare those, of course, they would have to be adjusted for HFA and for Shea's park factors. Regardless, it is clear that his h/r splits are indeed extreme for 2003.

Now let's look at his (road) stats at Shea only versus his road stats at all other parks in 9802. We actually "expect" the Shea stats to be better than at other parks, because it is an extreme pitcher's park, but if the "mound thing" has any merit, and it existed during those years, we would expect to see worse stats at Shea after doing the park adjustments.

Shea (199 PA)

s=75, d=15, t=0, hr=20, bb=33, so=78

All other parks but ATL (road stats) (1874 PA)

s=79, d=21, t=1.1, hr=10, bb=38, so=72

So are these adjusted or what? Care to explain the data?

Chris, you do good work, but you apparently don't recognoze or understand your limitations.

Would that we could all be perfect specimens such as yourself, Mitch.

Whatever. Get over yourself, son.

The "attacks" and barbs are getting silly. I for one will stop. I really only made one point in all of my long diatribes. And that is that from "our" perspective it is much more likely that NOTHING but random flucs "caused" Glavine's home/road splits than anything else we can speculate on.

Of course I made up the number 1% (I think I made that clear). The point and the reason I made that number up was to illustrate how unlikley it is that ANYTHING "caused" the extreme splits. Again, because we find no "player control" in h/r splits, as evidenced by the almost or complete regression in those splits, it is a FACT that the chance of Glavine's splits, from our perespective, being "caused" by anything, let alone something specific among many possibilities, is small.

How small? Of course I don't know the exact number. I can say with reasonable certainty based on the observed regressions that it is less than 10% and my best WAG is that it is somewhere in the neighborhood of 1%. But the exact number is not important at all to the thesis. What is important, is that the regressions are near 100% (there is of course sample error there). I'll gladly cede the credit, BTW, for this revelation (that extreme h/r splits are not "sustained," i.e., that we know that they are due almost entirely to luck).

As far as what to do with that information, that's entirely up to the researcher and Glavine and his cohorts. The funny thing is that one of my first thoughts was "If Glavine thought that the mound at Shea were affecting his game, don't you think he would have had the groundskeeper rectify that?" (Please don't comment on that - it is too easy to make that a red herring).

In any case, of course Glavine is free to investigate what if anything may be causing his "problems" at Shea, if he indeed thinks that there were any "problmes" beyond just bad luck at home. My WAG is that he chalked it up to luck, fluctuation, or just "one of those things" as players like to say. His buddy Maddux says that all the time, when they interview him after poor perforemances, or after poor stretches, IIRC.

The researcher is also welcome to ivestigate further if he wants. Everything beyond my one and only point in all of my posts (that the chance that it is anything but luck...) was rhetoric and hyperbole, including the "insane" comment. I often do that to illustrate and emphasize a point. I NEVER (at least rarely) do that IN LIEU of a well-evidenced point! Anyone who truly knows my research and my posts, which often go hand in hand, and doesn't have an ax to grind, knows that there is NO ONE (more hyprebole) that takes his research and what he says more seriously and presents them more carefully than I!

Anyway, the next logical step to speculating that something about Shea (which is not new) may be influencing Glavine's stats is to look at his history at Shea. I did that and found nothing remarkeable, other than a high HR rate. In fact his BB and K rates are better in Shea, even after park adjusting them. BTW, Chris, those are just raw stats that are park adjusted - that's all.

I honestly don't know what else to do to research the possibility that Glavine's splits are due to something other than luck, short of talking to Glavine or his coaches. That's why I say, and I think rightfully so, that I and others have used the "scientific method" to investigate this matter. One, the regressions, which required some research, tells us that the chance of something other than luck "causing" the splits, is very small. Two, looking at his prior stats at Shea decrease that chance even more. What else can you do? And what other conclusion can you draw other than (I'll say this only one more time, AND THAT WAS MY ONLY THESIS), that:

"The chance that Glavine's extreme h/r splits in 2003 were due to anything other than stone cold luck is very small..." If it were me, I would make that clear (I have). I would also make it clear that any other potential explanation is mere speculation and that there is no evidence to indicate otherwise, other than the splits themselves. But that's just me. Everyone else is free to express their ideas as they please.

Maybe I read the article wrong, but my impression was that the implication was that there is a fairly high likelihood (30%? 80%?) that his splits were due to something other than chance and that the mound was a good explanation. That being my impression, I felt compelled to point out that that isn't true - that unless we are privy to some inside information, there is a very small, as opposed to a high, fairly high, or medium, likelihood that his splits are due to anything other than chance. Also, as usual, I explained why this is true (the regression thing that SOMEONE ELSE came up with - I just duplicated the work to satisfy myself, which I often do), and then went one step further - looked at Glavine's prior stats in Shea to see if perhaps there were any INDEPENDENT evidence that his true perforemance level at Shea might be significantly different than elsewhere (at least in 98-02).

What more can anyone say or do? When I do research or write an article, I welcome all (legitimate) comments, good or bad, especially the bad. I recognize that some people will "reject" even constructive criticism when it comes across as being an attack, which I tend to do. That's a shame of course (both my style of criticism and the reaction). It would be nice if certain people could recognize constructive criticism for what it is regardless of how it is couched. But alas, human nature is what it is. I'm sure I am not immune to that as well...

MGL,
my flaw as a writer is that I assume the readers here *start* at the point of "it is probably luck/random fluctuation".

Once we make the assumption that happens, we work from *that* point - not point zero "what happened?"

That's what Bill James such a great writer - he managed to startpeople at the right point.

And yes, ad hominem attacks aren't very productive, so when you start calling things "piss poor" or "insane", don't expect to get very far; nor should you expect polite responses.

And yes, ad hominem attacks aren't very productive...

But they're oh so tasty. Some of us have very little else to offer the world.

:)

FIP (Fielding Independent Pitching) is a Tangotiger-created formula that, like BABIP and DIPS, shows the pitcher's own effectiveness divorced from the fielders around him. I believe the formula is:

FIP = ({HR*13}+{BB*3}-{k*2})/9

It's funny - with all the rancor being thrown back and forth I forgot all about the Questec thing! If there is an "explanation" for Glavine's poor stats at home (in NY), and I mentioned this in my original post, it is Questec! Being somewhat of a Braves fan, and because of TBS, I've watched many a Glavine game. When he throws that borderline outside pitch, you almost hold your breath waiting to see whether it gets called a ball or strike. And when the umpire refuses to give him that outside strike, unlike Maddux, who can throw many different pitches in many different locations, you wonder how Glavine's can get anyone out without getting that outside strike call. Especially with his velocity down, that inside alternative becomes less and less attractive. It would not be surprising to me if Glavine were particularly affected by Questec. The first step, of course, is to see whether and by how much Questec parks differed from other parks in terms of ball/strike ratio or bb/so rate. That is on my "to do" list.

A "boor"? How about "a boar"?

BTW, I have a pretty bad ERA projection (below average by around .25 to .5 runs) for Glavine next year, whereas Maddux still has a great projection (at least according to me), at around .5 runs or more below average...

That should be "above average" for Maddux...

I'm a bit late to the discussion, but all the talk about adjusted home/road split fluctuations being almost all due to luck got me wondering about another example.

Maybe this has already been covered somewhere else, but what about a player like Mike Cameron, who seemingly had a really hard time at Safeco. Is that a real player-specific park effect, or is it simply(bad) luck?

theberle,
Aaron Gleeman (I think), or someone, recently asked about the exact date of the new hitter's background in Safeco, to assumably look to see if that impacted any specific hitters. You may want to search his blog archives for Safeco hitters background (or even the web in general).

Read my article on extreme home/road splits. Of course an extreme sample home/road split may have something to do with the player, but it is overwhelmongly likely that it is due in most part to luck. In fact, all we can do is assume that it is luck and project a player's splits to be normal in any subsequent year. Keep in mind that Safeco is an extreme pitchers park anyway, so that SOME of Cameron's extreme splits ARE normal.

Because Cameron is a good hitter anyway, especially for a CF'er, and because he is probably one of the, if not THE, best defensive CF'ers in the game and his peripherals (e.g. baserunning) are also good, he was a bargain. He is worth around 8 mil per year.

"From MGL: BTW, I have a pretty bad ERA projection (below average by around .25 to .5 runs) for Glavine next year, whereas Maddux still has a great projection (at least according to me), at around .5 runs or more below average... "

So, how are those projections working out for you so far?

;-)