my questions on dips

[cgambill]

Exactly where have I been either nasty, aggressive, or ignorant? You said yourself you don't like DIPS and it's counter-intuitive, therefore my thinking is intuitive. Like I said, I'm finished dealing with you, but I'm sure that you'll continue to make a bigger ass out of yourself without my help. Lucky for you that you're not 'looking over my shoulder'. I will hang around here if only for the fact that the thought of it probably gets your panties in a wad. Oh...by the way...if I were you I'd think about giving Prozac a try. Have a nice life, pal.

[gmo]

Originally posted by KMac
Those of you who question DIPS on the basis of "incomplete information" may want to look up "slugging average on balls put in play." BigBadBaseball.com published it a bit ago, though I have no idea if it was their brainchild or not. Use it and tie it in with DIPS. Basically, it creates a greater separation among pitchers than DIPS does based on a similar principle.

KMac, can you point to a specific link talking about this in detail? Searching the blog at BigBad only yielded a mention that said basically only what you said above.

[cgambill]

After reading on this DIPS model most of the afternoon, I have to admit the the numbers seemingly back up his claims. However I still have reservations about it.

The average ball in play from, say, Jimmy Haynes is not going to be the same as the average ball in play from a Pedro Martinez. This component of 'slugging avg. on balls in play' seems like it could help alleviate those reservations.

I would have to be convinced as well that team defenses as a whole deviate enough from the average team defense to make such a difference so as to disregard hits allowed entirely. Especially with so much player movement from year to year. Someone who knows a lot more about statistical analysis than I do could somehow calcuate range factors and post some deviations from the average in terms of team defense. Poor fielding shouldn't matter, I don't think, because those do not contribute to earned runs, but would increase pitch count and thus decrease IP, and possibly lead to more fat pitches, so there's probably some factor there, but how to quantify it?

I would also think that the 'luck' factor of balls in play falling for hits would tend to even out through the course of a season. Guys will give up duck snorts but they'll also have some balls ripped right at people. The only thing is McCracken's wild variance of BABIP for guys like Maddux and Pedro, so it seems he is correct that they do not even out. The thing is, common wisdom is that duck snorts and hard line-outs even out for batters over the course of the season, but McCracken claims they don't for pitchers, and seemingly has the stats to back that up. Therefore, shouldn't/couldn't someone make some kind of model to correct batting averages for this?

This is a very interesting model since it's been explained/read correctly. Thanks to those who provided thoughtful responses. I think the actuality is too complex to model perfectly and this may be a step toward baseball's 'Theory of Everthing', however I'm not sold on it replacing conventional thought, more augmenting it. I'd be very interested to hear some other's thoughts on the questions I raised...

[BarryZito75]

Originally posted by cgambill
Therefore, shouldn't/couldn't someone make some kind of model to correct batting averages for this?

Batting average is a skill. The BABIP of a pitcher isn't.

[cgambill]

Why? If BABIP is not a skill because of the random luck of a ball in play landing as a hit, why does that same element of luck not also apply to the hitter? You can't say that luck plays a part on one side of the equation and not the other. If batting average is totally skill, then it follows that hits allowed indicates the pitcher's skill at combating the batter's skill, but that's not the case. Guys hit weak grounders the just squeak throught the infield, or hit routine grounders into the hole on a hit-n-run, yet they also blister some balls that go right at people for routine outs. If the variance in Maddux's BABIP is not due to a difference in skill of Maddux, then it can't be due to a sudden increase in skill of the batters he faced either.

[BarryZito75]

Originally posted by cgambill
Why? If BABIP is not a skill because of the random luck of a ball in play landing as a hit, why does that same element of luck not also apply to the hitter? You can't say that luck plays a part on one side of the equation and not the other. If batting average is totally skill, then it follows that hits allowed indicates the pitcher's skill at combating the batter's skill, but that's not the case. Guys hit weak grounders the just squeak throught the infield, or hit routine grounders into the hole on a hit-n-run, yet they also blister some balls that go right at people for routine outs. However McCracken says that's not the case. If the variance in Maddux's BABIP is not due to a difference in skill of Maddux, then it can't be due to a sudden increase in skill of the batters he faced either.

BABIP is more of an inevitability than it is random luck. I think that the varying batters that a pitcher faces results more in the year-to-year changes of BABIP rather than luck. Does BABIP know that one pitcher faced the Red Sox 4 times while he faced the Devil Rays 0 times or vice versa? BABIP is still in the early stages of development. Like any other baseball statistic it'll either evolve into another one or become obsolete due to a better system.

[rcbuss]

Originally posted by Henry
For those that still think DIPS is "dippy" let me pose you a question. Why is it the new DIPS engine works ? Why is it when I run 100 years of baseball I get statistics that are more believable than if I run 100 years of baseball in Version 5 ?

Henry

In reviewing career pitching statistics over the 100 years of your simulation, did you notice if there were any more "outliers" on the BABIP spectrum than you would expect to occur by chance (particularly amongst those with very long careers). If so, this could indicate that OOTP6 does reflect the reality that a select few pitchers do, indeed, have the ability to control BABIP slightly, as well as the general improvement of having the DIPS engine in place.

Looking forward to hearing more about your test runs in this regard.

[cgambill]

That's a very good point, Zito, but you could make the same argument for a batter's average and the effect of facing the BoSox's rotation 4 times and the DRays' rotation 0 times. Whatever you want to attribute the variance to, I don't think you can separate the BABIP of a pitcher, counting home runs, from the batting average of the hitters he faced. And I think random luck does play a part in a ball's result. Not every swing, but enough to make a difference over the course of a season. If a pitcher was lucky, as McCracken says, for one season and has an uncharacteristically low BABIP, then over the course of the season, the batters who faced that pitcher were uncharacteristically unlucky when facing him. Thus, if you adjust for the pitcher, shouldn't you also adjust for the hitter?

[BarryZito75]

The only difference between a pitcher and a hitter is there is more of a chance for a hitter to even out because a hitter is in approx. 4.5 times the amount of games than a pitcher. A pitcher facing the Red Sox in 12 percent of his games and the Devil Rays in 0 percent, is MUCH more statistically significant than a batter split such as 16 percent vs. the Sox and 14 percent vs. the Devil Rays.

[cgambill]

That's true as well. The only exception I can think of might be the limited # of games in inter-league series against only certain teams from a division, say the RSox, while another team only plays the DRays. But that is the only difference. The unbalanced schedules would even out the pitchers facing some teams more than others, at least in their own division. It could still be possible that a guy from the Royals faces the RSox multiple times and not the DRays due to the unbalanced schedule, but there's no guarantee of that. Someone could break down a pitchers appearances. For a top tier pitcher I would think that he would face everyone more or less evenly since a manager should want to get him as many starts as possible, except for someone like Pedro, who you might rest against the DRays to keep his IP low. This aspect gets real complicated real fast!!!

But I still say that you can't attempt to correct for luck for a pitcher, when the stats seem to show it doesn't even out, and then say that it does even out on the batter's end. If a pitcher is lucky on a ball hit in play, then the batter MUST be unlucky on that ball hit in play, and vice versa.

[NickG]

Well, DIPS corrects for luck and defense, which we simplify (perhaps slightly inaccurately) as luck.

FWIW, I think OOTP's pitcher fatigue model should get a bit of paying attention to for the next model (Will Carroll's "Saving The Pitcher" should help a great deal ).

[Malleus Dei]

Originally posted by cgambill
I don't think you can separate the BABIP of a pitcher, counting home runs, from the batting average of the hitters he faced.

Then you think wrong. K's are not a factor of BABIP, but are a factor of BA, and K's vary by player.

[Malleus Dei]

Originally posted by cgambill
Exactly where have I been either nasty, aggressive, or ignorant?

Right here in this thread. And far too often. Search on "jackass" for an example.

Like I said, I'm finished dealing with you

Oh, no, you're not. Dream on.

but I'm sure that you'll continue to make a bigger ass out of yourself

Than you? Not possible.

Lucky for you that you're not 'looking over my shoulder'.

Ah, but I am. Get used to it.

Oh...by the way...if I were you I'd think about giving Prozac a try.

If you were me I'd think about giving suicide a try.

Now go back and read some more. You might actually begin to understand the basics soon.

[cgambill]

Originally posted by Malleus Dei
Then you think wrong. K's are not a factor of BABIP, but are a factor of BA, and K's vary by player.

Keep up the good work, chief! You're making this too easy. At least you contributed this, but you'd simply have to take K's out of batting average to equal them out just the same.

Anyone have any ideas about team defense, specifically how much deviation there is between any one team defense and the average team defense in terms of range factor? Range factor is the only measurement I'm aware of that quantifies defensive ability, and I know it has it's flaws at certain positions, but those should equal out if applied to every team. I would be surprised if any one team defense stood out that much from the average, either good or bad, but I could well be wrong as well.

[Johnny Slick]

Okay, this is the part of the thread where I toot my own horn and point out my own very, very minor contribution to the whole DIPS thing:

On Wed, 5 Sep 2001 11:57:05 -0700, "John Craven"
<john1974@u.washington.edu> wrote:

>
>"Voros" <voros@baseballprimer.com> wrote in message
>news:3b9598ee.3265552@news.neiu.edu...
>> On 04 Sep 2001 15:41:52 -0700, Vinay Kumar <vinay@baseball.org> wrote:
>>
>> >voros@baseballprimer.com (Voros) writes:
>> >
>> >: I'm currently working on a way to better estimate defense independent
>> >: hits per balls (as opposed to giving evreybody the same rate and going
>> >: home) in play by using things like strikeouts, handedness, and whether
>> >: the pitcher is a starter or reliever (all of which seem to matter a
>> >: little bit). I'm leaving the ground ball/fly ball stuff alone for this
>> >: year as it's a can of worms I'm not ready to open yet (I have concerns
>> >: about how indpendent of defense these numbers really are). I'd also
>> >: like to give the knuckleballers their due, and I might ask the group
>> >: for help on that one.
>> >
>> >On the Fanhome board, Mickey Lichtman separated "balls in play" into
>> >fly balls, line drives and groundouts. He then looked at the $H on
>> >each of those classifications, and the year-to-year correlations both
>> >on the number of each kind of batted ball, and the $H on each. Do you
>> >think that would be a good idea?
>>
>> There is a major problem here that I can't get an answer from anybody
>> who has the data:
>>
>> "Where is the cutoff between a flyball and a linedrive?"
>>
>> Because the differences we're looking for here are necessarily going
>> to be small, a reverse caustation bias here (more hits in play causing
>> more line drives rather than the reverse) no matter how small could be
>> a _real_ problem here. This would compromise the defense independent
>> aspect of the stat.
>>
>> Another problem is that this data isn't available to the general
>> public, nor for players in the past. As a general point, information
>> can be be important but also not very useful. I think trying to set
>> something up based on line drives might fall into that category.
>>
>> If we can skin that cat using other methods that we have plenty of
>> data for (I think we can), and that we have solid knowledge that these
>> stats are independent of defesne, I think we can get at it much more
>> effectively.
>>
>> >I'm just curious for your thoughts, as I don't really remember you
>> >participating much in that Fanhome thread (which is understandable,
>> >because mgl kept referring to his work as "debunking DIPS", while I
>> >don't think he totally understands the value of DIPS).
>>
>> That was the comment I made to him in the thread. That any of the
>> things he mentioned were perfectly compatible with DIPS and could be
>> worked in if the situation warranted it. He's under the impression my
>> goal is trying to predict future performance, when that really is a
>> secondary concern, usually only brought up as additional information
>> in the debate.
>>
>
>I wouldn't go that far, myself. I've been trying to post DIPS numbers for
>the Mariners' pitching staff in the M's NG every now and then this year, and
>I think I've caught where you've done wrong. I think you're including park
>effects too early. I'm going to invoke Craven's Law here, which is the
>following:
>
>Park effects do not determine how well an individual player would hit or
>pitch in another park. They reflect how valuable his performance was in his
>home park.
>
>So what I'd do is not make ANY adjustments for walks, strikeouts, and
>homeruns, then figure out non-homerun hit values (the average value of a
>single, double, or triple), hits per balls in play, and outs per balls in
>play as though half of a given pitcher's games were being played in his home
>park. Figure out raw numbers that way and then, when you've put together an
>ERA, THEN you can adjust THAT number by park factor.

I don't know. I understand what you're saying here. There is an
advantage here in that if someone does what you have been doing
(concentrating on a single team) before adjusting for park, they have
a good way to compare the DIPS stats with the actual stats for all the
players on the team and get a good idea of it on a team basis.

But in the end DIPS is an ability stat and not really a value one (I
guess you could say it is "value due to ability alone"). As such, your
suggestion adds in an element of value that really might not be
related to the pitcher's ability.

I do use specially constructed park factors that match the constructs
in the formula (IE, instead of just adjusting for Homers I adjust for
Homers per Non-Strikeout At Bat), and I'm not unpleased with the
results.

Obviously, the formula is flexible and if you understand the concepts,
you don't really even need to adhere to it to devise different systems
to accomplish different things. I generally don't develop formulas
unless there's some sort of tangible concept I want to explore, and if
the concept is sufficiently understood, the formula itself is a bit
unimportant. The BP article didn't have the formula but still I think
explained the _basic_ concept sufficiently.

--
Voros McCracken

<stretches arms out and leans back in chair> Yep, I got Voros to concede that he was figuring out DIPS all wrong. That's right. Award me a frigging medal.

[Henry]

I think we keep looking for the "flaw" in the analysis - but keep missing the intention... here's a Voros quote...

What does this mean? Essentially it means that if a pitcher posts a very low $H rate one year, you really can't expect him to repeat that with any level of certainty at all. However if a player posts a very high $SO rate, there is a level of comfort in thinking he'll have a good one the following year as well. I cannot stress how important I think this is. Think about a second. How much value would you give to a stat, if you KNEW that it meant virtually nothing towards that players future stats? IOW, we don't give Aaron Sele credit for the seven runs a game his team scored for him because we know that he had little to do with it. Those runs are very real and very valuable but there isn't a real reason to give Sele credit for them. In my opinion, Hits Allowed deserves the same treatment.

If there is no correlation to hits, then hits ARE the result of the batter and defense. You can't say the pitcher has control over this variable if there's NO correlation.

Having said that, the ONLY tree we should be barking at is the remote POSSIBILITY that SOME pitchers (knuckleballer's?) MAY have some slight control over hits - HOWEVER -

We aren't the only ones that have been looking. Some of the masters in the field have looked again and again from every angle they can think of and they come up with the same thing...

Even those pitchers that have some slight control, have so little that the law of averages (luck) and other variables (ballpark. defense, etc.) simply overwhelm that ability.

Henry

[Viva Los Expos]

Originally posted by cgambill
I would have to be convinced as well that team defenses as a whole deviate enough from the average team defense to make such a difference so as to disregard hits allowed entirely.

But nobody is asking you to assume that. The DIPS model isn't based on finding differences in BABIP and concluding "oh, that's just defense." I don't understand your point here.

Poor fielding shouldn't matter, I don't think, because those do not contribute to earned runs, but would increase pitch count and thus decrease IP, and possibly lead to more fat pitches, so there's probably some factor there, but how to quantify it?

Fielding matters because poor fielding doesn't just show up as errors. It shows up as hits that a fielder with better range would have gotten to, or throws that would have been in time coming from a fielder with a better arm, etc. -- and the reverse of all those, such as a great fielder "robbing" the batter of a base hit.

I also don't understand your reference to earned runs.

[Malleus Dei]

Originally posted by Henry
We aren't the only ones that have been looking. Some of the masters in the field have looked again and again from every angle they can think of and they come up with the same thing...

Even those pitchers that have some slight control, have so little that the law of averages (luck) and other variables (ballpark. defense, etc.) simply overwhelm that ability.

Henry, I have pointed this out over and over. You have pointed this out over and over. The problem is that (1) some people have no comprehension of sabremetrics and refuse to accept DIPS; (2) some people are too dumb to understand DIPS, (3) some people refuse to do the homework and read the articles and keep bringing up the same stupid points over and over again, and (4) some people think DIPS is somehow negotiable - that they are a Secret Master of Sabremetrics and are fit to dispute this with the real masters (not that they will ever do that; instead they spout total nonsense here where only a few us will slap them for it).

Let me repeat this one more time: PITCHER EFFECTS ON BABIP HAVE BEEN DEMONSTRATED TO BE STATISTICALLY IRRELEVANT.

[Malleus Dei]

Originally posted by cgambill
you'd simply have to take K's out of batting average to equal them out just the same.

And if you do that then it's not BA. Are you really this dumb or is this all just an act?

You really need to go read all the articles on DIPS. You keep trying to argue silly points that were smashed flat by experts years ago.

[treedom]

Mal, the theory of gravity just doesn't sit well with me. I've read all the literature, and I've even done some experiments dropping stuff from my balcony. The observed behaviour seems to support the theory, but something just doesn't feel 'right' about it. Can you explain it to me?