Saber and the Pitcher

by thatdog on Mar 9, 2008 1:09 AM PST Comment 54 comments

Neat story about how Brian Bannister is trying to use stats and pitch f/x to improve his game. He seems to be approaching it realistically and with some intelligence.

There's also a couple of little nuggets in there about BABIP. Bannister had a somewhat low BABIP last year and apparently Zito did as well. Last year's leaders for BP members. For those that are not members, the top five are:
Orlando Hernandez .228
Chris Young .246
AJ Burnett .262
Brian Bannister .264
Barry Zito .264

It's suggested by Bannister that good curveball pitchers might have better BABIP than expected. Well, Zito has done pretty well with BABIP over the years:

.238
.288
.251
.244
.299
.249
.287
.264

Now the low BABIP didn't necessarily mean success for Zito last year. I just found it interesting because I thought BABIP should average out to .270 to .280. I also think it's fantastically interesting to hear a player talk about stats in this way. It's a decent little read. Passan wrote it, but don't let that stop You.

I hope Bannister has continued success. Here's to more players approaching the game in a similar manner.

This FanPost is reader-generated, and it does not necessarily reflect the views of McCovey Chronicles. If the author uses filler to achieve the minimum word requirement, a moderator may edit the FanPost for his or her own amusement.

0 recs | Comment 54 comments

Story-email Email Printer Print

Comments

Re: Saber and the Pitcher

While we're on the subject of pitchers, did anybody know that Messenger got sent down??

Nattowear | artstuff | Durham? I know 'im!

by Natto on Mar 9, 2008 1:23 AM PST reply actions 0 recs

Sorry...Me disculpo...Je fais des excuses

OK, the mea culpa is on GFIS' diary. I started my post and by the time my dim brain could pound out the words, GFIS had posted his. Honest mistake, I swear.

Just the same, the article is worth reading.

-- Born Yesterday

by thatdog on Mar 9, 2008 1:45 AM PST up reply actions 0 recs

Natto there's this cool video too!

Some of the 1983 Dodgers pitchers (along with other players) made this video you should watch it!

Coming to you from the Land of Many Beers

by WalrusMan on Mar 9, 2008 9:07 AM PST up reply actions 0 recs

Re: Saber and the Pitcher

You're such a Lowell snob. :)

by E Ticket on Mar 9, 2008 11:38 AM PDT up reply actions 0 recs

Zito has always done well on hits on balls in play

According to BPro, Zito over his 8 year career, has allowed 112 hits fewer than he "should".

Tom Glavine, (in)famous defier of DIPS, has 53 fewer hits than he "should" have. His "luck" was in allowing far fewer runs than he should have, 93.

Pedro, 69 fewer hits than he "should" have.

Greg Maddux, 60 fewer.

Koufax, 94 fewer.

Duque, 75 fewer.

Burnett,50 fewer.

Young, 55 fewer.

Beckett, 29 fewer.

And Matt Cain, 37 fewer.

Also, DIPSomaniacs should read this post by Tango, on career DIPS numbers, and also some of the comments:

This means that if you have a pitcher with 3700 BIP, you regress his sample BABIP 50% towards his teammates' BABIP. If you have 1850 BIP, you regress two-thirds toward your teammates' BABIP. With 7400 BIP, you regress one-third toward your teammates' BABIP. For your typical pitcher with 500 BIP in a season, the regression amount is 88% (i.e., r=.12, rSquared=.01). It's on this basis that you will often hear that pitchers have little influence. What's really true is that one-season's worth of stats is hardly indicative of his skill."

This is why we say that pitcher's haev little influence on balls in play. What it ACTUALLY means is that the metric hits per ball in play explains very little, if you only have 500 balls in play. If you had 37,000 BIP, it would explain almost everything. The problem is that no pitcher is involved with 37,000 BIP in his career.

by rfloh on Mar 9, 2008 7:21 AM PST reply actions 0 recs

Useless

Which means that if it's impossible to get a large enough sample size to test the hypothesis, then the parameter is useless.

by DrBGiantsfan on Mar 9, 2008 9:45 AM PST up reply actions 0 recs

I'm not sure what your point is

yes, with most things, if the sample is not large enough, you cannot draw any meaningful conclusions?

So?

by rfloh on Mar 9, 2008 10:50 AM PDT up reply actions 0 recs

Re: I'm not sure what your point is

If I understand correctly, you're implying that it makes no difference if a pitcher throws groundballs or flyballs or foulpopups?

Are you saying that there is no way to statistically compare pitchers once a ball is put into play?

Or are you saying that this particular attempt to quantify a pitchers ability is full of shit?

None of the above.

(I pick no. 4 personally)

by E Ticket on Mar 9, 2008 11:41 AM PDT up reply actions 0 recs

4, I think.

I'm sure you have seen various posts debating whether Lowry was "lucky" or not in 2007.

Also, posts on whether Zito has been "lucky" throughout his career.

DIPSomaniacs, contend that MLB pitchers have little / no control over hits on balls in play, ie, the pitcher has no control on whether a ball that is put in play ends up as a hit or not; hence you might see statements that Josh Towers is as good a pitcher as Carlos Silva. Some take this further, and say that not just MLB pitchers, but all pro pitchers have no control over hits on balls in paly

The article / post / comments I linked is saying that this is NOT true. It is saying that controlling hBIP IS a skill. But a skill that is difficult to measure and separate from various other factors.

Hence the need for a reasonably large sample, ~5 seasons, before an individual's career numbers on controlling hBIP as an indication of actual skill.

by rfloh on Mar 9, 2008 12:18 PM PDT up reply actions 0 recs

Jury's Out Indefinitely

And at his current rate, Lowry will take 12 seasons to give you 5 complete.

by Moggeee on Mar 9, 2008 12:25 PM PDT up reply actions 0 recs

Re: Jury's Out Indefinitely

Games or seasons. Never mind. He's high. On a Hill. Where little cable cars climb halfway to the stars and fly-balls land in the bay

by E Ticket on Mar 9, 2008 3:43 PM PDT up reply actions 0 recs

Lowry, too, Out Indefinitely

Yeah, he's high. On anesthesia.

On a gurney being pushed halfway to the recovery room.

by Moggeee on Mar 9, 2008 5:38 PM PDT up reply actions 0 recs

Re: 4, I think.

I couldn't agree more. If hBIP was not a skill, batting practice pitchers would be in the rotation. But how one measures this objectively is fairly freeking impossible as you suggest. Too many variables.

Thanks much for the explanation

by E Ticket on Mar 9, 2008 3:41 PM PDT up reply actions 0 recs

Re: 4, I think.

While one could argue that Barry Zito has been lucky, he has been "lucky" enough for long enough that it would appear to be at least partially something else.

Now Noah Lowry -- he was almost certainly flat-out lucky last season.

by sharksrog on Mar 10, 2008 10:50 PM PDT up reply actions 0 recs

So.....

We keep hearing BABIP brought up in arguments as evidence of one thing or another. I'm just saying, there's no point in bringing it into any discusssion if it's impossible to get a valid sample size.

by DrBGiantsfan on Mar 9, 2008 12:00 PM PDT up reply actions 0 recs

But that is NOT what Tango is saying

First this is regarding pitchers, not hitters.

Secondly, what Tango is saying that controlling hits on balls in play IS a skill. HOWEVER, it is a skill that is difficult to separate from just random noise.

Thus, you need a reasonably large sample, before you can look at a pitchers career numbers, instead of just regressing his numbers to team / league average, and conclude whether he is better, or worse than average at controlling balls in play.

by rfloh on Mar 9, 2008 12:08 PM PDT up reply actions 0 recs

Yes, I Understand.....

but, if you can't get a large enough sample size, you can't draw any conclusions, which makes it a useless stat.

by DrBGiantsfan on Mar 9, 2008 12:13 PM PDT up reply actions 0 recs

Ok, I apologise

I'm saying this very badly.

What I am trying to express, poorly, is that the smaller the individual's sample, the more you should "regress it to the mean", ie team / league hBIP.

This applies to all the other stats too. If a player puts up a 500 OBP in 10 games, most people are not going to conclude that he is the 2nd coming of Barry Lamar Bonds. They expect that he will regress to the mean.

As for hBIP, the section from the comments that I quoted

which is basically saying that you need 5-6 "normal" seasons, before you can look at someone's hBIP's numbers, and conclude that he has more / less / equal levels of skill at controlling hBIP as league average.

by rfloh on Mar 9, 2008 12:29 PM PDT up reply actions 0 recs

Re: So.....

there's no point in bringing it into any discusssion if it's impossible to get a valid sample size.

Because of the small sample size, when a pitcher has a season or two with good ERA thanks to a very low BABIP, but unimpressive K, BB and/or GB rates, we expect the pitcher to regress soon.

BABIP is used to measure how lucky a pitcher got in a certain season because it's unreliable, not in spite of it.

You deserve to be struck out, when your first name's a verb.

by Cookyman on Mar 9, 2008 12:13 PM PDT up reply actions 0 recs

Re: So.....

I am saying the whole thing is a bunch of poppycock from people who pretend to understand statisitical analysis.

by DrBGiantsfan on Mar 9, 2008 3:07 PM PDT up reply actions 0 recs

Re: So.....

I am saying the whole thing is a bunch of poppycock from people who pretend to understand statistical analysis.

Well it's hard to argue with that point, but do you agree that:

A season or two is a very small sample size for measuring BABIP.
A season or two is a good sample size for measuring K/9, BB/9 and GB rate.
K, BB and GB rates are better indicators of how a pitcher pitched in a season than BABIP.
Therefore, when a pitcher has a low ERA thanks to a low BABIP but bad peripherals, we consider him lucky, and expect him to have a much higher ERA in the next season.

You deserve to be struck out, when your first name's a verb.

by Cookyman on Mar 10, 2008 11:22 AM PDT up reply actions 0 recs

What I Think

BABIP is a useless stat.

K/9 and BB/9 are useful stats for evaluating pitchers, but only when viewed in context.

GB rate tells you what kind of a pitcher you have, but not how good that pitcher is.

Matt Cain would be an example of a pitcher who you would be likely to severely misjudge by those numbers alone. You have to actually see him pitch from game to game to really understand what we have in him, and that he ain't regressing to no mean.

by DrBGiantsfan on Mar 10, 2008 1:42 PM PDT up reply actions 0 recs

Re: What I Think

he ain't regressing to no mean

And yet he already has. Remember all those discussions we had in April and May of last year about whether Cain's fantastically low hit rate was sustainable? Turns out it wasn't. He regressed toward the mean.

by Evan on Mar 10, 2008 2:46 PM PDT up reply actions 0 recs

We Don't Know That.....

because we don't have a large enough sample size to know what Matt Cain's mean is.

by DrBGiantsfan on Mar 10, 2008 4:13 PM PDT up reply actions 0 recs

Re: We Don't Know That.....

Matt Cain isn't mean, he's probably nice.

by xanthan on Mar 10, 2008 4:23 PM PDT up reply actions 0 recs

I Agree

by DrBGiantsfan on Mar 10, 2008 4:30 PM PDT up reply actions 0 recs

Re: We Don't Know That.....

True enough. It is possible that Matt Cain is far, far better at preventing hits than any other pitcher who has ever lived. We can't prove otherwise. If you enjoy approaching the game that way, have fun with it.

by Evan on Mar 10, 2008 4:58 PM PDT up reply actions 0 recs

He Is.....

up to this point. That's really all we know.

by DrBGiantsfan on Mar 10, 2008 10:04 PM PDT up reply actions 0 recs

Re: What I Think

BABIP is a useless stat.

I think that is what this work tells us.

K/9 and BB/9 are useful stats for evaluating pitchers, but only when viewed in context.

And what context would that be?

GB rate tells you what kind of a pitcher you have, but not how good that pitcher is.

I think the problem with all these super-de-dooper pitching stats is that it's very hard to track LD%. If you break up all the pitchers PAs you get:

K (good)
uBB (bad)
IBB (bad for team, but no bearing on pitcher)
HR (very bad, and very rarely not the pitchers fault)

And "BIP".
These are readily divided into 1B, 2B, 3B, incredibly rare HR, GO, FO, Foul Outs, E, etc. Some 70% of these are outs, and the rest are hits.

But everyone knows that weak choppers that are legged out for singles and balls crushed to the gap that Mike Cameron somehow fields are not treated correctly by BIP.

And all results of BIP are effected by the defense behind the pitcher, as well as how hard / how high / where the ball was hit.

Strong DIPS supposes that SINCE good/bad BABIP regresses to the mean, it must not be a skill. This is obviously an oversimplification, but the fact is - it's not obvious which pitchers (from year-to-year) are good at this. Only over several years do your Glavines and Zitos emerge... and even then, I have never seen a report that this was statsically significant.

I think as the tech gets better on BIP, we will finally learn how defense and BIP are coupled and you will get things like LD% in your USA today.

by zenbitz on Mar 10, 2008 4:05 PM PDT up reply actions 0 recs

K/9, BB/9, Etc

K/9 can be one measure of pitching dominance, which is not the same thing as pitching success. Other measures of pitching dominance could be an extremely high GO/AO such as Brandon Webb turns in with regularity. Another could be all the pop ups Matt Cain induces. K/9 seems to be the easiest to measure, plus it's the most exciting so that's the one we tend to gravitate to.

A low K/9 coupled with a low BB/9 may mean that a pitcher is pitching to contact and winning by keeping his WHIP down, or it may mean he's not a good pitcher.

A low K/9 coupled with a high BB/9 usually means a bad pitcher, but there are exceptions for extreme GB pitchers who may induce more GIDP's and keep HR's down.

A High K/9 coupled with a low BB/9 is probably an elite pitcher because he is able to force batters to swing and miss at pitches in the strike zone.

A High K/9 coupled with a High BB/9 may mean the pitcher is achieving the K's by trying to force batters enlarge their strike zone, or it may mean they pitch with high velocity and are just plain wild.

As you can see with most of these, context is the key to a proper interpretation.

by DrBGiantsfan on Mar 10, 2008 10:17 PM PDT up reply actions 0 recs

Re: What I Think

Dr., I thought you were far too wise to say that BABIP is a useless stat. I'm just a layman, but I use it in several useful ways.

Saying that BABIP is a useless stat is akin to saying that GB% or a host of others are useless stats.

by sharksrog on Mar 10, 2008 10:54 PM PDT up reply actions 0 recs

Re: What I Think

I would say that BABIP is a fairly useful retrospective stat. Year to year, though, it's not a lot more than a general equalizer without a significant amount of context.

He is Vengeance. He is the Knight. He is Dave Righetti. PRAY TO HIM!

by howtheyscored on Mar 10, 2008 11:09 PM PDT up reply actions 0 recs

Wise

No, I am wise enough to know it's a useless stat.

by DrBGiantsfan on Mar 10, 2008 11:21 PM PDT up reply actions 0 recs

Re: Wise

Wise men know what they don't know. One can certainly find fault with BABIP, as with any stat, but it most certainly isn't useless.

Give me the name of a position player, Dr., and I'll very likely show you how BABIP analysis can or has improved our ability to look forward at that player.

You pick the player, and I'll see what I can do with BABIP and that player.

by sharksrog on Mar 11, 2008 10:48 PM PDT up reply actions 0 recs

Re: What I Think

If you need 37,000 BIP to get a good sample, then, yes, it is useless.

by zenbitz on Mar 11, 2008 10:19 AM PDT up reply actions 0 recs

37000?

I think that had an extra zero on it, but still, 3700 BIP's is close to half a career's worth to esablish a mean. Couple that with the fact that career trajectories are not horizontal lines, well, it's still a useless stat.

by DrBGiantsfan on Mar 11, 2008 10:52 AM PDT up reply actions 0 recs

Re: What I Think

Unreliable is not the same as useless. Coockyman nailed it up above:

Because of the small sample size, when a pitcher has a season or two with good ERA thanks to a very low BABIP, but unimpressive K, BB and/or GB rates, we expect the pitcher to regress soon.

BABIP is used to measure how lucky a pitcher got in a certain season because it's unreliable, not in spite of it.

by Evan on Mar 11, 2008 11:01 AM PDT up reply actions 0 recs

Absurd

Sorry to be blunt, but that is one of the more absurd statements I've read lately.

by DrBGiantsfan on Mar 11, 2008 2:22 PM PDT up reply actions 0 recs

Re: Absurd

I know I'm going to regret this, but -- why?

by Evan on Mar 11, 2008 3:26 PM PDT up reply actions 0 recs

Why?

You can't use it to measure luck, because you don't know whether it's luck or not. Your sample size is too small to be significant, so the best you can say is the deviation might be due to luck, which also means it might not be. You simply cannot draw any conclusions from an indequate sample size.

by DrBGiantsfan on Mar 11, 2008 8:59 PM PDT up reply actions 0 recs

Re: Why?

You can't draw any firm, ironclad conclusions, but you can make some educated guesses. If you have ten pitchers who post a low BABIP in a given year (say .260 or below), nine of them will see their BABIP rise, often by a lot, the following year. If you're a GM paying a pitcher on the basis of his ERA, that's information you really need to know.

by Evan on Mar 11, 2008 9:54 PM PDT up reply actions 0 recs

I Don't Agree With That

by DrBGiantsfan on Mar 11, 2008 10:49 PM PDT up reply actions 0 recs

Re: What I Think

It's useless because it's only the "remainder" of a stat.

You calculate DIPS-RA (see, for example: http://www.hardballtimes.com/main/article/batted-balls-and-dips/)

And look at the difference between actual RA to get a feel as to whether a guy was lucky or not. DIPS-ERA assumes that the BABIP is constant for all pitchers. What the actual BABIP is for someone isn't really relevant.

You can't calculate his DIPS-RA, show that it's worse than his regular RA and then say "it's because his BABIP was low". It's implied.

by zenbitz on Mar 11, 2008 3:39 PM PDT up reply actions 0 recs

Re: What I Think

You can't calculate his DIPS-RA, show that it's worse than his regular RA and then say "it's because his BABIP was low". It's implied.

When did someone here say that? I missed it.

If all you're interested in is a consolidation stat that gives you a more predictive ERA simulacrum, fine. But BABIP is something completely different. It's a mere record of what happened on the field.

by Evan on Mar 11, 2008 10:01 PM PDT up reply actions 0 recs

Re: What I Think

I quote (not you): Because of the small sample size, when a pitcher has a season or two with good ERA thanks to a very low BABIP, but unimpressive K, BB and/or GB rates,

That is effectivly what I said. The fact that there is a difference between ERA(real) and ERA(K/BB/GB) - he left off HRs, critically - implies there the pitcher has a low BABIP. There is no need to really actually calculate it, especially when you know it has sample size issues.

BABIP is kind of like batting average when you already have OBP. It tells you very little.

OK, maybe I am wrong (after all I am agreeing with DrB, which I basically never do).

You want to predict pitchers who will regress/progress based on babip. Is this list difference than the list of pitchers that will regress/progress based on DIPS-ERA - ERA?

by zenbitz on Mar 12, 2008 9:34 AM PDT up reply actions 0 recs

Re: What I Think

A player's BABIP in a given season compared to his career BABIP can show us if he is likely to bounce back from a poor season or fall back from a good one.

And we can apply a .300 BABIP to a player's minor league record in order to get a better approximation of how high he might be able to hit in the major leagues.

Try it. You'll be surpised at what you may learn from it.

I'll offer you the same challenge I offered the good Dr. Give me the name of a position player, and I likely can tell you something you can or could have learned about him from his BABIP.

BABIP isn't a be-all, end-all. But do you know any stat that is?

by sharksrog on Mar 12, 2008 11:23 AM PDT up reply actions 0 recs

Re: Zito has always done well on balls

Didn't Tommy John have a silly low BABIP over a silly long career?

Feliz gone, team still bad

by oldjacket on Mar 9, 2008 11:56 AM PDT up reply actions 0 recs

Per BPro,

John allowed 152 hits MORE than he should have. His career BABIP is 289, which is not silly low.

You can find the "lucky" hits numbers on BPro's DT player cards. You do not need to be a subscriber, I believe.

It's the column labeled as D(elta) H(its).

Where John excelled was in controlling runs. He allowed 97 runs fewer than predicted.

Like Glavine, this might have been due to him changing his pitching strategy with men on base. Or maybe random noise.

by rfloh on Mar 9, 2008 12:42 PM PDT up reply actions 0 recs

I'm sorry

but this just seems a bit obscure to me. To track a hitter's "lucky" hits or a pitcher's "lucky" outs is an exercise in futility.

For every lucky hit that Pete Rose had, over his 20+ year career, he must have had 1 +/- .001 "unlucky" outs. So, over time, a players numbers are going to reflect his abilities with pretty darn reliable accuracy. There are very few outliers.

Right?

Barry Zito: Throws until there's nothing left -- which is usually around the 3rd inning. Barry's a GAMER!!!

by Woody Wins on Mar 9, 2008 1:09 PM PDT up reply actions 0 recs

Firstly, there is a difference

between hitters and pitchers, in terms of "lucky" hits.

It has been shown in various studies, that hitters do have more control on whether a ball they put in play ends up as a hit or not.

Many studies have shown that pitchers' abilities to control hits on ball in play is much less.

The first and most famous, is Voros McKracken's D(efense) I(ndependent) P(itching) S(tats) study. Which is where the idea that league average BABIP is around 290-300, and that a pitcher who has allowed a much lower BABIP, say 260, is "lucky", and thus going to regress to the mean, comes from.

The mean being team / league average. So, what this delta hits stat does is, it uses Voros' DIPS formula to calculate whether a pitcher was "lucky" or not relative to his teammates.

From BPro's glossary:

The number of hits above or below average for this pitcher, based on his own number of balls in play and his team's rate of hits (minus home runs) per ball in play; (H-HR) - BIP * (team (H-HR)/BIP). Essentially, the Voros McCracken number.

by rfloh on Mar 9, 2008 1:25 PM PDT up reply actions 0 recs

Also, yeah,

the post that I linked up top, is saying that over time, a players numbers are going to reflect his abilities with pretty darn reliable accuracy.

The problem is how long a time.

This is why we say that pitcher's haev little influence on balls in play. What it ACTUALLY means is that the metric hits per ball in play explains very little, if you only have 500 balls in play. If you had 37,000 BIP, it would explain almost everything. The problem is that no pitcher is involved with 37,000 BIP in his career.

by rfloh on Mar 9, 2008 1:28 PM PDT up reply actions 0 recs

Re: Also, yeah,

That makes sense, and I see the hitter vs. pitcher differences.

I think it's just hard for me to wrap my mind around the concept when there seem to be so many intangibles: day game versus night game, defensive positioning, different ballparks, weather, umpire's strike zone, etc. When you take all of these things into account it can skew the numbers I would assume. Even with adjustments for league-averages I have to think that there is a substantial degree of error. Very interesting stuff though, thanks for the info. I've learned a ton from all these different perspectives, and that's the beauty of the game!

Barry Zito: Throws until there's nothing left -- which is usually around the 3rd inning. Barry's a GAMER!!!

by Woody Wins on Mar 9, 2008 2:19 PM PDT up reply actions 0 recs

Yeah, intuitively

DIPS, that pitcher control over hits on balls in play is much less than expected, is a concept that does not appear to make sense.

It's why, Brian Bannister, the pitcher linked in the main post in this thread, is becoming a cult favourite among statheads: he not only is aware of the concept, he acknowledges it, is intrigued by it.

by rfloh on Mar 9, 2008 2:27 PM PDT up reply actions 0 recs

Re: I'm sorry

There certainly aren't going to be many lucky or unlucky hitters with as many at bats as Pete Rose had. On the other hand, didn't he have more than anyone else?

by sharksrog on Mar 10, 2008 10:55 PM PDT up reply actions 0 recs