SB Nation - Login for mobile commenting

Lookout Landing

More On Strikes & Expected Strikes

Yesterday I presented a simple method of figuring out which pitchers have been working with the most generous zones, and which have not. With a range from 3% more strikes than expected to 3.6% fewer, this definitely has the look of being somewhat significant, with the upper quartile getting more strikes than the lower quartile by 30 per 1000 pitches.

The next step in anything like this is to examine the year-to-year correlation, to see whether you're looking at a repeatable trait or a bunch of random noise. Towards that end, I grouped together the 199 guys who threw at least 750 pitches in both 2008 and 2009, calculated their expected strike rate in each year, calculated the difference between that and their actual strike rate in each year, and then looked at the relationship. The result?

Strexpstrcomp_medium

That's an r value of 0.034 and a significance F of 0.632. In other words, based on this data, there is no significant relationship between Diff% in Year X and Diff% in Year X+1. Suggesting that getting good and bad calls is less a systematic thing and more just umps choosing certain days to be funny.

This is by no means conclusive, and it's entirely possible that certain individual pitchers may be able to game the system a little bit. Since 2005, for example, Jamie Moyer has an average Diff% of +0.6%, which works out to an extra 18 or 20 strikes a season. But what does that really mean? Does it mean Moyer's doing something special, or is it statistical happenstance? Between 2005-2008, Greg Maddux comes out a slight negative. Jake Peavy's near the top of the leaderboard in 2009, but finished in the negatives in each of the four years previous. My inclination is to believe that, if there is skill involved in getting calls, it makes such a small difference that it's lost in the noise.

Of note is how heavily this regresses. A table, using data from 2005-2009:

Average Diff%
Pitches Top 10 Bottom 10 Range
3000+ 1.1% -0.7% 1.8%
2000-2999 2.2% -1.4% 3.7%
1000-1999 2.5% -2.0% 4.5%
750-999 2.2% -2.5% 4.6%

Pitchers who threw at least 3000 pitches in a season showed little variation in their Diff%, while pitchers who threw between 750 and 999 pitches in a season showed something more substantial. This is another bit of evidence that, if getting calls is one part skill, it's also about twenty parts luck.

Umpires: annoying? Yes. Manipulable? Not as far as I can tell.

1 recs  |  48 comments

Comments

Only a couple more steps until we have UIP

All of this (today and yesterday) assumes that all umpires have roughly the same strike zone (or variance to strike zone, or # of games where they just go street rat crazy). Or possibly, thought of another way, it assumes that all umpires move around the league enough that all pitchers see them an even amount of times.

It’s possible that some pitchers are seeing specific umpires way more often. Or that some pitchers are compatibile or incompatible with specific umpires, and happen to see more compatible or more incompatible umpires more often.

It just feels like say, a disproportionate # of CB Bucknor appearances for a given pitcher could really throw your calculations out of whack. Enough to be beyond the margin of error.

If we could ever get them, umpire stats should be factored in. E.g. overall strike zone accuracy, their consistency, how often low-strike-zone umps seemed to be in games with guys who general throw lots of low strikes, times where umps were paired with particular pitchers or particular teams a disproportionate amount of times.

So what's Rob Johnson's cDiff%?
I don't know how signficant the results would be

But if a pitcher has been screwed on something like 30-50 strikes on the season, it would stand to reason they could be slightly better than their results, yeah? How viable a tool would this be for undervalued assets? Or would it not be that great since 30 strikes is probably only 4-6 more strikeouts or so?

It's something to look at, just in case
I would think that getting hosed on 30-50 strikes a season would be very very significant.

Especially if you could take into account when these strikes are called balls. I don’t know if it is even possible to collect that data, but getting hosed on a 0-0 count would weigh higher than getting hosed on a 2-1, 2-2, 3-2 count.

It’s possible, not probable, that every one of the 30-50 strikes came in high leverage situations. That could lead to more mistake pitches, and more pitches in favorable locations for the batter. I think that could easily make an above average pitcher look way worse than he actually is.

Getting hosed on a 3-2 would be way worse than on a 0-0
Wow, I missed completely.

That was all backwards. Getting hosed on a 2-1, 2-2 etc is way worse than a 0-0 is what I meant.

I got off light

Getting hosed hurts

From Tango’s analysis of hitting by pitch count in the 2006 Baseball Handbook, it appears the “best” situation for a pitcher to get screwed is on an 0-0 count, when the blown call changes the wOBA from .283 to .371. By contrast, going from 2-2 to 3-1 raises the wOBA from .290 to .490, and of course there are the results where the blown call awards a walk or takes away an out.

The question for someone more mathematically inclined than me is to figure out what it would mean over the course of a season for a pitcher to face 30 batters whose wOBA is artificially inflated by the blown call.

It seems to me that the 'best' and most common blown calls are the free strikes umps give out on 3-0
Right

But the interesting thing is trying to measure how under-valued or over-valued a pitcher may be if they are at either end of the blown call spectrum. Those common blown calls in the pitcher’s favor aren’t really harmless either since the batter goes from a walk to a 3-1 count where their wOBA is .490. What’s the wOBA of a walk?

.720
In other words

Whatever the count, blown calls have a big effect on any at bat. When a strike is called a ball, the batter’s wOBA goes up anywhere from 90 to 250 points, or even more when they have two strikes on them.

If you take 30 batters a pitcher faces over a season and artificially inflate their wOBA by, say, an average of 180 points, I think that works out to an extra 4-5 runs allowed. Not a huge effect on how to evaluate a pitcher, but not irrelevant.

I'll have to do a really big look at catcher framing one of these (offseason) days.
Huh, this project and the Adrian Beltre heat map covered the two items that interested me the most.

Whether or not pitchers with control reputations like Jamie Moyer could expand the zone significantly and whether or not pitchers can expand a batters zone. I’m not sure what the value of that is exactly, but a pitcher of beer ought to suffice.

Good analysis.

Especially the final table….throw more pitches? Regress further to the mean.

br

You have 'br', whatever that represents, already in your signature.

It’s not necessary, and actually annoying to some, to repeat yourself and waste vertical space, especially on something as trivial as that.

vr, Xei

I finally saw it.

The guy makes some good posts on True Blue LA

but the sig thing annoys me.

I think the Hebrew meaning of your name, "Gift of God," sometimes goes to your head.

The “br” is out of habit from my workplace. We all sign our names in two letters. It usually is a quip depending on the week but I’ve always stayed consistent.

Also, i’ve never posted on True Blue LA. I’m Mariner tried-and-true.

No offense, brother. Just a force of habit.

br

I hope this wasn't intentional
Nobody gives a fuck about your habits

I would advise that you quit the backtalk.

Your tone is unneeded and you apparently confused a suggestion with a formal request.

Knock it off.

The True Blue LA comment had nothing to do with you.
Would a catchers relationship with an umpire

make any difference on strikes% and expected% throughout a pitchers career? I realize this may be somewhat of an obvious question.

Do you think you are just measuring some combo of

umpire error and pitch location error?

That 1" location error is going to have such a huge effect
Well there is strike zone location error which I'd assume would have a bigger error than the pitch tracking

And its not like we are talking about huge deviations here.

30 pitches out of 1000 isn’t too insane. It sounds somewhat significant but after you start thinking about random error from pitch f/x, pitch f/x strike zone location, and umpire ability (listed from lowest error to highest) then I think it starts sounding like a reasonable amount of error.

Are you seriously suggesting that pitch f/x operators are unable to locate knees and/or letters?
The stringers make errors. It's not uncommon at all.
Yeah

The knees are >> 1". People move. Everything is a little unknown.

If you just think of the strikezone as a rectangle surrounded by a fuzzy area that is 2 inches wide (a little under the diameter of a baseball) where the system is right/wrong 50% of the time, you are still going to get a significant amount of wrong calls. Especially because this is the area most pitches will be located since nobody is trying to throw extremely far out of the zone or down the middle of the plate.

The stringers make a lot of errors

Beyond just inputing the top and bottom strike zone. Sometimes, pitches have the wrong pitch f/x data from. Mike Fast has personally found a lot of errors just on first glance.

The problem is there is no way to weed them out, as far as I know.

So there is still human error in the system?

This is interesting.

It's still probably better than umpires

But I think there is a lot of error in pitch f/x location still. That would explain the discrepancy between Jeff’s method, which doesn’t use pitch f/x, and my/jon/dan’s method which does.

They are mostly, if not all, weeded out by hand in the data used by MLB.

Not to mention the significant pitch f/x park effects that need correcting as well.

Are these still pretty serious?

I know they were a few years ago, but I’d hoped the most egregious park effects had been addressed this year. Maybe not….

Probably not serious

but still statistically significant. It’s mostly a speed reading issue.

Yeah; better (for this purpose) than the errors in pitch location that seemed to plague the system

in years past.

Let it be considered as well,

that the pitch f/x available to the public is not as good as the pitch f/x available to teams. I would guess a healthy portion of skewed results in a study like this are systematic or stringer errors.

I'm not sure that Fangraphs' Zone% is based on PITCHfx
It uses BIS
How does BIS spot the pitches?
I'm not sure

Zone % goes way further back than Pitch f/x data, so they are obviously using something different.

It's interesting how the Dodgers' and Angels' Zone% in 2007 are so completely fucked
That can't be right

less than 40%?

Something clearly went way wrong

Kind of threw my 2007 calculations for a loop.

I actually have access to BIS data for THT

I’ll take a look at how the location data compares with Pitch f/x.

You must Login with your SB Nation account and be a member of Lookout Landing to post a comment.