THE BOOK--Playing The Percentages In Baseball

It looks like Bannister is not just following sabermetric theory, he is actually leading it.  He is correct that the difference in BABIP by pitch count is significant even when differences in batter quality are accounted for.  The problem he faces is that not many pitchers have had success in controlling the pitch count.  But at least he has defined the problem correctly, which many so called sabermetricions have not done.

Pitch by pitch analysis really is the next big frontier.  How do you get to 0-1 and 1-2 counts without getting your early count strikes hit all over the park?  What’s the correct game theory distribution of pitches for various repertoires?  What’s the best mix of pitching from your strength and pitching towards hitter weakness?  Very interesting questions that we now have the data to investigate.

Sky/2: You actually did good work on Adam Dunn, which I linked to here:
http://www.insidethebook.com/ee/index.php/site/comments/hitting_by_count/#14

***

And this chart here should be required posting and reading:
http://www.insidethebook.com/ee/index.php/site/comments/pitch_analysis_of_eric_bedard/#17

The big thing I’m missing in the chart is the frequencies of each.

***

After that, a breakdown by pitch type thrown would be ideal.

This way, Brian Bannister will know the run value of each of his pitches at each count… under the very large provision that the opposing hitter may be aware of this information and may tailor his approach to this new knowledge.... under the very large provision that Bannister knows that the hitter knows, and will tailor his approach and the batter’s approach.... under the very large provision that.... you get the idea.  Poker. Game theory.

Here is something that was in a Peanusts strip once. Charlie Brown is talking to himself:

Frame 1: “This guy will never be expecting a fastball...”

Frame 2: “With the bases loaded he’ll be expecting a curve. But he also knows I know what he’s expecting...”

Frame 3: “So if he’s expecting me to pitch what I know he knows I know he knows he’s expecting...”

Frame 4: “Where was I?”

http://youtube.com/watch?v=3EkBuKQEkio

“Are you the sort of man who puts the poison into his own glass?”

I’m hoping stuff like this will be automated by the likes of Josh Kalk or Mike Fast soon.  How often does a batter swing at 1-0 pitches?  How often does he get thrown 2-1 strikes?  How well does he hit curveballs ahead in the count versus behind?  How often does he see curveballs followed by inside fastballs?  Does order of pitches matter?

Sure, there’s sample size issues, but there’s got to be some smoothing possible.  I have to believe an organization like the Indians is all over this, no?

Wow. I’ve never seen such a smart ballplayer. At least, I’ve never seen a smart player be so open about what he thinks. Greg Maddux is supposed to be the smartest pitcher, but in his interviews he comes off as an “aw shucks” guy who doesn’t really want to divulge his knowledge (I don’t blame him). In theory, Bannister is right about the way to beat DIPS BABIP--control the count. That means throw strikes. But haven’t some of you noticed that pitchers who have good stuff, but concentrate highly on not walking batters, seem to have a higher BABIP (just a casual observation--might be wrong).

Checking now…

Using my database of 728 pitchers with at least 1000 TBF since 1993, the average BABIP of the 30 guys with the fewest walks allowed per TBF is: .297

And the 30 guys with the most walks has: .290

There may be something to it.  You’d have to control for K rates as well, along with starter/reliever.

There would be some selection issues with that data.  If you have a lot of walks AND a high BABIP you are not as likely to be getting 1000 TBF.

tango/8: if you walk more guys it means you’re throwing more balls and you’re living on the edges of the zone (or outside), thereby throwing the ball in microzones where BABIP is lower. Didn’t Joe Sheehan start to look at something like that (people who’d rather walk a hitter than throw a meatball)?
If you’re a guy who’s control-challenged, and therefore both walk guys and throw meatballs, then I believe Peter’s point would apply.

I don’t recall all studies by heart, but I believe I saw somewhere that hitters are more likely to take a breaking ball very early in the count, thereby giving up a strike (assuming you’re able to locate the ball) and putting the pitcher in a better position. Joe Sheehan’s last article at Baseball Analysts shows that hitters just won’t swing at a first pitch curveball. If you get ahead, then you can pace the AB from there.

Moreover, I’m actually taking a look at pitchers who beat league-average BABIP, and the first thing that jumped at me while looking at players like Rogers, Timlin (this season), Litsch and Glavine is that they throw in weird patterns. Rogers and Glavine throw tons of changeups. Litsch and Timlin are “sliderballers” (they throw more sliders than anything else, in fact they throw very few fastballs/sinkers overall, just enough to keep hitter honest). Those pitches inherently produce a lower BABIP than others (on average), and if appropriately mastered and controlled, they’ll probably produce a better-than-average BABIP if thrown way more often. Of course I don’t have the time right now (I’m working at other stuff) but I would love to see a BABIP-breakdown between pitchers who throw more and less than league-average FBs. I bet that guys who throw less fastballs (including sinkers) and more breaking balls (and changeup) will have a vastly better BABIP. My assumption is that if you can control a curveball and have one that you can throw for strikes (early in the count), even with marginal stuff you can get the count in your favour and then help your BABIP by getting the hitter to guess.

Pizza Cutter takes a look at the Bannister Theorem:
http://mvn.com/mlb-stats/2008/01/31/is-brian-bannister-on-to-something/

I commented:

Granted all that, what happens if you take the MLB BABIP at each count, and apply that to each pitcher’s frequency at those counts?

That is, let’s say that Curt Schilling is the greatest pitcher in terms of getting into a pitcher’s count. His frequency distribution at the 0-2 count might be say 10% if the MLB level is 5% (all numbers for illustration only). Let’s say then that you apply the MLB BABIP level to his 0-2 count frequency.

Do this for all pitchers at all counts. What’s the resulting BABIP for all the pitchers? That is, how much does one side of the equation (controlling the frequency of counts) affect the overall BABIP. I will guess that 1 SD will be less than .005 points.

These next 3 posts were written by Rene.  For whatever reason, the blog software keeps flagging it as spam.

Rene:
I realize how obvious are my conclusions from my last comment regarding less fastballs = lower BABIP. It’s obvious that throwing your best BABIP-pitch will lower your BABIP and fastballs will tend to be the worst BABIP-pitches of a pitcher’s arsenal.

Rene:
I only wanted to say (and poorly expressed) that I believe that most XXX challenged guys (Jeremy Bonderman comes to mind) fall into poor pitching patterns. Bonderman has a terrible fastball (looking at PitchF/X data, he just doesn’t get swinging strikes and hitters kill it when they put it in play) but still uses it a ton. Pitchers who use secondary pitches (when they have reasonably effective ones) more often will beat XXX more easily.

Rene:
I believe that too many pitchers have similar BABIP just because they fall into similar pitching patterns regardless of what their true abilities are. Just because they have to “establish the fastball” and other conventional wisdom stuff. Occasionally you’ll have guys who “get it”, like the four I mentioned in my previous comment, and they’ll beat BABIP even with extremely marginal stuff.

One of Rene’s paragraph’s had too many BABIP I guess.  I changed it to XXX and it took.  Very weird.  Is BABIP some sort of drug, or is the software coder not a fan of Voros?

Tango - I would suggest that you do your study both ways.  Also give each pitcher the major league average count distribution and his on BABIP on each count.  The standard deviation will be greater than in your proposed study.

Good idea, back as well.

Cross-posted from StatSpeak, my response:

Tom, I just ran the analysis you suggest. (2003-2006, min 200 BIP) All expected BABIP’s were between a range of .295 and .300. Standard deviation (N = 929) was .00068.

Further I ran a correlation between actual BABIP and projected BABIP, done in this manner. The result was a measly r = .065.

The effect size is probably pretty small, but it’s there.

It would seem therefore that Bannister, while technically correct, will come to the realization that there’s little to be gained here (on the BABIP side).  Even given the best pitcher in terms of working the count (who is he by the way?) compared to the worst pitcher in terms of working the count, there’s just a smidge of difference on balls in play.

In 2006, it was Mariano Rivera.  At least according to this framework of expected BABIP, which I suppose penalizes guys who get to a 0-2 count and then strike the batter out rather than let him hit the ball in play.

I didn’t run 2007.

I believe that Bannister is finding life to be good with him without fully understanding how and why he’s controlling BABIP. I really maintain that’s it’s all just about throwing secondary stuff. Voros did find out that the only pitchers who seemed to constantly control BABIP were “trick pitchers”, such as knuckleballers. The reality is that whoever relies heavily on non-fastballs is going to achieve different results from “ordinary pitchers” (who tend to fall into relatively similar patterns with heat and secondary stuff). Here is the average BABIP for pitch type this season according to my data:
Knuckleball .275
Change .290
Slider .290
Curveball .298
Fastball .307
Cutter .309
Sinker .319
Splitter .336

So why can knuckleballers beat league-average BABIP? Because they use different patterns from most pitchers, throwing more efficient pitches. I mentioned Rogers, Glavine, Litsch and Timlin as examples. The first two throw tons of changeups, while the last two throw tons of sliders, and they had a great BABIP last season.

So why do pitchers’ counts yield better results? Because pitchers are more willing to go to their secondary stuff, while in hitters’ counts they’re more likely to stick with fastballs/sinkers (whatever their main heat is), which are worse pitches on average. If pitchers were willing to pitch “backwards” and throw in “random” patterns regardless of the count, they wouldn’t really notice any difference by count in terms of BABIP and there would possibly not be such a thing as a hitter’s count anymore.

So why don’t pitchers throw more breaking balls on 3-1? We know that in hitters’ counts most pitchers (if not all) are going to throw more fastballs, and we generally assume that it’s because they can’t afford to walk the hitter (walks are obviously bad), so they’re just going with the pitch they control best. But what isn’t really known is that on average, there isn’t a whole lot of difference in strike% (control) between fastballs and breaking balls, at least not nearly as much as commonly perceived. Again, according to my data:
Cutter 65.1%
Knuckleball 65.0%
Splitter 64.1%
Sinker 64.0%
Fastball 63.8%
Slider 63.6%
Change 61.5%
Curveball 59.6%
So sliders are as easily controlled as fastballs, on average. Quite obviously, we shouldn’t care about “average” trends, but we should look at individual pitchers. There are pitchers who control their breaking balls better than their fastball (or just as well), but are still going to throw their heat, just because they are taught to do so in a hitters’ count. So take my favourite whipping boy, Jeremy Bonderman. According to Josh Kalk’s data, he throws 58.3% FB, 36.5% SL and 5.2% CH. His FB has a BABIP of .354, his SL has .333 and his CH has .300. But let’s take it a step beyond: when hitters swing at his pitches, they generate a .321 slugging average against his fastball, .260 against his slider and .400 against his changeup (quite clearly a poor pitch, I would say). This makes his slider his best pitch. Now, his strike% for his fastball is 66%, while it’s 65% for his slider. Not a huge difference, right? It would seem that his control is roughly the same. He should be able (and willing) to mix those pitches up in every count. Actually, he should be more willing to throw his slider in “dangerous” counts, since it’s more effective when hitters swing at it.
Well, Mr. Bonderman, with Pitch F/X working, has had 17 3-1 counts. Sixteen times he has thrown a fastball. Does it really surprise anybody that during the season Bonderman has conceded 1.118 SLG in 3-1 counts? I mean, his fastball is a poor pitch (poorer than the slider anyway), he HAS to throw a strike and the hitter knows it’s coming. Given that he has to throw a strike, he could at least throw his best pitch (which he controls as well as the fastball) a little more. He MIGHT walk a couple more (and I’m not even sure about that), but at least he might not serve up 5 XBH (3 HR, 2 2B) on 17 BIP.

Joe Sheehan studied pitchers whom he assumed were more willing to walk people rather than serve meatballs, but the real difference was in the repertoire: they tended to use more offspeed stuff. They were not afraid to throw better BABIP pitches, getting better results. These pitchers get it: if you can control your breaking ball or changeup as well as your fastball, then there is no actual reason to throw the fastball in hitters’ counts. In fact, it’s possibly a bigger reason to throw a changeup or a slider or a curveball. You simply need to stick to your best pitches, which is not what Bonderman is doing.
In some cases, it’s easy: Chris Young has the fastball as his best pitch and he throws it a ton. In some cases, it’s not so straightforward, but take Tom Glavine, a famous BABIP-beater. He throws his fastball less than his other pitches (he throws it as much as his changeup, and obviously less than changeup+slider). His fastball allows .360 BABIP and .248 slugging average per swing. His changeup allows .262 BABIP and .197 SLGSWG. So it’s pretty obvious why he relies so much on it, right? The only two counts in which he throws more fastballs are 0-0 and 1-1. In the others, it’s either very even or he throws many, many more changeups. For comparison’s sake with Bonderman, on 3-1 (according to Kalk) he has thrown 19 changeups and 12 fastballs. That is: he relies on his best pitch more, and that’s despite controlling his changeup (55% strike% as opposed to 59% with his fastball) much less than Bonderman does with his slider both in absolute terms and in proportion with his fastball. He doesn’t give in to hitters’ counts and still goes with his best stuff.
So, all this to say a really simple thing: pitchers should ALWAYS go with their best pitch, even in hitters’ counts, and they’ll get better results. And I know that while “trusting your best pitch” is something which sounds incredibly obvious, some simply don’t do it. After all, it’s also obvious to see that fastballs and sinkers yield a worse BABIP and therefore, unless you have a special one (like Chris Young, who also benefits from that park holding tons of his flyballs), you’ll be better off by throwing many, many more other pitches, but this doesn’t really happen. I just don’t understand the “establish your fastball” stuff. You don’t really need an overpowering and frequently thrown fastball to succeed. You can easily just have one that keeps hitters honest while surviving with your secondary stuff. It’s depressing to see a guy with apparently great talent like Bonderman perform worse than his DIPS ERA every season and never work out his presumed talent, but he’s clearly hurting himself with pitch selection. Not everybody can be as smart as Glavine or Rogers (it’s also a byproduct of experience) but I think a lot of blame should go with the mindset that is created that implies that fastballs are much easier to control and that should be thrown more often than other pitches. Here are some “slider extremes” (pitchers with at least 500 pitches seen by Pitch F/X):
Michael Wuertz 58% .262 BABIP on sliders .705 DER
Carlos Marmol 53% .255 .742
Justin Speier 50% .259 .767
Relievers, I know, but the point I’m trying to make is that those guys who have above-average sliders (I’m using sliders, but it goes for non-fastballs I guess) and aren’t afraid to throw them more than their other pitches will end up getting good results, even though those percentages likely go against conventional wisdom.

Man, I feel like I wrote a ton and didn’t really say anything we didn’t know

Rene - It is tempting to think of strike % as control but it really isn’t.  If pitchers use their fastballs to pitch to locations nearer the margins of the strike zone (and its pretty clear that they do) than they have less of a margin of error before it is out of the strike zone.  Plus one of the main advantages of the non fastball is that it is that it is slower than a fastball so that it disrupts timing resulting in strikes even if its location is not well controlled by the pitcher.  The reason that these pitches are not thrown as much on batters counts is because the batter has the option not to swing on those counts, so the pitch has to be in the strike zone to be of benefit to the pitcher.  On a pitchers count the pitcher has all the options and the batter has none.  The pitcher can throw any pitch in his arsenal and can aim for a difficult to hit location as well.  The batter must be prepared for all the pitcher’s pitches and does’t have the option of letting a pitch go by near the edge of the strike zone that may end up being a called strike.

And you can’t use the BABIP to measure the effectiveness of pitch types because the BABIP is reflective of when in the count those pitches were used as well as their inherent effectiveness.  There is also the factor of to whom they were used (the quality of the batter) which you haven’t considered.  There is also a problem of self selection.  The quality of a pitchers “secondary” pitches is likely to vary more from start to start.  A catcher should recognize early on in the game when the pitcher has his best stuff and when he doesn’t.  When he doesn’t, the catcher is likely to call fewer of those pitches.  So the secondary pitches are likely to appear more successful because they are only called when they are working well.  It doesn’t mean that the pitcher has an option of throwing more of them when they aren’t working well.

Pizza: Ah, Mariano Rivera.  The answer to all questions.

You are absolutely correct that the K and BB will let you get out of the BABIP data.  But, the main point is the degree to which you can control the count in terms of using that to lower your BABIP.  Regardless of whether it was Mo or any other pitcher in MLB, the entire range of BABIP, in terms of controlling for the count, is .295 to .300.  That’s quite a tight range.

Another way to analyze the data is from here:
http://www.insidethebook.com/ee/index.php/site/comments/pitch_analysis_of_eric_bedard/#17

You apply the linear weights figures from an average pitcher to each strike, ball, and inplay at each count.  You can use whatever combination of frequency of pitcher or frequency of league, rate of pitcher, rate of league, etc, to give you the breadth of whatever it is that you are really looking for.

That chart in post 17, more than anything, should be the bible to which MLB pitchers should refer to.  That, and what their personal post17 looks like.

Peter: I did think of many of those things actually, but I didn’t want my post to be too long. Let me reply concisely.
Strike% is actually control. I believe what you are referring to is command. Command is fundamental, obviously and I don’t expect the majority of pitchers to be able to paint the zone with their breaking balls. But I do expect the majority to be able to throw at least one secondary pitch for a strike. So, while I absolutely do agree that when they need a pitch on the corner they might go to their fastball for reasons other than the ones I’ve mentioned, it also happens that pitchers use fastballs when all they need is a “get me over” strike.
As far as BABIP goes, it’s just the tip of the iceberg really. The discussion started there but the whole point is to “pitch better”, right? That’s why I mentioned slugging average per swing. When you combine control with “damage done” every time a hitter swings at a pitch, you can at least start evaluating a pitch. I mean: if I throw strikes with a certain pitch, and when you swing at my pitches you’re not hurting me, then it’s pretty clear my pitch is quite good. Command ends up being quite subjective because regardless of our perceptions, what we care about is getting outs. If for some weird reason I can throw strikes (even in the center of the zone, whatever) and get outs, does it really matter which part of the zone I’m hitting?
As I said, I didn’t do a rigorous study of all the pitchers in the Majors (I really didn’t have time), but I did bring an example such as Jeremy Bonderman, whom I hoped could just spark a discussion (and maybe prompt someone to do that rigorous study).
I find all your objections noteworthy, but regardless of this, I really can’t find a reason for a pitcher to throw one pitch in 16 out of 17 occasions. Self-selection, secondary pitches not working, small sample size, anything, but 16/17 smells like a clear decision-making process, such as “on 3-1, I’m going with my fastball”.
As far as strike-throwing, I showed that he doesn’t have much difference between fastball and slider. As far as effectiveness when hitters swing, the slider is much better. Therefore, there is no reason to be that extreme, especially when results are really not supporting the strategy.
Bonderman has been touted as a great prospect and a guy with great stuff. And yet he has only had one season with ERA+ above 100. And yet he constantly underperforms his DIPS ERA. And yet if you just look at his performance, so far he has been a poor pitcher despite great stuff and despite not having horrible control. He is quite simply being knocked around.
On the other hand we have pitchers like Glavine who have much more marginal stuff at this stage of their careers, but throw their secondary stuff much more. Glavine walks more hitters, but he’s not concerned with that since he limits the damage when hitters swing. And he does that by throwing his best pitch in terms of slugging average conceded by swing. If the price to pay is an extra walk, he’ll take it. After all, it’s one base, while quite often Bonderman will give up extra base hits by sticking to his poorer pitch. As I said, Joe Sheehan studied pitchers who are more willing to throw their secondary stuff (or are more willing to walk a couple of extra hitters) and found out better BABIP. The gains in terms of BABIP outweigh the damage of extra walks, apparently.
Aside from this, however, I guess that I believe that pitchers like Bonderman, for whom things are clearly not working, should at least try changing their ways. At worst, they’ll still not be working, right? And for guys like Litsch or Rogers or Glavine throwing much more secondary stuff actually works.
As I said, it would take a rigorous study. I took a look at individual pitchers, so group studies might contradict what I’m claiming, but so far I’ve seen success in those pitchers who are throwing more often their best SLGSWG pitch (again: the pitch which surrenders the lowest slugging average per swing). I promise that if I have time I’ll try to dig deeper in the data and prove what I’m saying (I also have to figure out how to combine SLGSWG and Strike% exactly), but for the moment I’ll stand by my claim that in order to be more successful you need to throw your best pitch (in terms of strike% and SLGSWG combined) more often, until its performance starts decreasing to a point where you’re breaking even in terms of performance with your other pitches.

I’m deeply sorry for not being able to be more concise… I’ll work on that!

Tango - If what you were asking in post #19 was who threw the highest percentages of his pitches while he had a pitcher’s count and not a batter’s count, I did some research on that last year.  For 2006 and pitchers that faced at least 500 batters the top 5 are:

----------------BFP----PC-----BC---PC%
Mike Mussina----796---1736---643---73
Curt Schilling----818---1862---707---72.5
Johan Santana--908---1992---758---72.4
Brad Radke------678---1320---522---71.7
John Smoltz-----945---1947---791---71.1

Clearly if you are able to keep the count in your favor your odds of being a successful pitcher are very good.  But thats not really news nor does it mean that these pitchers have better than average BABIP.  Rivera faced 288 batters and had a &#xPC; of 71.5.  I probably defined pitchers counts (0-1,0-2,1-1,1-2,2-2) and batters counts (1-0,2-0,3-0,2-1,3-1,3-2) differently than other analysts, but I wanted to include all counts other than 0-0.

The count PC% range is fairly narrow.  No pitcher faced more than 150 batters and threw more pitches while behind in the count than when ahead.  The fifth from the bottom with over 500 BFP was Daniel Cabrera at 56.1%.

Rene - A meaningful research project for Bonderman would be to separate his games into successful games and unsuccessful ones.  Did he start out throwing a mix of pitches in the early innings of his successful games at about the same rate as his unsuccessful ones?  Did his offspeed pitches get hit better in his unsuccessful ones?  Did his rate of throwing offspeed pitches go down in the later innings of his unsuccessful games?  Was his fastball speed lower in his unsuccessful ones?  Was his first pitch being hit more or taken for a ball more in his unsuccessful games?

He may have a lot of games where his offspeed pitches are just not an option or his lack of success may be due to an inconsistant fastball.

Peter: you’re absolutely right. Unfortunately, we have very limited data and it may suffer from sample size. I fear that in order to carry a meaningful study of any sort on a single pitcher we may need multiple years of data, after all Bonderman was unfortunate enough to have less than 1000 pitches recorded this season.
So far the only thing I’ve noticed in pitchers who have better than average BABIP (or effectiveness rather, not just BABIP but everything together) is that they throw their best pitches more often, which, surprising as it may sound, isn’t something that every pitcher does.
If I get around to do some in-depth checks, I’ll post the results here.

I think there’s a good chance that the issue Rene and Peter are discussing could be resolved simply by looking at BABIP in different count-pitch cases - say, fastballs thrown on a 1-0 count vs. fastballs thrown on an 0-2 count.  Is there enough data available to run this sort of analysis?

Well, ok, I now have all the data regarding pitches thrown last season. Unfortunately I couldn’t get the data from MLB.com (it won’t allow connections from me for some reason) so I had to get the data from Josh Kalk’s online tool and then manually tweak the pitches his algorithm got apparently wrong (around 150-200 or so out of 1946).
I’ll take a look at the data over this weekend and I’ll try to find a pattern for effectiveness or something. Well, if there’s anything interesting I’ll share it.
I maintain that there are two important factors in judging the effectiveness of a pitch: strike% and SLGSWG (slugging average per swing). Command of the zone, unhittability and all the rest are going to be factored in SLGSWG. If you throw many strikes, and those strikes are effective when hitters take cuts, then your pitch is good.

Please note that I realize the shortcomings:
- I know that some frequently thrown pitches are going to be less effective at face value than others thrown less frequently. But right now I just want to know which pitches were more effective. I might deal with frequency at some other time.
- I know that getting a called strike MIGHT turn out to be an actual skill (such as deceiving the hitter and make him think it’s going to be outside the zone), but I believe it’s not that much of a factor when compared to effectiveness on swings and geeneral strike%.
- I know that some pitchers are going to throw their best pitch a lot less because they’re willing to use it only when its value is really high (take Papelbon, who really uses his splitter only with 2 strikes in order to get the hitter out without having given him a previous look at the pitch). I should use the run value of every single pitch, but one step at a time. Plus right now I don’t have splits (by count, by hitter handedness), just overall pitcher data.
- I know there might be all sorts of reasons for whatever the findings will be, but I’ll deal with those after I get my results. As I said, I’m really just concerned with combining strike% and SLGSWG right now.

I hope I’ll have time, and I hope I’ll get worthwile results…

I had a personal assumption that the percentage of called strikes isn’t really indicative of a pitcher’s skill, since I have a hunch that the called strike% will regress to the pitch’s league average in future years. Of course that’s just a hunch, so I figured that instead of working with SLGSWG I should use SLGSTR (slugging average per strike thrown), so whatever ability does a pitcher have in painting the zone and getting more called strikes would be factored in.
The huge problem with this is that I should calculate the run value of every single strike in every single game situation, based especially on counts. I can’t work with the average value of throwing a strike, because of course pitch selection highly depends on count, batter handedness and game situation. Since I don’t have that kind of data, I figured out that whatever I did would fall short of my goal anyway.
So I decided to just have some random quick and possibly meaningless fun and see what came out of it. Hey, OPS isn’t great but it correlates pretty well with offensive skills. I first calculated SLGSTR+ (not park or league adjusted, but just opposed to the mean) for all pitches with at least 100 instances. The top 5 leaders and trailers:
1. Fausto Carmona (slider) 186
2. Carlos Zambrano (slider) 183
3. J.J. Putz (splitter) 183
4. Manny Acosta (slider) 182
5. Wandy Rodriguez (curve) 182
...
887. Edwar Ramirez (fastball) -23
888. Scott Kazmir (slider) -23
889. Jose Capellan (fastball) -30
890. Dave Bush (change) -49
891. Jorge DeLaRosa (change) -52

Obviously this hinges on the quantity of strikes thrown, so without too much thinking (since, as I said, the “average” value of a strike doesn’t make much sense to me) I just multiplied this number by strike%:
1. Bobby Jenks (slider) 136
2. Matt Capps (fastball) 122
3. Rafael Betancourt (fastball) 122
4. Manny Acosta (slider) 119
5. Jonathan Papelbon (fastball) 119
Once again, the effectiveness of a pitch highly depends also on how often a pitch is thrown, so I just multiplied this number further by the percentage of use:
1. Rafael Betancourt (fastball) 104
2. Matt Capps (fastball) 97
3. Jonathan Papelbon (fastball) 89
4. Mariano Rivera (cutter) 87
5. Troy Percival (fastball) 86

Those numbers mean absolutely nothing as they are. I was just having fun, as I said, but I would love seeing someone who actually has detailed pitch data work out the run values of pitches (depending on the count and game situation) and then combine that with the pattern of use and see if anything meaningful arises. As I said, I don’t have that kind of data, but it was fun to just do random stuff and see some of the best pitches in baseball end on top anyway. Aside from Kazmir’s slider (which apparently sucks), the results are pretty effective, which means (I assume) that once the run values are calculated appropriately and the math is worked out, a combination of how many strikes are thrown, how effective those strikes are and the percentage of use, should be pretty effective in determining the best pitches around. Groundbreaking, I know.

Aside from all this, when I see that Carmona’s slider has the best slugging average per strike, I really believe that it’s a pitch that should be thrown more often. I know that throwing a pitch more often will reduce its effectiveness (which really says a lot about Mariano Rivera’s cutter, which he throws 95% of the time and is still unbelievably effective), but pitchers who are having little success (such as Bonderman, once again) should seriously consider throwing a lot more their best pitches. I’m not going to complain with Papelbon’s splitter or Carmona’s slider, since they’re having good success with their current patterns, but Bonderman and others?

Well, it’s all moot. I wanted to do more and was set to work out things seriously, but I’m lacking the necessary data. Too bad, we’ll see next time.

I don’t know what happened to my previous post. Anyway… I took the 10 qualified pitchers who had the best FIP-ERA and the 10 who had
the worst, according to THT data.
I calculated the SLGSWG+ of the 2 groups. The ones who outperformed their FIP have a SLGSWG+ of 99.2. The ones who didn’t, have 80.0.
Now take a look at a few pitchers who, according to the conventional crowd, have great talent, but whose results hasn’t matched that talent so far: Belisle, Bonderman, Willis, Felix Hernandez, Bush, Francis. For every pitcher I show: type/percentage of use/strike%/SLGSTR+

Bonderman
fastball/58.3%/66.0%/45.5
slider/36.5%/65.0%/77.2
change/5.2%/54.9%/19.1

No question: should throw way more sliders. But he’s overrated, his pitches aren’t good when thrown for strikes. He would be more efficient by employing his slider more and possibly dropping his awful changeup.

Belisle
fastball/59.2%/66.7%/49.6
curveball/20.8%/52.9%/168.3
cutter/10.1%/67.4%/87.6
slider/9.8%/60.7%/71.0

So, he has a great curveball but can’t throw strikes with it. He should throw his cutter more though, since it’s so much better than his fastball. Would decrease the effectiveness of the cutter, but he would get better overall results. And I also think he should throw his curveball a little more anyway.

Bush
sinker/50.5%/65.1%/90.6
splitter/22.7%/68.7%/83.2
curveball/18.1%/56.9%/111.5
change/8.7%/73.0%/-48.8

Absolutely drop the changeup. He’s being killed the 8.7% of the time he’s throwing it. He should probably throw more curveballs, but there isn’t a whole lot to be gained. His pitches are poor.

Francis
fastball/57.5%/62.2%/91.5
change/28.5%/70.1%/131.7
curveball/14.0%/52.7%/121.7

His changeup/fastball looks like a Glavine combo. Glavine manages to survive by throwing an equal number of them. Francis should definitely throw his changeup a lot more, especially since he also throws strikes. He would lose effectiveness, but possibly would gain it on the fastball until he reaches the breakeven point. In pitchers’ counts he should pound with the curveball. Anyway, he’s got talent and is using a sub-optimal distribution of pitches.

Felix Hernandez
sinker/55.8%/66.5%/90.6
slider/21.2%/61.7%/122.3
curveball/13.6%/63.7%/96.8
change/9.3%/64.1%/150.1

He should throw many, many more changeups, and many, many less sinkers (that’s his worst pitch). More sliders to righties, more changeups to lefties. Basically, more secondary stuff. He could end up with three above average pitches, and underperforming his peripherals at this stage is the last thing keeping him from greatness. His stuff isn’t questionable.

I really think most of these pitchers would greatly benefit from a optimal pitch selection. I know there might be reasons for throwing a pitch more or less, but they shouldn’t have the luxury of “setting up” the curveball when its run value is at the highest, because they’re being unsuccessful. I’m not arguing with Putz’s selection of the splitter because he’s being successful and he can afford to throw it when he wants. That’s not the case for these kids though…

I’ve also worked out with the run values of the various pitches.

As I said, even though there shouldn’t be anything like an “average” strike, I still went ahead with run values. I got the run values of every event (ball, strike, single, out, HR, etc) and calculated overall effectiveness of every pitch (thrown at least 100 times) this season.
This heavily penalizes strikeouts obviously. We know that strikes and balls are not created equal, and I would truly love someone who has count data to work out how many runs every pitch actually cost or saved, rather than working with lame average values. At least the “BIP” values are good, since you can’t argue with outs or hits, so for low-strikeout, low-walk guys my values should be reliable.

I calculated three things: how many runs a pitch saved this season, with Pitch F/X working (obviously all the leaders play in fields where Pitch F/X has been working since the start of 2007), how many runs a pitch would have saved if thrown 2000 times (rate stat, comparing all pitches to the same baseline) and how many runs a pitch would have saved with the pitcher’s distribution if the pitcher threw 2000 overall pitches (counting stat basically, depending on a pitcher’s choice).

So, which are the top 10 leaders and trailers in runs saved, as a rate stat (I made the cut at 100 instances of every pitch)?
1. Jenks (slider)
2. Capps (fastball)
3. Putz (splitter)
4. Burton (cutter)
5. Betancourt (fastball)
6. Bradford (sinker)
7. Papelbon (fastball)
8. O’Flaherty (slider)
9. Rauch (fastball)
10. Percival (fastball)
...
882. Bacsik (changeup)
883. Rheinecker (cutter)
884. Dumatrait (fastball)
885. K. Wells (changeup)
886. Kazmir (slider)
887. Stauffer (fastball)
888. Capellan (fastball)
889. Edwar Ramirez (fastball)
890. DeLaRosa (changeup)
891. Bush (changeup)

But the interesting one of course takes into account the distribution of pitches, so as a counting stat, here are the best and the worst:
1. Capps (fastball) 87
2. Betancourt (fastball) 83
3. Bradford (sinker) 79
4. Papelbon (fastball) 72
5. Burton (cutter) 71
6. Percival (fastball) 71
7. Thatcher (sinker) 67
8. Dotel (fastball) 61
9. Howry (fastball) 58
10. Wassermann (sinker) 57
...
882. Buckner (fastball) -33
883. Coffey (fastball) -33
884. Stanton (fastball) -34
885. Peralta (fastball) -35
886. Gutierrez (cutter) -38
887. Olsen (fastball) -44
888. Dumatrait (fastball) -46
889. Edwar Ramirez (fastball) -48
890. Stauffer (fastball) -55
891. Capellan (fastball) -61

So if Capellan threw 2000 pitches, with his current distribution (62.8% fastballs), according to the run values of events, his fastball would be costing him 61 runs. Likewise, Capps would save 87 runs with his fastball. Rivera’s cutter is 14th, by the way. Of course, as I said, this is really incomplete because it penalizes strikeout artists, since I don’t have data by count. I could retrieve it by hand for individual pitchers and pitches, but there can’t be a comprehensive study of which is the best pitch (which pitch saved most runs during the 2007 season). I guess after 2008, with Pitch F/X working in every park, studying this stuff is going to be more accurate and exciting.

The following is from Rene:

==========================================
I took the 10 qualified pitchers who had the best FIP-ERA and the 10 who had the worst, according to THT data.

I calculated the SLGSWG+ of the 2 groups. The ones who outperformed their FIP have a SLGSWG+ of 99.2. The ones who didn’t, have a SLGSWG+ as a group of 80.0. It probably has something to do with the quality of the two groups (even though the overperformers have the fine Chico, Livan and others). And maybe it has something to do with the fact that the overperformers tend to throw their best pitches more often by optimizing pitching patterns. So I went and picked guys who, according to “conventional” wisdom have talent and yet didn’t manage to outperform their FIP. It’s tougher to do it the other way around, because critics tend to be lenient to guys who outperform their DIPS, and end up claiming it’s a byproduct of their talent.  But there are guys like Bonderman, Willis, Belisle, who apparently have talent but don’t really show it on the field. Does it have anything to do with their patterns? I know there is selection bias, but who cares at this point with all the mess I’ve already done, it’s curiosity. I checked:  Bonderman, Belisle, Willis, F. Hernandez, Bush, Francis.

For each pitcher I specify: type/Percentage of use/SLGSWG+

Bonderman

fastball/58.3%/52.6

slider/36.5%/80.4

change/5.2%/16.0

I already spoke about him. He is vastly overrated apparently, since no pitch is above average in terms of swinging efficiency. Having said this, he throws his fastball way, way more than his slider. There might be reasons for that, but since he isn’t racking up results, I think he should try throwing his slider more. It’s a poor pitch anyway so I wouldn’t expect miracles, but I would try changing something.

Belisle

fastball/59.2%/53.6

curveball/20.8%/170.9

cutter/10.1%/106.3

slider/9.8%/87.6

He has a great curveball apparently. His cutter is also above average. And yet, he pounds hitters with his fastball which is easily his worst pitch.  This is a case where I really think his patterns are hurting him despite some really good talent. I know that if he started throwing his curveball and cutter more they would be less effective, but given the gap with his fastball, the breakeven point is so far that he would get much, much better results by mixing up his pitches more and better.

Bush

sinker/50.5%/87.3

splitter/22.7%/101.2

curveball/18.1%/117.6

change/8.7%/-20.0 (yup, negative)

Leaving aside his awful changeup, which he should just drop, I also notice the fact that he is overusing his worst pitch. Once again, more curveballs and more splitters. He should be able to be at least an average pitcher.

Francis

fastball/57.5%/87.3

change/28.5%/141.4

curveball/14.0%/102.7

Great changeup, above average curveball, below average and overused fastball. There might be a Coors effect at work. If I were him, I’d still throw the changeup more often and the fastball slightly less. I think he would end up with three slightly above average pitches. He isn’t a fluke, apparently.

Felix Hernandez

sinker/55.8%/89.3

slider/21.2%/118.5

curveball/13.6%/85.0

change/9.3%/149.5

So, the King has two dominant pitches, and two below average pitches. His sinker isn’t nearly as good as people think when hitters swing at it. He would be more successful if he used his secondary pitches more often (more sliders to righties, more changeups to lefties). His secondary pitches would lose effectiveness, but his sinker would likely gain some. He would end up with three above average pitches and an average fourth one, and be really dominant. The fact that he underperforms his peripherals is the only thing keeping him from greatness.

All these pitchers throw their worst pitches way, way too often. As I said, other pitchers do something like that, but I’m not really interested in guys who are successful (and therefore might be willing to maximise the effectiveness of their best pitches by throwing them in 2 strike counts, or whatever). These guys don’t have (or shouldn’t have) the luxury of thinking about setting up their best pitches for strikeouts because they’re unsuccessful. They absolutely should change something in their approaches, and given that what they’re doing isn’t working, they should try throwing their best pitch more often. Hey, at worst they’ll keep being unsuccessful, right?

Mike Fast’s three-part look at Bannister and BABIP:
http://mvn.com/mlb-stats/2008/02/28/winning-with-an-89-mph-fastball-an-analysis-of-brian-bannister-part-3/

Thanks, Tango.  I got a nice response from Bannister on the series in the comments to Part 3.

Where can you find Clemens FIP numbers at?  For his entire career?

Fangraphs.

Bannister’s comments are here:
http://mvn.com/mlb-stats/2008/02/28/winning-with-an-89-mph-fastball-an-analysis-of-brian-bannister-part-3/#comment-1763

More about Bannister, from Joe Poz:
http://www.kansascity.com/sports/royals/v-print/story/550237.html

Poz says BABIP rhymes with crab dip. Am I the only who takes joy in pronouncing it (in my mind) baw-BEEP?

More Bannister, this time from Josh Kalk:

http://www.hardballtimes.com/main/article/anatomy-of-a-player-brian-bannister/

Our hero keeps talking:
http://kansascity.royals.mlb.com/news/article.jsp?ymd=20080604&content_id=2837801&vkey=news_kc&fext=.jsp&c_id=kc

For those with an interest in statistics, a pitcher’s BABIP, or the percentage of balls in play (not including home runs) that result in hits, are largely out of a pitcher’s control when averaged over the course of several years. A favorable batted-ball mix (the ratio of line drives/fly balls/ground balls/popups) can help lower a pitcher’s BABIP over the course of a season, but in general, it will migrate towards .300 over time. As pitchers, we can only control our strikeouts, walks, home runs allowed and groundball/flyball ratios (the latter based on adjusting our pitches and arm angles).

A good way to get a rough idea of what a pitcher is doing to improve his long-term sustainable ERA, independent of luck, is to look up his FIP, or Fielding Independent Pitching, on a site such as http://www.fangraphs.com. For example, Paul Splittorff’s career FIP was 3.72, and his career ERA ended up at 3.81, while Bret Saberhagen’s career FIP was 3.26, and his career ERA ended up at 3.34.

I think it is very important for young pitchers, especially those with a desire to play in the Major Leagues, to understand how they can make themselves more effective for their organizations and fans long-term, and not simply rely on the luck and emotions of a few games. Striking out more batters and walking less, without increasing your home run rate or throwing extra pitches per inning, is what every pitcher should strive for. Making low-contact pitches with two strikes will increase strikeouts, while throwing strikes early in the count will reduce walks and pitches per inning.

Throwing pitches that hitters consistently hit on the ground will lower your home runs allowed, and also increase your chances of getting double plays. Continue to focus on what will make you a better pitcher long-term, and don’t let the luck of the game affect your emotions in the short-term.

The reason that a pitcher’s ERA does not always match his FIP is that the timing of his hits can vary from year to year. The luck of those hits/homers are much more detrimental with runners on base, which is recorded as percentage of runners left on base, or LOB%. A common LOB% percentage is in the 70-80% range, with anything above that range representing good luck and below that bad luck.

Therefore, you can now see how your favorite pitchers, such as Zack Greinke, are improving from year-to-year. Zack is currently posting a career-best 3.56 FIP, and has the 2.88 ERA to match because of some great pitching with runners on base and an increased ground-ball rate, which has resulted in less home runs allowed.