The statistical coaching aid

The idea of computers dictating the on-field happenings of a baseball game has been a long-professed fear of the more traditional school of baseball thought.

I believe that fear is quite misguided. Today’s managers and coaches, the vast majority of whom would fall into that traditionalist camp, are already being steered by these computers about as much would be conceivable for the foreseeable future. Advance scouting data is marked, parsed, analyzed, tested, and fed to a manager before a series.

Matchup data (regardless of its validity) and platoon statistics are collected and passed on to the field marshals. Front offices have meetings with managers and coaches about bunt and stolen base strategies, using data collected and analyzed for that purpose. Advanced video analysis software is used to help a coach explain to a hitter why he’s dragging his bat head, or to help a pitcher relocate his arm slot.

Are these methods detracting from the authority and validity of a manager or coach? On the contrary, I think they add to it. These are tools that are put into the hands of experienced, professional people, to use at their discretion in order to help achieve the goals of the organization. These tools empower managers and coaches to attack problems and make decisions in ways they weren’t able to before. With all these tools at their disposal, a manager or coach who utilizes them effectively is more important than ever, not less.

I am putting forth a prototype of a new tool to put into the hands of those coaches and managers, based on two areas of relatively recent refinement in the statistical community:

1) We know what gives a pitcher value, in terms of components (walk rate, strikeout rate, etc)
2) We have a strong idea of what types of pitches lead to what types of outcomes (example: sinkers lead to ground balls, which lead to fewer home runs and higher BABIPs).

If we have some solid information about a pitcher’s abilities/strengths/weaknesses/approach, we should be able to develop a theoretical road map of sorts for optimizing his value.

That first bit is the difficult part for us as fans (but not for an organization). The sample player I’m working with here is one of my friends currently playing minor league ball, so I was able to speak directly with him to get the information you’ll see in the questionnaire below. I also have a contact within the front office of his parent organization, who has suggested I should leave the player’s name and organization blank, just for the sake of confidentiality.

I gave my friend the opportunity to choose his own pseudonym, to which he replied “you can use anything you want, just don’t make it stupid. So in lieu of shaming him with the nickname “Mama’s Favorite” or “The Big Cuddler,” I’ll just refer to him as “Pitcher X” from here on out and leave the organization blank. (But dude, you get to throw baseballs for a living… my jealousy pretty much requires I have to take a little jab at you when I can.)

Overall Value

Name        Year  Team      IP   mle FIP  mle ERC
Pitcher X   2006  XXX      89.7    6.31    7.34
Pitcher X   2007  XXX      65.3    6.71    8.44
Pitcher X   2008  XXX      53.7    7.18    6.92

I should mention; during the above seasons, Pitcher X was ages 21-23, respectively. The MLE FIP/ERC are my own calculation. MLE stands for Major League Equivalent. FIP stands for Fielding Independent Pitching and ERC stands for Component ERA, which is similar to FIP but includes the impact of hit rate.

Obviously, the overall numbers and their lack of progress seem to leave something to be desired for the ages of 21-23. However, I feel 2008 was a considerably better season than it appears on the surface, as I’ll discuss in the component evaluation.

Component View

Pitcher X 2006 A+ 89.7 .253 .195 .069 .085 .016 .018 .662 .702
Pitcher X 2007 A+ 12.3 .122 .183 .035 .084 .017 .017 .826 .716
Pitcher X 2007 AA 53.0 .191 .197 .116 .091 .017 .019 .678 .693
Pitcher X 2008 AA 53.7 .242 .199 .127 .096 .034 .019 .596 .702

In this table, the pitcher’s stats are highlighted in yellow, while the league stats are next to the yellow boxes (for easy comparison). BF stands for Batters Faced, so each of the stats (such as K/BF) are expressed in terms of batters faced (strikeouts per batters faced).

Taking all the data into account, we can generalize Pitcher X’s components as follows:
{exp:list_maker}Strikeout abilities just barely above average (the league figures include starters as well as relievers. Most of Pitcher X’s innings have come as a relief pitcher, and if the data just examined relief pitchers, the league rates would be a bit higher)
Slightly below average control
Slightly below average HR prevention {/exp:list_maker}Now, with regard to the 2008 season, Pitcher X’s HR rate was obviously an aberration. While his ground ball rate did take a bit of a dip from previous seasons, the real culprit was that 13.8% of his fly balls allowed were homeruns, which is obviously a fluke. If we regress his HR rate to the league average (which I find fair, based on his GB/FB rates and history), the MLE FIP/ERC become 5.88/6.08.

Replacement-level relief pitchers come in at 4.75 under my methodology, and pitchers peak earlier than hitters (about age 25), so Pitcher X will certainly have to “jump” the aging curve at some point to have a major league career (the same could be said of the vast majority of pitching prospects, but Pitcher X’s jump will need to be rather large, and he’s short on time). The fact that he demonstrated a meaningfully above-average strikeout rate in the high minors last year is an encouraging sign that the aptitude is there.

Based on the component evaluation, I have listed three objectives (with their relative priority) of which some combination could send Pitcher X down a Major League career path:
{exp:list_maker}50 percent: Improve walk rate
25 percent: Improve HR rate
25 percent: Improve strikeout rate {/exp:list_maker}Consider those objectives through the questionnaire I use for pitchers, with Pitcher X’s answers:

1) What pitches do you throw, and roughly what percentage of the time do you throw them?
Slider: 50 percent
Two-seam fastball: 40 percent
Change-up: 10 percent

2) Rank your pitches in order of how you perceive their effectiveness.
My slider is plus. Fastball is average. Changeup at times is above average but overall average.

3) What pitch do you most often throw with two strikes when you are trying to strike a batter out?

4) What is your general approach to a right-handed hitter versus a left-handed hitter?
Lefties I stay hard away, change up away, late fastball in, slider backfoot. Righties I stay hard in, sliders down away.

5) In general, how does your approach differ when facing a hitter whom you perceive to have strong power?
My approach doesn’t differ. I try to dominate every hitter. By treating the lesser hitters as potent ones I don’t get hurt by them. I try to get ahead with either fastball or slider then stick to the gameplan of question four. In other words, I don’t see the hitter or get caught up in who it is, I just see the catcher and execute my pitch.

6) How is your approach different when used as a starter versus as a reliever?
Starting is more pacing and setting up hitters for the second and third time through the lineup. Relief work I focus on outs and going at each hitter aggressively.

7) Describe your command/feel for each of your pitches.
Fastball command a bit shaky. Can always hit away and in but tend to leave it up a bit when I miss. When slider is on, it takes hitters off my fastball. Slider command most of the time is good. Sometimes overthrow it which takes away from the depth but I feel I can throw it for a strike anytime I need to. Changeup command is decent. I only throw it to lefties and keep it down and away. Occasional missed spot tends to be high for a ball.

8) Please include any other information that you feel makes your approach/style/effectiveness unique amongst other pitchers.
I feel my pitching style is different from other pitchers in that I’m not afraid of contact. I pitch to contact but because of my slider get a lot of misses and strikeouts. Sometimes come into games a bit hyped up and tend to get behind sometimes which leads to walks and a big inning. Eliminate those walks and my stats would look a whole lot better.

Given the objectives and the information provided by the questionnaire, I’ll first go through each objective independently and brainstorm possible solutions. At the end, we’ll see if some common themes emerge.

Improve walk rate (50 percent)
This is the question with the fewest obvious solutions, because most of the issue seems to be related to the fastball, which is obviously fundamental. And it’s not like Pitcher X is being too fine; he says he’s trying to pitch to contact.

However, one possibility would be mixing in (or just moving to) a four-seam, for two reasons; A) He might be able to control it better, B) Most pitchers gain a couple ticks in velocity when throwing a four-seam versus a two-seam. The additional velocity might allow him to pitch more comfortably in the zone, and draw more swings on pitches outside the zone.

Also, though he claims he already throws it 50 percent of the time, it may not hurt to throw the slider just a bit more. He says he feels he can throw it for a strike at any time, and it will generate more out of the zone swings than a fastball. Over 56 percent of the pitches Brad Lidge threw last year were sliders, and that’s a guy with a mid-90’s fastball. If it’s such an effective pitch, it’s not so crazy to suggest Pitcher X throw it even a bit more often.

One more slightly unique characteristic is that Pitcher X has much more control trouble with right-handed batters (10.2 percent walk rate) than lefties (7.2 percent). That’s the inverse of the normal distribution. Even more curious is that his strikeout and home run rates are virtually identical versus either handedness. For a fastball/slider guy, being more effective against opposite-handed batters than same-handed is, too say the least, odd. Maybe he would be well advised to change his approach to RHB; perhaps he’d be more comfortable (control wise) using the fastball outside more often.

Improve home run rate (25 percent)
Considering that the three pitches he throws are all at least somewhat conducive toward ground balls, it’s curious that Pitcher X hasn’t ever sported terribly strong ground ball numbers. I think the likely culprit is that the two-seam isn”t effective towards this end (which is really the majority of the value in throwing a two-seam).

Improve Strikeout Rate (25%)
To effectively pitch in a Major League bullpen without an Aaron-Cook-or-better groundball rate, Pitcher X will need to miss more bats than he does now. He may need to reconsider backing off the “pitching to contact” approach just a bit.

Working the slider a bit more often would also help.

Another obvious potential solution would be moving from a contact oriented fastball—the two-seamer—to the higher-octane four-seamer.


1) I would suggest that Pitcher X start throwing more four-seam fastballs—instead of only throwing a two-seamer, he should throw a four-seamer at least 50 percent of the time. The two-seam isn’t currently doing very much for his home-run prevention skills, is almost certainly limiting his strikeout rate, and is possibly a part of the cause for his sub-par walk rate. Any losses in HR prevention should be at least offset by the gains in other areas.

2) I would suggest Pitcher X throw his slider even a bit more often than he has in the past. This would almost certainly improve his strikeout rate, likely improve his home run rate, and would appear to not hurt his walk rate (it may even help it).

3) I would suggest Pitcher X modify his approach to right-handed batters. For a pitcher with his arsenal, it is nearly inconceivable to me that he is more effective versus lefties than righties. The “stuff” is obviously not the cause of the reverse split; it must be the approach.

This may mean working the fastball outside more often, mixing in more changeups (maybe it’s not only effective on lefties), or switching to the more aggressive four-seam fastball.

Again, this is a prototype. This is currently the stone-age version of a hammer; still quite useful, just not optimal. New developments in information (whether scouting methods, PITCHf/x or other technology, or simple enhancement of this technique) can and will tweak this model towards greater utility. We could construct a model of the range of possible outcomes from the suggestions, based on Linear Weights values—or, we may be able to replace the questionnaire with PITCHf/x (or similar) data.

But for right now, that’s unnecessary. As the hammer’s purpose was to bash things, the purpose of this tool is to help a pitcher be more effective. The above recommendations are by no means absolute; they are theories worth investigating by those who know the situation best. This method represents a way of organizing information which, when put into the hands of people who coach pitchers for a living, ought to be able to help diagnose and attack problems more effectively than without it. For now, that will do.

Comments are closed.