February 6, 2001
From The Mailbag
Special Edition: Pitching and Defense
I wonder if you have looked at whether there is evidence that pitchers can "control" extra-base hits. By this, I mean, are there pitchers off whom batters generally do not hit the ball as hard or hit the ball on the ground? If there is no correlation between a pitcher and "giving up" hits, is there a correlation between a pitcher and the kind of hits he gives up?
The short answer is that you probably won't find anything there. Pitchers may differ in this ability--they probably do a little just as they probably do a little on hits on balls in play--but it will be difficult to find it on a case-by-case basis.
However there are promising leads if we want to look at pitchers in groups.
There is one "variable" that you did not address that I am curious about: the difference between ground-ball and fly-ball pitchers. I am not suggesting that there is, or even should be, a difference between these groups; rather, I am very curious about whether there is any pattern.
One of the funny things I read in some letters was the argument that ground-ball pitchers should give up fewer hits on balls in play. Funny, because I frequently got the argument that fly-ball pitchers should give up fewer.
I've wanted to take DIPS in this direction for some time now. However, this is fraught with all sorts of difficulties. Checking last year's splits against ground-ball and vs. fly-ball pitchers, the (H-HR)/(AB-HR-SO) (a hitting equivalent of DIPS) rate for fly-ball pitchers was .293, the rate for ground-ball pitchers was .305.
This deserves more digging, much more than I can do here. The difference between an extreme ground-ball pitcher like Mike Hampton and an extreme fly-ball pitcher like Denny Neagle is only about 136 plays. In order for that to amount to that entire 12-point difference, the rate at which ground balls must drop in has to be about 5% higher (or 51 points of batting average) than it is for fly balls. If I recall correctly, the rate is lower than that, only about 2%. If that's true, there needs to be another explanation for the rest of the difference.
Also it must be mentioned that for pitchers with 600 BIP in a season (a fairly large number in this day and age) that only amounts to around seven hits between fly-ball and ground-ball pitchers in a season.
And in fact, GB-to-FB ratios are not exactly defense independent. There are a number of balls in play--mainly line drives--that are judged as neither ground balls or fly balls. There must be a certain number of plays where judgment of what a line drive is and what a fly ball is difficult to ascertain. This is also done after the play is finished. Whether the scorer saw the ball land for a hit or be caught must have some effect on the call. If part of the determination on who winds up being a fly-ball pitcher and who doesn't is dependent in some way on how many hits in play they give up... I shouldn't speculate too freely in this area, and it is not a major issue, but I think it is one worth mentioning.
Finally, and most importantly, any advantage fly-ball pitchers do have in this area seems to be mostly canceled out by the extra frequency with which these pitchers give up extra-base hits. So while GBers may tend to give up slightly more hits in play than FBers, a smaller percentage of those hits (around 2%) go for extra bases. The net result in a pitcher's value would have to be pretty small.
Lots more needs to be done here. Unfortunately, data is often hard to come by.
It would seem to me that it is not just the strikeout and the base-on-balls that is isolated from the defense. It is their base components--the strike and ball. I believe you will find that a pitcher who throws more strikes than balls will get ahead of the count more and I think that there is a correlation between the count and AVG/OBP/SLG percentages. So a pitcher should have the ability to limit hits/balls in play if he gets ahead in the count more often than behind in the count. Is there a flaw in my logic here?
The hits per balls in play rate in the AL (we want to get the pitchers out of this one) in 2000 when pitchers were ahead in the count was about .015 lower than when they were behind.
Sounds fairly significant, doesn't it? The problem is that our perception of how often some pitchers are ahead is greater than the reality. Keep in mind that it's not just the difference between how often the pitcher is ahead, but how often the pitcher is ahead AND the ball is put into play.
To put it simply, if you do the math on this, you'll find that if we assume that .015 point difference is real (it might not be; the difference may be due to better hitters being ahead in the count more often), the difference between Pedro Martinez and Dan Reichert in an entire season comes out to a little less than two-thirds of a hit per season.
The huge differences in the behind in the count and ahead in the count numbers are due mostly to the fact that there were no strikeouts when behind in the count, and whole bunches when ahead. Home runs also have an effect there.
Two things. I was wondering whether there is a difference in the stat for knuckleballers, who intuitively (though intuition is not always correct) cause players to hit off balance and therefore weakly.
Two good questions, and one embarrassing one. I thought for sure when I wrote the article that I mentioned what I like to call "trick delivery" pitchers. It is obvious I didn't.
There certainly seems to be good evidence that knuckleballers and extreme submariners have lower rates than most, and logically this ought to be the case. When I did the career rates of active pitchers divided by their team averages in 1999, number one on the list was Tim Wakefield. I wrote several times in the past that "trick delivery pitchers may be an exception here, though there are sample-size issues. In any event, if someone wanted to disregard this for guys like Tim Wakefield and Scott Sullivan, I wouldn't argue with them."
The second question is more difficult to assess and I've really only offered what I felt was the best solution, which was to completely reconstruct the pitching line so that we can use weights that have already been determined before. I'd also add that HR rate probably shouldn't be taken lightly. It is a key component in the success of pitchers like Tom Glavine and Mike Hampton, and also what keeps John Wasdin from having sustained major-league success.
It strikes me that this is an area that would be hugely impacted by ballpark effects, but I haven't seen any analysis of that.
Ballpark effects would certainly come into play, though they would affect Voros's original study more than my followup. I was comparing pitchers' ball-in-play rates to their team's overall average balls in play rate. Since the team as a whole has the same home ballpark, comparing the results from one pitcher to essentially all the other pitchers on the team minimizes the park effects.
I am curious about whether you might find a trend for teams, rather than individual pitchers. If there is no particular trend for a team, what does that say about the +/- effect of a team's defense? Rey Ordonez, or Geoff Blum? How many batted-ball outs (Ordonez) are converted into hits (Blum)? Why not just let Chipper Jones play shortstop?
While most of the discussion has been focused on whether individual pitchers have varying abilities to affect ball-in-play average, there is clearly more work to do.
There is a significant correlation between teams' balls in play average from year to year from 1979-99 (about +0.50), indicating that there are trends for teams to consistently play above or below league average (even before trying to control for personnel turnover between seasons).
For example: the worst team average (other than for Colorado, where park factors make defense much more difficult) is the 1997 A's, a team that used Jason Giambi, Geronimo Berroa, and Matt Stairs frequently in the outfield. The best in a non-strike year is the 1982 Cardinals (the 1981 Tigers and A's did slightly better), a classic Whitey Herzog team with all-world defenders Keith Hernandez and Ozzie Smith, and where, according to baseball-reference.com, every regular position player in fair territory had an above-average range factor except George Hendrick; even Lonnie Smith had a good RF that year.
So even if teams can't rely on pitchers to generate easy fielding opportunities, they certainly can help themselves with good defense behind their pitchers. And with teams averaging around 4,500 balls in play per season, the advantage amounts to more than a hundred hits per season between typically good vs. typically bad defensive teams, and more than 300 hits per season difference between the most extreme teams.
What would happen if you did parallel research for hitters' batting average? Would the correlation (splitting players' careers into even and odd groups and using the even average to predict the odd average) be greater than +0.53? I think it would be important to do a parallel study from the hitters' side in order to assess what's really going on on the pitchers' side.
That's a good suggestion. I did a quick look at the correlation between batters' half careers for balls in play, and the result was very different than that for pitchers. The correlation from even-career to odd-career was +0.82, with a range of ball-in-play averages from around .250 to a few approaching .400. The chart of even vs. odd shows a strong clustering around a linear relationship (I may try to show the chart in a followup article). Eyeballing the standard deviation, it looks to be about 10 points, meaning that a .300 balls in play average even-career hitter has a 95% chance of having his odd-career average be between .280 and .320 (this isn't rigorous, though, just an visual estimate).
It suggests that the outcome of a ball in play depends most heavily on the batter at the plate, with lesser weight on the pitcher and fielding team behind him.
You can test the second argument from your conclusion (that long career pitchers have selected themselves into the sample because they are good at preventing hits on balls in play) by running 70 separate regressions (in the larger dataset) of balls in play ratio on dummy variables for each of the 70 players in the smaller sample. If more pitchers turned up as statistically significant positive predictors of BiP ratio, i.e., if the coefficient on each dummy was positive and significant, then you'd have evidence that their careers may have been longer because they prevented hits on ball in play. More negative than positive predictors would suggest the opposite, and rough equality would imply no difference.
Thanks for the great suggestion. I expect to do some additional followup work along these lines, and have gotten several good ideas from our readers. With luck, my schedule will allow me to have another article or two on the topic in the near future.
Voros McCracken is a student living in Chicago. He will be writing a weekly column called the "Baseball Skeptic" in an upcoming webzine. For more information on DIPS and McCracken's work, check out his Web page at http://www.baseballstuff.com/mccracken.