17 March 2015

FIP and the Ball-Strike Count

Voros McCracken argued in 2001 that pitchers have little ability to prevent balls in the field of play from becoming hits and therefore a pitcher’s ability to record strikeouts, avoid walks, and limit home runs separates good pitchers from bad ones. He argues that one reason for this is that there’s a significant lack of year-to-year correlation in BABIP.

One of the questions I’ve had about this is what this means for ball-strike counts. If pitchers have little impact on balls in play then presumably they should give up a similar percentage of singles, doubles, and triples on a 0-2 count than on a 3-0. This seems unlikely and unintuitive because a pitcher can’t be as selective in a 3-0 count than in a 0-2 count. Using data from Retrosheet, I built a data file with all at-bats from 2000-2013 that ended up in either a single, double, triple, home run, or out in play and determined the count when the ball was hit into play.  Below is the data.

These results do partly support FIP because as the count becomes more hitter friendly it becomes considerably more likely that the ball will be hit out of the park. Hitters hit a home run four times more often on a 3-0 count than on a 0-2 count. The idea behind FIP is that pitchers have more control over home runs than over balls in play and these results certainly back that idea up.

However, the results also show that triples and doubles are more likely in a hitter's count and singles are more likely in a pitcher's count. I think this chart is showing that pitchers are giving up harder hit balls on hitter's counts than in pitcher's counts. While I wouldn’t have expected these results, they do make sense because it’s easier to hit a lucky single than a double, triple, or home run.

The BABIP stat is very interesting because while it indicates that hitters do have a better BABIP in a hitter's count than in a pitcher's count, the impact is minimal. If one was just looking at BABIP and didn’t look at the type of hits then the differences would appear to be minor. This means that a lack of significant year-to-year BABIP is mostly irrelevant. The reason why pitchers have better results on balls in play in a pitcher's count rather than a hitter's count is because they’re giving up fewer extra-base hits and not fewer hits overall. Presumably there should be some value in giving up singles rather than doubles or triples.

It appears that this data does indicate that pitchers do better in pitcher's counts than in hitter's counts. The next question is whether certain types of pitchers are more likely to allow contact in pitchers counts rather than hitter's counts or whether they’re able to give up lighter contact. Basically, if we can identify pitcher groups that give up weaker contact than other groups then this would potentially show a flaw with FIP. Alternatively, if all pitchers allow contact at the same rate for each ball-strike count then what I found above may be interesting but is purely academic.

I’m not sure which metrics are the best to predict ball-strike count or the type of contact. The two metrics that I did use to see whether this is the case was K% and LD%. If pitchers with a high K% strike out a lot of batters then it seems possible that they’ll face more pitcher-friendly counts than those that have low K%. If so, logic would indicate that they’d give up more contact in pitcher-friendly counts. Likewise, pitchers that give up fewer line drives than average should probably give up lighter contact than those that give up more line drives than average. I’ll start by looking at K% and the data is below.

The data is interesting. It indicates that pitchers with above-average K% do considerably better in hitter's counts than those with below-average K%, and it would be interesting to figure out why that’s the case. It’s possible that it's primarily due to small sample size. However, all in all, pitchers with below average K% give up .03% more singles, .11% more doubles, .01% more triples, and .04% fewer home runs. Furthermore, this next chart shows how often they give up contact for a given ball-strike count.

Pitchers with below average strikeout rates do give up slightly more contact in hitter-friendly counts than in pitcher-friendly counts. The problem is that the difference is minimal. Simply put, this metric doesn’t show that pitchers with high K% either give up lighter contact or pitch in significantly more pitcher-friendly counts than those with low K%. Next up is LD% rates and below is the data.

Basically, the data show that pitchers with above-average LD% rates (lower than average) give up fewer singles, doubles, and triples than those with below-average LD% rates while allowing the same amount of home runs. This suggests that FIP may be improved by looking at the type of contact that a pitcher gives up but ultimately isn’t helpful for my purposes. The chart below shows that pitchers with above average and below average line drive rates allow contact at roughly similar rates for each unique ball-strike count.

Since I already looked at BABIP before realizing that it wouldn’t be helpful, I suppose it makes sense to discuss that quickly as well. Below is a chart with data.

Basically, it also shows that pitchers with good BABIPs allow roughly 18% fewer singles and doubles as well as 12% fewer triples than those with bad BABIPs. Pitchers with good and bad BABIPs also allow contact at roughly similar rates for each unique ball-strike count.

At this point, I’ve shown that pitchers have control over primarily their home run rates but also balls in play based on the count. However, I haven’t been able to show that there are certain types of pitchers that are able to be successful because they are able to minimize contact allowed in pitcher's counts. This could be for any of at least three reasons.

The first reason is that I simply could have picked poor metrics. Just because the metrics that I used to see whether certain pitchers minimize contact in pitcher's counts didn’t do that doesn’t mean that other metrics won’t be more successful.

The second reason is that this could be a fringe skill. It may be the case that only a few pitchers are able to consistently avoid hitter's counts and therefore give up fewer extra-base hits. If so, perhaps looking at only 10% of the population instead of 50% of the population will provide better results.

The third possibility is that pitchers aren’t able to control when they allow contact and therefore noting that avoiding hitter's counts reduces extra-base hits is merely academic.

Next week, I’ll look at this question with a different metric and see whether that changes the results.


Mike said...

Very interesting article. I do think however that it is a bit of a dated perspective about pitchers having little "control" (effect is probably a better word) on the quality of hit balls. We should expect the location of a pitch to impact the quality of a hit and counts impact pitch location quite a bit. Other research has also shown that velocity (as well as a couple other things like spin) impact hit quality as well. I think this work fits well into the existing body of research.

realist said...

Interesting post, great analysis showing a higher percent of all XBH's in hitters' counts.

That said, I've noticed a mistake in the BABIP calculation, at least in the first chart. Your numbers still include HR's in the denominator for BABIP rather than excluding them entirely. I believe that, after fixing this formula, we will will then see somewhat higher BABIP in hitters' counts. For example, BABIP on 3-0 counts should be .313. BABIP on 0-2 counts should be .285. (I only calculated a couple because I didn't want to spend the time to 10-key all of the numbers in the chart.)