Is there any predictive value to % Line Drives?

by beastball on Sep 19, 2007 12:24 PM EDT

I remember exploring this issue some last year and finding that %LD didn't really have much correlation to performance. Considering I didn't have access to that information before last year, I couldn't test if it had any predictive value, but I felt that it didn't.

Lately, I've been seeing people use that data point to argue for or against a certain prospect, and I'm curious to know why.

My question is, has anyone done further analysis on %LD or %GB to see if it really has any predictive value? I haven't see any myself, but I haven't been able to follow baseball as closely this year as I have in the past.

Tweet Comment 14 comments | 0 recs |

Do you like this story?

Comments

hardball times

did a study that showed that Line Drives become hits something like 72% of the time. Groundballs something like 24%, and flyball something like 12% (excluding home runs). I don't know if they included popups into flyballs, or was its own category, not worthy of discussion since such a tiny fraction of popups fall for hits.

by Galt on Sep 19, 2007 1:10 PM EDT reply actions

memory

all numbers from memory. Probably a bit off, but directionally correct - don't have time to google it, but I'm 99% sure it was hardball times that had the study.

by Galt on Sep 19, 2007 1:11 PM EDT up reply actions

LD%

The Hardball Times studied the run-value of different types of batted balls and found line drives to yield the highest run-value, more than twice the value of the next highest, outfield flies. I believe it appeared in the 2007 book.

I've been looking at the predictive value of LD% and the repeatability of the metric. I don't have solid conclusions yet, but it seems that LD% has a pretty good correlation to slugging and batting average, which supports THT's findings.

What is intuitively certain is that a line drive is better, for the batter, than a strikeout or infield fly. So if a player can reduce those rates and increase their LD-rate, it's certainly a benefit and a positive move in statistics will emerge.

by kschellenger on Sep 19, 2007 1:17 PM EDT reply actions

Here's one thread from last year...

and just consider, Upton had one of the two lowest %LD of the top prospect hitters of last year, and now he's the #1 prospect. Braun was towards the bottom in %LD last year also.

It just doesn't seem like %LD is a useful data point, not when predicting future success.

http://www.minorleagueball.com/story/2006/12/23/144442/10

And I agree, it seems logical that it would mean something, and maybe it's a better tool to examine current data and not something with much predictive value.

by beastball on Sep 19, 2007 1:31 PM EDT reply actions

well

That may be true over periods of like 400 ABs or whatever you've got in that other post. But who's to say that LD% over, say, 1000 ABs isn't predictive of future LD%?

As it is, the stat seems problematic to me. There's not, AFAIK, a definitive cut-off between LDs and FBs and GBs and LDs, so at the margins you're going to get some distortion. And even if there was a standard (something like "ball trajectory leaving the bat of X degrees above the horizontal at at least Y velocity"), I doubt there'd be competent measures in place to accurately observe it.

The other possibility is that, as players develop in the minors, their swings change and they hit line drives at differing percentages. Maybe minor league LD% just don't translate to the majors very well.

by mraver on Sep 19, 2007 2:47 PM EDT up reply actions

Yeah...

I was hopeful that these percentages would yield some fruit. I seem to have better luck predicting success with pitchers than with hitters, so any other way to slice their numbers to help predict success would be great.

As it is, we have to prospect using smallish sample sizes, so if a trend doesn't present itself in a year's worth of play (~600 PAs or less), then it's not particularly useful. Trending is useful when tracking progress, so these percentages might be useful in that regard, but I don't see a correlation between these percentages and present or future success.

I couldn't take a set of these percentages and have any idea how the player had performed or will perform in the future.

by beastball on Sep 19, 2007 3:04 PM EDT up reply actions

LD%

Minor league batted ball numbers for line drives seem to differ methodologically from Major league batted ball numbers. MLB hitters are credited with way more line drives than minor leaguers.

by limozeen on Sep 19, 2007 3:41 PM EDT reply actions

That could possibly be

because major league hitters are better, and therefore hit more line drives? Does this go across top prospects as well, or are you looking at pitchers' LDA numbers? It would make sense that worse prospects would have more trouble making solid contact and getting line drives.

Vice-Chairman of the Sonnanstine Underground Railroad

by Brickhaus on Sep 19, 2007 6:46 PM EDT up reply actions

no

It's for all prospects.

by limozeen on Sep 19, 2007 7:40 PM EDT up reply actions

Distortion

There's not, AFAIK, a definitive cut-off between LDs and FBs and GBs and LDs, so at the margins you're going to get some distortion.

It's not perfect, but it's better than not categorizing batted balls, not only for hitters but for pitchers. All things considered, pitchers who give up a lot of line drives tend to pitch poorly. All things considered, hitters who hit a lot of line drives tend to hit well.

For short periods of time a hitter can have an extraordinary run of bad luck and hit a lot of line drives right at people. But over the course of a season, that luck tends to even out. So if a guy has a high line drive percentage and a low batting average, you can predict that his batting average will go up over time.

For example, both Jason Bartlett and Jason Kubel had high LD % and low BA early in the year. Over the course of the year, their BA has tended towards their LD%. In this way, LD is a useful stat.

It's not useful for fine-grained analysis because it's fuzzy at the margins as you point out. But it's a pretty good broad brush approach.

cmathewson

by cmathewson on Sep 19, 2007 4:21 PM EDT reply actions

I agree with the logic...

but is this statement demonstrably true: "All things considered, hitters who hit a lot of line drives tend to hit well."? And are those things directly related? So do those who don't hit a high LD% also hit poorly?

When I looked at it briefly it seemed that those who hit poorly were hitting just as many line drives as those hitting well. The worst hitters in the league, statistically, were hitting as many LDs as the best hitters. Someone like Burroughs hit over 20% LDs, while Maybin hits less than 10%, but what does that really mean?

I suspect that LD/GB ratios with pitchers have some predictive value, although I don't use it myself since, as I said, I do pretty well with pitchers. But I don't know that there's any value to those numbers for a hitter.

And as I said, you could stack up these percentages and not really have any idea how well the hitter really did. High LDs doesn't mean the hitter did well, and low LDs doesn't mean they did poorly. So what value does it have, really?

by beastball on Sep 19, 2007 4:42 PM EDT up reply actions

I think

for minor leaguers it doesn't have that much predictive value. It's pretty easy to see Maybe hits the ball on the ground too much. But the guy is probably changing his swing and will be completely different numbers wise in a year or two.

I think it's very good for players that have been around and are not changing much. Or maybe to give you an idea that a player or prospect MIGHT be changing their plate approach.

by pedrophile on Sep 19, 2007 6:10 PM EDT up reply actions

my two cents

a couple things that jumped out at me in this thread:

are you trying to predict, or to explain? I could see very different applications for each. I think LD% is really helpful for explaining departures from the expected, such as waht cmathewson points out for Bartlett and Kubel, above.
If LD% is being used to predict, what is it you want predicted? I'm sure it correlated with something - the question to me would be is that something a useful aspect of production? It might correlate better with doubles and singles, than with HRs, for example. Pick 6 variables and give me a dataset (with those 6 plus the LD%), and I can run the stats for you (siddfynch@yahoo.com);
Just because some guys with low LD% can still show success in one measure of production is not proof that LD% isn't positively associated with that variable. There can be false positives and false negatives to anything - in fact, most statistical tests for research are set up to let a false positive slip through once every 20 events.
If you want to predict success of a 19-yr-old, it makes sense that there will be more error involved - especially someone athletic who hasn't put it all together. A low line drive rate for Justin Upton might not be as good of a predictor for next year's LD rate as it would be for Ichiro or Jeff Conine, because he is still learning his craft. I think you need to look at the dataset as a whole (100 minor leaguers, not a select few) and see if it correlated in general.

OK, I guess that was about 4 cents. Warn me here if you send that email, because I only check it weekly.

Oh, and I can testify to Artie's skill with pitcher IDs - I leaned heavily on his commentary in here last year, with good results.

by siddfynch on Sep 20, 2007 1:59 AM EDT reply actions

I actually agree with most of what you said...

I'm looking for some ratios that could help predict future success for hitters. The only reason I follow the minors is because I'm prospecting for my fantasy team, so finding a set of data points that can collectively identify the future success of examined hitters is my goal.

Currently, I don't have a good way to identify good hitters given a relatively small sample size (a full year of PAs), which is really what you need if you want to identify the next big prospect.

I only posted this because lately people seem to be using %LD as a data point that can predict a prospect's future success or failure, which is a connection I haven't been able to make. I'm just wondering if they've found a way to examine that data which is useful to predict future success.

I did examine about 15-20 players last year, and found wildly different results. Like a guy hitting 16% LD could be a top prospect, a former prospect, or have the worst OPS in his league. The top prospects LD% ranged from 18% to 9%. While Kemp hit 20% LD in AA last year with great results, Sean Burroughs hit 24% in AAA, and he's a total bust. So I didn't see how that number could be useful, and thought I'd ask.

But I do agree with you, I think it may be useful in examining current performance, but what does it tell you, really? If a guy has a low LD%, but is performing well, does that mean it's not sustainable? Or conversely, if a guy has a high LD%, but is struggling, does that mean he's been unlucky? Cause if that trend were true, then it would be useful in identifying false positives. A month-by-month breakdown would be useful to test that theory.

by beastball on Sep 20, 2007 10:02 AM EDT up reply actions