Another big advantage of this method of calculating xBABIP is that it works easily for pitchers as well.

Both the other methods supplied all look very interesting, but it requires a lot of data that I don’t have on players(and I’m not sure how to get). The method I developed can very quickly determine an expected BABIP based on some stats that are readily available via fangraphs, for both batters, and pitchers. I’ve been playing with it a lot, while doing my 2010 projections, and I’ve been fairly happy with it’s results. There are a few outliers that I’ve noticed though, that are probably worth mentioning:

Matt Kemp: my system projects him with a .347 BABIP, which is well below the numbers he’s been posting, is this a flaw in my system, or is kemp just getting lucky?

Ryan Howard: my system projects him for a .349 BABIP (due to his high HR/FB rate, and extremely low IFFB rate, and high LD rate. Has howard been unlucky with balls in play? or is this a flaw?

Nyjer Morgan: my system projects him for a .328 BABIP (low LD%, relatively low IFH% (compared to other speedsters)). Has nyjer been lucky thus far in his career? or a flaw?

Brandon Phillips: my system projects him for a .310 BABIP (lots of ground balls, and a descent IFH%), has he been unlucky in recent years? or another mistake?

Besides these anomalies (or are they?) the other thing I’ve noticed, is that IFH%, IFFB%, and LD% are all vary greatly from year to year themselves (via luck), so even breaking BABIP out into this batted ball data, it’s still difficult to project accurately, but it’s still better then just looking at BABIP itself. You can do things like "this guy has really good speed, and bats left handed, so I tend to believe more in his higher IFH%, and tweak accordingly.

I don’t think you really need to find projected batting average to the thousandths of a point.

I found the following post:http://www.hardballtimes.com/main/article/batters-and-babip/

In comparing my results to theirs, They get an R square around 35%, my results give me an r square of 44%. So it would appear that basing BABIP on these stats does indeed do a pretty good job.

Of course, the nice thing about their method, is that it’s based on more skill based stats, while LD%, IFH%, IFFB%, are still fairly luck based statistics (and vary a lot from year to year themselves).

But this method should be an interesting way to project at a glance, a batter, or pitchers expected BABIP.

And another article: http://www.hardballtimes.com/main/fantasy/article/whats-the-best-babip-estimator/

At 44%, again, this method shows to be the most accurate (Again, disclaimer is, it’s hard to project what a persons true LD%, and IFFB, IFH% are).

I ran this comparison, and here’s what I found:

First, I compared the difference between a batters BABIP from 2008 and 2009, and his 2008 xBABIP vs his 2009 BABIP.

51% of the time, the xBABIP was closer then just using the actual BABIP was.

Next, ignoring the range of difference, I totaled up the number of times that xBABIP accurately predicted that a batters BABIP would rise, or fall the following year.

65% of the time, the xBABIP accurately predicted a rise/fall the following year.

I used a sample of about 100 players who had a minimum 500 plate appearances in both 2008 and 2009.

So what does this mean? Well, I do feel that the batted ball data that I’ve identified has a very strong correlation with a players BABIP. Unfortunately, that same batted ball data will vary greatly from year to year itself, making those stats themselves, impossible to predict. So what good does this do?

Well, it lets us break up BABIP into something that makes a little more sense to us, let me explain:

LD%: Based on my research, more often then not, a players career LD% fluctuates around the 19.5% mark. So, if the batter you are looking at has a LD% that’s over, or under that, it seems to imply that they have been a little lucky/unlucky with their flyballs. You can regress that number closer to the 19.5 mean, and odds are, you’ll have a better guess at his LD% for the next year

GB%/FB%: In most cases these remain fairly consistent over the course of a players career, i’ve found these are easier to predict more accurately, with a few noticable exceptions (sometimes strangely a player will flip flop and become a flyball hitter one year)

IFFB%: Players with better power tend to post better (lower) IFFB% over the course of their career. Overall, this stat tends to fluctuate a lot, but if you notice an IFFB% that stands out from the rest of their career, or fluxuates greatly from the league average (8.9), then you can regress it towards the mean.

IFH%: Better for left handed hitters, and much better for speedy players. You can rule out a flukey low IFH% and regress it towards a career average, or towards the mean (5.97 league average).

Really quick I took 3 players, and attempted to ignore the existence of 2009, and using the ideas I outlined in my last post, and my xBABIP calculator, posted my results below.

Soto: 14.7 HR/FB 20.25 LD% 37.7 GB% 42.05 FB% 7.65% IFFB% 5.97 IFH% Predicts: .319 BABIP
Actual 2009: .251

Comments: I predicted it would go down (most would have), and it did, but it did so by a LOT!. Drilling into the numbers of 2009, we can see why: 18.1% LD% 4.8% IFH% His line drive went down, as I expected, but by much more. And his IFH% went down even more then I predicted, though I can’t say this is really all that unexpected, given his speed.

Maur: 8.5% HR/FB 21.5 LD% 49.2 GB% 29.3 FB% 2.4 IFFB% 3.6 IFH% Predicts: .335 BABIP
Actual 2009: .377

Comments: I incorrectly predicted it would go down, I thought his LD% would regress further then it did, and I couldn’t have predicted his spike in HR/FB% (which helps your BABIP, because that’s less fly balls in play), his IFFB% also went lower then I would have predicted.

Votto: 18.5 HR/FB 22.35 LD% 44.1 GB% 33.55 FB% 3.8 IFFB%, 4.34 IFH% Predicts: .342 BABIP
Actual 2009: .373

Comments: His IFH% did indeed go up, as I expected it should have, while his IFFB% improved to 1.4% which helped aid in a big increase in his actual 2009 BABIP over my predicted one.

This is a very small sample, but it gives you an idea of how people (even smarter people then I) could leverage the ability to predict a BABIP using batted ball data.

OK, after learning a little more about how to do correlations correctly, I came up with the following results:

Comparing 2008 BABIP to 2009 BABIP: .35 correlation
Comparing 2008 xBABIP to 2009 BABIP: .48 correlation

It turns out that this is indeed a better method of predicting BABIP then simply using last years BABIP. Of course, in the grand scheme of things, .48 still isn’t that great. What I’m most interesting in discovering at this point, is how to better predict what a persons IFFB%, and LD% is going to be (based on skills) since those things vary so much from year to year, and have such a large impact on your BABIP. If I can figure this out, I’ll have a much better idea. Better still, would be to introduce spray into the equation (but I can’t figure out how to get spray data).

Using this formula, and the 2008/2009 data, I adjusted the 2008 data towards the mean for each of: IFFB%, IFH%, and LD%, and used the true 2008 GB% and FB%. Doing this, got me up to a .57 correlation. Further analyzing the results:

65% of the predicted BABIP’s were within 30 points of the actual
55% were within 20 points
25% were within 10 points
14% were within 5 points
and 80% of the time, the xBABIP predicted correctly whether the BABIP would rise or fall the next year.