Baseball Musings: Line Drive Variation

January 24, 2006

Line Drive Variation

Yesterday these charts showed the variations in different types of balls in play over the last four seasons. The increase in line drive outs led many to wonder if there was some change in the scoring that's causing the difference. Baseball Info Solutions sent me this information in regards to my questions on the subject:

We did some research on this for another customer, and found the difference was real. There were no changes made to our scoring practices to suggest the difference was as a result of scorer decision.
We do believe our scoring has improved each year, so we are confident of the number. What we're not sure of, however, is the normal variance of these numbers since we don't have any data to compare against other than the four years of our data. It will probably take a few more years of data until we can full gauge whether 2005 is an usual year.

So until we know better, I'm willing to trust the data. However, in presenting tables for fielders, a further breakdown in needed. Since line drives are volitile, we'll also look at tables with only the predominant type of ball in play. That's grounders for infielders, fly balls for outfielders. The shortstop data will be available soon.

Posted by David Pinto at 02:41 PM | Probabilistic Model of Range | TrackBack (0)

Comments

That's a great solution for 2005. For both GB and FB, actual and predicted DER are quite close.

However, you'll still have a problem if you use the 4-yr model to compare one year to another. It will appear that almost all players were above-average in 200-03 (Lake Wobegone!) but below average in 2004-05, since DER has dropped for both GB and FB.

BTW, these two statements from BIS:
"There were no changes made to our scoring practices" and
"We do believe our scoring has improved each year"
cannot both be true. :>)

Posted by: Guy at January 24, 2006 03:18 PM

Good catch, Guy. Marketing people try to put a good spin on everything, even when it leads to a contradiction! :) I'm sure the defensive model would work better if it accounted for year-to-year variation in the data.

Posted by: Jason at January 24, 2006 03:43 PM

The precision in this type of analysis will always be influenced by the accuracy of the stringers. The organizations who track this sort of stuff have become better with cross checking over the years, but there is still a great deal of subjectivity in the "rulings". It doesn't help that the turnover of these stringers is very high from year to year.

Posted by: J.P. McIntyre at January 24, 2006 04:07 PM

We actually have a potential cross-check on this. STATS Inc also does hit typing; their groundball/flyball counts for 2002 -2005 are available (eg at the ESPN webstite). I'm not aware that their bunt/line drive numbers are directly available, but assuming their bunt count would be very similar to BIS; with the following procedure on the MLB totals I estimate their line drive total:
(AB+SF+SH-K)-gb-fb-bunts
this is basically batted balls "in play" minus the accounted for hit types. David's definition of a ball in play removes (over-the fence?) HR because hardly any of them are fieldable, so the year by year totals he gives would have any line drive HR subtracted from the totals. With this estimate I don't try to do that. BIS line drive totals with HR added back in should be slightly higher, and the BIS year to year variation might smooth out slightly.
here are the totals
BIS STATS
2002 27,852 26,417
2003 29,779 26,783
2004 24,885 26,210
2005 27,775 26,121

Apparently according to the STATS data, line drives do not fluctuate from year to year so significantly. I find that a more plausible result, in part because if you look at MLB totals for other sorts of counting results (singles,doubles, strikeouts,hit by pitch, whatever), you don't see fluctuations so extreme. I still believe that Guy had the right explanation.

Posted by: joe arthur at January 25, 2006 07:20 AM