THE BOOK--Playing The Percentages In Baseball

My point in the SIERA thread about ground balls should probably be here.

Each event has a weight. But, is there a study that shows a weight for each batted ball type?

For example a ground ball can be single, double, grounded into a double play, ground out, fielder’s choice, and an error (I think I got them all). Could we take the run results and the chance that each of these events happen per batted ball type and turn it into a value for the generic batted ball type (in this case a groundball)?

I have to run, but I’ll check the comments later.

Yes. You can use run expectancy to find the “delta” or change in RE value for events however you want to aggregate them. That is, in fact, what tRA does.

Hey JD,

Those numbers have been included in the past few THT Annuals, I believe.

A run scored on the one triple play in 2006, so it was kind of bizarre case.  Only triple play of 2006 too.

Hmm… I get some weird values for batted balls:

groundball    flyball    linedrive    popup
-0.0723904    -0.1014103    0.3391058    -0.2653174

My run values are calculated using GameDay data, 2007-2009, using John Walsh’s methodology, so they are adjusted for count.

for 0.04 Advanced on Innterference… can someone explain why this is so low?  Are catchers sticking their mitts out more when run leverage is low?  Or are fielders just grabbing runners in a bear hug when they’d most likely advance anyways?  It seems to me that advancing on interference would be similar to the value of a stolen base or a walk.

Nick/5, those values seem reasonable to me.  You’re including hits and outs together for each batted ball type, correct? 

So it makes sense that line drives are the only type that is positive, and they’re worth on average a little less than a single.  Popups are almost always caught, so they are close to the value of a generic out.  Groundballs and flyballs should be close to the same value and slightly negative.

Mike/6: sample size.

***

JD: this has been shown in the THT annual, as well as this blog several times.  The numbers are something like:

-.30 pops
-.12 GB (including DP)
-.12 FB (excluding HR)
+.03 FB (including HR)
+.34 LD

So, this is why you can treat pops like K and LD like walks.  And, if they are unrelated to skill (mostly) just ignore them.

That leaves you with GB=FB, once you account for the HR.

Mike, I’m just averaging the run values of each pitch that was in play and hit for that given batted ball type (and I excluded home runs obviously), so yes, it includes hits and outs.  So how come my run values differ so much from Tango’s above and the ones on statcorner?

http://www.statcorner.com/tRAabout.html

Okay, so now I see Tango’s #8 comment, and those pretty much match mine.  Nevermind.

Nick - Tango’s values don’t take the count into consideration.  That’s why your values differ from his.

Mike #6 - I believe that the author just used the Retrosheet category for interference for calculating his numbers.  But Retrosheet includes several different kinds of interference in that category, some that result in outs and some that result in men on base.  So the average is near 0.

There are many mistakes and misleading values on this list.  Categories that the defense has some choice about when they occur are better calculated by the alternative method that Tango uses for Table 4 in The Book as I described by article on valuing the intentional walk at THT last summer.  Intentional walks are worth about .09 runs by this method.  Defensive Indifference drops to about .05 runs.  Also, Strikeout values should be higher (less negative) because strikeout double plays should be moved to a separate category.  Fielder’s choices need to have separate categories for fielder choice outs and fielder’s choice safe.  Error needs to be broken down into Error - Batter reaches base, Infield Throwing Error on hit, and Outfield Throwing Error. 

I also break down singles into infield singles, ground ball singles through the infield, and outfield singles.

Thanks everyone for the comments. I’ve been following along through the e-mail alerts.

Tango/8: Bringing this full circle, if GB = FB, and we assume that the the ability to suppress a HR doesn’t exist (therefore HR/PA or HR/BIP are luck based) then why do tRA and SIERA give weights to GB%?

FWIW, I bought my first THT annual this year. Thanks for filling in a few holes for me!

I should note that when David generated the list, he had no intention to do anything other than spit out based on the event categories that BIS has.

In no way should we treat these as some sort of game-state or base/out-state categories.

JD: if you have a FB/10 term, then you don’t need to treat GB differently from FB.  If you do not have a FB/10 term, then you need to treat them differently.

Basically, just ask yourself: how are they handling the HR?

How was a CS DP better than a CS?