THE BOOK--Playing The Percentages In Baseball

Random question, but why did we ever start using BABIP instead of linear weights on balls in play?  And does linear weights on balls in play have a higher skill ratio given that doubles generally are hard hit balls?

I see on Fangraphs that ZiPS is projecting Strasburg to have a .333 BABIP for the rest of the year. Is ZiPS assuming that the freakishly high number that he currently has is indicative of some susceptibility to a high BABIP? I’m trying to figure out why it would assume that the number will remain inflated. (Even if it thinks it will be less so.)

Because knowing how many H and HR a pitcher allowed is all we knew prior to Retrosheet.

Thanks Colin, I didn’t know that.

Excluding HR, and splitting the BIP into Pops, Liners, GB, and FB: the run value of a GB is virtually identical to the run value of a FB.

So, while you have a higher BABIP on GB, you also have a higher DP rate and a lower SLG.  Put it together, and you’ve got a wash compared to FB.

Ideally, you’d have wOBA that included DP, and that’s what we’d be quoting.

Neil/2, that .333 BABIP projection is most likely just that .360 BABIP regressed toward the ML average. I don’t think that it means Strasburg is particularly vulnerable to base hits.

The point is that a .333 forecast is unreasonable.  It means a 50% regression, based on only 50 BIP. 

And if you include his minor league data (of which he had a great BABIP), it makes the .333 even more improbable as an estimate.

Clearly, ZiPS has issues with BABIP, with regards to players with limited BIP.

Maybe Zips doesn’t like the Nationals defense.  Who do their other pitchers project?

Is there any correlation between pitchers that throw hard and higher BABIP allowed? I’m just wondering if Strasburg throws so hard that batted balls tend to fly farther/harder/be hits more often…

Rick/9, higher, no.  Lower, maybe.  The ball would go farther if it was contacted just as squarely, but it may be harder to contact a faster pitch squarely.

We talked about it here:

http://www.insidethebook.com/ee/index.php/site/comments/danvenport_translations_strasburg/#8

So, bottom line, he has a 1.78 ERA and .947 WHIP while being unlucky… Small sample, so perhaps he’s been lucky along some other dimension? Stranding runners, maybe?

"So, bottom line”.

No, that is NOT the bottom line.

Doesn’t seem too strange considering the 26.9% line drive rate.

Tom, you can satisfy your curiosity to how much in-season ZiPS is regressing current season performance. The original spreadsheet is available for download:

http://www.baseballthinkfactory.org/files/oracle/discussion/2008_zips_projections_in_season_projection_tool/

So that should tell you want you want to know.

are we sure about the throwing harder = lower BABIP?  I mean it makes sense to me.  Can’t square up resulting in weaker contact. 

I’ve heard alot tho faster in = faster out....which would mean harder contact and harder thrown pitches.

Don’t forget that some batted ball disappear from the numerator and denominator… they become HR.

So, you have to be careful about that.  If the hard throwers throw a disproportionate number of HR relative to all BIP, then the reason they have a lower BABIP than league average is because doubles turned into HR.

I’m not saying it’s happening.  I am saying you have to be careful.

Re #15 and #16, I linked in post #10 to the discussion we had earlier about this.  It’s sorta silly to repeat all that again and speculate without looking at the data.  The data is presented in that thread.

Thanks.

I’ll take a look at it.

Robot batters might hit the ball harder on contact if pitchers throw faster, but human hitters generally would be inclined to shorten their swings, choke up, etc. when the ball is harder to hit.  High-K pitchers have lower BABIPs, and I’m inclined to think that’s because batters don’t swing as hard in an effort to avoid striking out.  That’s probably why you don’t see the faster in/faster out effect.

Thanks for the link to the previous article and further explanation about BABIP and pitcher velocity.

Yeah, looking at the tool that is used to calculate the numbers, I think it’s just a simple weighting of a player’s actual stats and his pre-season projections, without regressing those actual statistics at all.

For whatever reason fangraphs doesn’t have Strasburg’s preseason projection, so I’ve entered his actual stats and tried to reverse engineer the projections.

I’ve gotten a nearly identical RoS projection when I input a preseason projection of 20 starts, a 10-10 record, a 4.00 ERA, 115 IP, 115 H, 13 HR, 35 BB, and 111 Ks.

So that’s 495 batters faced, and therefore 115 hits on 371 balls in play, for a projected BABIP of .310.

Thinking about Strasburg’s slightly better-than-a-month-ago BABIP and how it is affected by his team’s defense, it reminded me of a question that hopefully Rally or somebody else can answer.  On the pitcher pages at baseballprojections, do the runs in the H+ column consider how many runs a pitcher himself prevents compared to league, or is it runs saved by hits on BIP by the pitcher plus his defenses?  So if Maddux is +56, does that mean he prevented 56 runs compared to his teams on BABIP or just league on BABIP?  The glossary says “compared to league with no contextual adjustments”, but the site also says that the X column includes all of Maddux’s defensive support, and the H+ column and X column can’t both contain his defensive support.  If it is true that Maddux received 64 runs of defensive support (most due to BABIP reduction) and only prevented 56 compared to league on BIP, that would imply that Maddux’s real skill in BABIP was non-exsistant.  That leads me to believe that H+ is comparing pitcher to his teammates. Is this correct?

So...can anybody help me with my question?

The H+ column compares him to his league.  There are no adjustments at all.  Those are 56 runs that could be saved by Andruw Jones in CF, by Maddux’s own pitching, by dumb luck, ballpark, or whatever.  Just simply that when he was pitching, there were fewer hits on balls in play, and worth that many runs.

Then you add up the runs from walks, hits, strikeouts, HBP, and homers, and get a total of +496 for Maddux.  But he actually gave up 542 fewer runs than league average.  So the X is a catch-all for everything else.  It could be defensive support, picking off baserunners, sequencing, luck, lots of DP balls, or any other explanation you can muster.

I get that, but I am still confused about something.  If Maddux prevented 54 runs on hits compared to his league, and his defense saved him over 60 runs (due mostly, I am assuming to BABIP prevention), than wouldn’t that imply that Maddux did nothing on his own merit in regards to BABIP?  But we know for a fact that he prevented tons of hits on BABIP compared to his teammates.  How does that all work out?  Same with Glavine - his defense saved him 80 runs and he is +72 preventing hits on BIP.  However, we know that he prevented many hits on BIP compared to his mates. What am I missing?

One possible thing I can think of is: that a majority of the 60+ runs prevented by Maddux’s defense or 80 prevented by Glavine’s defenses are not hits prevented by BIP, but by DPs, outfielder arms, outfielders keeping singles from becoming doubles, etc.  That would still leave room to credit them for their BABIP differences between them and mates.  This may match up with the fact that the strength of the Braves defenses were their outfields too. Of course as ground-ball pitchers, TZ may be punishing them unfairly as a majority of their BIP went to their less amazing infielders.  I think PZR considers that and both Maddux and Glavine do have better PZR based WAR from 2001-2006 than rWAR.

They aren’t going to reconcile all that closely.  The right hand side stats are not park adjusted or anything.  They are there only because I thought they were interesting.

I have Maddux giving up 84 fewer hits than his mates, Glavine 64, for the record.  I’d hope that the more extreme cases reconcile better, like Charlie Hough.  He had about average D support, but is +193 on hits in play.

One example that could throw things off is ballpark.  Take a Fenway pitcher for example.  Let’s say that the team babip is .325, and league is .300.  Despite this, the defense still rates above average by TZ because of the ballpark.  Say a pitcher gives up a .315.  Now he’s showing above average defensive support, plus fewer hits than his mates, but yet his hit+ column shows a negative number compared to the league.  I know that’s an extreme example, and Tom and Greg didn’t pitch in an extreme ballpark like that.  I just made up an extreme example to show how things will not reconcile, and in the Maddux/Glavine cases we’re talking about 2-3 runs per year.

The DP/OF arm/Catcher runs you mention could be a part of it too.

Thanks, that makes sense.  So if Maddux allowed 84 fewer hits, than how many runs prevented is that?  About 45?

Probably 60-70.  It’s not just the hit, it’s the difference between a hit and an out.  It’s .75 for singles, though he almost certainly prevented some doubles and triples in there as well.

Does that consider his own baserunner environment (fewer guys on base) or just the league average baserunning environment?

Sorry for the bombardment of questions, but I have one more: on his “DIPS20” page (http://www.tangotiger.net/DIPS20.htm), Tom shows Maddux to have been +100 hits compared to teammates through 2000 and Glavine to have been +74. Even though neither probably increased those levels too much, they probably did a little bit.  What is causing the discrepancy in your numbers and Tom’s?  Infield singles?  Different hit databases used? 

Thanks again as always!