THE BOOK--Playing The Percentages In Baseball

The first 5 responses have been fantastic Tango. Thanks.

Speaking of Bert (circle me!), his Hall case (especially vis-a-vis Jack Morris) has become something of a barometer for the sabermetric community, almost an unofficial badge of belonging or a ring of fire through which someone who hasn’t been in the community forever must pass for acceptance in it. I should probably look at Mike’s site, but I’m curious to see how he’s touched on the RAB for Cooperstown issue, and if he’s dealt with the gospel-level of acceptance of Blyleven as a Hall-of-Fame-caliber player in the sabermetric community.

There are definitely ‘too many’ stats out there, but this is simply because people keep making improvements on them but not everybody stops using the old one.

For example, WAR is better than VORP at measuring value above replacement.  wOBA is better than OPS+ at measuring run production generated by a player (though it doesnt do park adjustments which OPS+ does).  VORP and OPS+ are in turn better than prior stats. 

The saber community is evolving new stats (and often, different segments of the community are each evolving their own stats).  Often by the time that one stat starts to get a following and some press and some fans start using it, the saber community has come out with newer and better stats. 

Eventually, I would hope that we will eventually have the following stats, which would later be adopted by the mainstream:

* One ‘uber’ batting rate stat (per PA), and corresponding counting stat.  Currently: wOBA for the rate stat, and ((wOBA - ~.300) * PA / ~1.20) for the counting stat, with the .300 and 1.20 changing slightly year to year.
* One ‘uber’ fielding rate stat and counting stat.  Currently: UZR/150 and UZR? - But these don’t account for catcher defense.  Will HitF/X give us something better? 
* One ‘uber’ pitching rate stat and counting stat.  (Some would say FIP here but I think this stat needs to capture the type of batted balls generated.  Line drives, ground balls, popups, and fly balls are very different).  I would say something like tRA+ for the rate stat?  Which is essentially linear weights for each outcome (HR/K/BB and batted ball outcomes), with the HR% regressed towards the pitcher’s career average.  Will PitchF/X give us something better?

These stats or the improved versions of them then give us our ultimate stat, WAR (reformulated to account for the new advanced stats).

Fangraphs taking over stats for THT could be the start of some consolidation, as some did not make it over.

I always thought VORP was for those inside BP who thought that BP’s fielding metrics were crap.

In defense of the “there isn’t too many stats” camp, one thing that I like about the multiple stats is the ability for newcomers to follow the sabermetric logic.

Some of the lesser stats are sometimes easier to understand and comprehend.  It may not be analogous, but if we got rid of all the Physics textbooks that dealt with Newtonian mechanics and forced all high school physics students to jump right into Quantum Mechanics, they wouldn’t learn anything, because they don’t have a basis.

Maybe I’m overstating the complexity of some of the new stats, but for those who are still learning, it’s nice to see the historic flow.

the saber punch needs a little more vintage but hopefully withg a little more seasoning from the marketplace it will be smooth and refreshing to the main stream palate.

i totally agree with tom tho in that getting saber stats listed on espn broadcasts is not the be all end all.  the stats are there for those who want them.  for those who dont, whatevs.

somewhat tangentially: if batting average was invented today and no other “saber stats” existed, would be also dismiss it as overly convoluted and invented so therefore not “real”?

Tim – Since you like to learn of the development of baseball statistics, read the following articles.  All the questions we are asking today were being asked 100 years ago and many of the answers were not too different than the answers that we have reinvented and are still using today.  If F. C. Lane had had Retrosheet at his disposal there would have been very little for modern baseball analysts to do.

http://www.la84foundation.org/SportsLibrary/BBM/1913/bbm112i.pdf

http://www.la84foundation.org/SportsLibrary/BBM/1919/bbm226k.pdf

http://www.la84foundation.org/SportsLibrary/BBM/1916/bbm165j.pdf

Particularly note the following article and the run values he calculates for each type of hit , and how close they are to linear weight values.

http://www.la84foundation.org/SportsLibrary/BBM/1917/bbm183o.pdf

Many of Lane’s other articles for BaseBall Magazine can be found here:

http://search.la84foundation.org/search?site=default_collection&client=default_frontend&output=xml_no_dtd&proxystylesheet=default_frontend&proxycustom=

Enjoy.

Maybe I misinterpret the question, but why should we really care about market saturation?

From all the questions, it really seems like Silva is convinced that the saberworld is out there trying to “sell” a product, rather than trying to analyze the game for a better understanding in depth. The questions range from the marketability of stats, to the saturation, to the way those are “sold” to the public (or to teams), to the future of statistical evolution. That is entirely missing the point. Tango gave an excellent explanation elsewhere of why there are other reasons driving research as opposed to trying to make money and deriving power from all of this, and he’s exactly right.

That is why we keep trying to advance and we do so at a good pace. That is why if there are margins for “better” stats, they should be analyzed, worked upon and introduced, regardless of market saturation and regardless of the previous stat having become mainstream. Take wRC+: undoubtedly better than OPS+. Silva had that “money trail” title in a piece that didn’t even care to address the value of the stat and its improvements as opposed to previous metrics. It really seems like Silva does have a beef with the stat world and its advancements, and not with the quality of stats themselves. I may be wrong, of course, but to me any objection that does not attempt to criticize the stat (the question the stat answers, the framework, its strengths and weaknesses), but only its presence, is entirely missing the point of statistical analysis and advancements.