THE BOOK--Playing The Percentages In Baseball

Also, did he really need to email someone to find out what OPS+ is?  He clearly has access to the internet.

Things like WAR and UZR are not formuli that can be posted.  Or Totalzone, which is a component of my WAR.  For me to show you the entirety of the process would take days.  You’d have to be looking at my computer, and the code for one program after another, all which must be run in the proper sequence to get the right results.  I’m not a great or efficient programmer.  My code gets the job done but will not often make sense.

Outside of the sabermetric oligarchs, your reaction to such a process might be:

1. Your eyes glaze over and you daydream about something else.  This might be best, and safest.
2. You are overcome with fear and flee the room.
3. You find you some level of understanding, and try to delve deeper.  This is a mistake.  Soon you are trapped, and as you go deeper the numbers slowly drive you insane.  At the end of the day you are an incoherent mess, babbling on about the game changing abilities of Jason Kendall.

I’m not really convinced… I mean, I followed the stuff about WAR in my own halting way, and cobbled together a version to use with my old High Heat 2001 baseball game.  Just plug in the numbers onto a spreadsheet and the formula takes care of the rest.  And since the game already calculates EqA, it’s a nice balance.  In fact, it kind of makes the game ridiculously easy.  This stuff is out there to find.

I’d really love to try to approximate Goals Vs. Threshold for my Eastside Hockey Manager… I don’t know if the game keeps details enough information, though.

Well, if anyone wants, say, a spreadsheet that calculates WAR from scratch, and an explanation of what is going on and why, they could read these:

http://www.hardballtimes.com/main/article/how-to-measure-a-players-value-part-i/

http://www.hardballtimes.com/main/article/how-to-measure-a-players-value-part-2/

http://www.hardballtimes.com/main/article/how-to-measure-a-players-value-part-3/

http://www.hardballtimes.com/main/article/updated-player-win-values/

The last link has a copy of the spreadsheet available for download. So if you absolutely need to know everything about how WAR is computed, and see how WAR changes as your assumptions change - there’s even a page that contains EVERY constant used in the formulas, so you can swap in your own constants and see how that affects things.

Or you could accuse people of not publishing things they have in fact published. Your choice.

Or let’s say you want to know how a defensive metric works. You have this:

http://www.hardballtimes.com/main/article/fleeter-than-birds-part-1/

http://www.hardballtimes.com/main/article/simple-zone-rating-takes-on-the-outfield/

Which includes SQL code that anyone can use to generate defensive ratings from Retrosheet data.

Comparing the 2008 financial crisis to the modern sabermetric community is completely absurd and highly ignorant—and I don’t intend to be mean;
ignorant is the only appropriate word. The crash in 2008 happened because the investment banks had quit trying to answer real questions and instead: starting splitting up financial assets into minute and complex assets, selling predatory loans (such as ARMs), and lending irresponsibly under the faulty assumption that real estate values were going to increase indefinitely.

How on earth does that compare to current sabermetrics? Even the most recent acronyms (SIERA) are still trying to answer legitimate questions ("how can we add context to pitching performance?").

Colin/4 was marked for mod and is now open.

(Hint: if you don’t put the http:// but do put www. in front, then the software does the link.  Otherwise, it treats it as spam.)

As someone who frequently asks questions of the oligarchs, I can attest that they usually respond very quickly and in a helpful way.  Tango has answered half a dozen questions for me and MGL was nice enough to critique a semi-primer piece I wrote for my blog.

The part of the community I interact with is very open.

Side note: Anyone else excited that the time stamps on the site now match the actual time, thanks to daylight savings?  Or is that just me?

Not in Oregon they don’t!

Me too Sky.  Subtracting 1 all the time can wear on you.

Okay, perhaps WAR, UZR, and WPA were bad examples (although the problem I have with WPA - it is based on a premise that all teams and players are equal - remains nevertheless legitimate.)

What I was trying to do back in 2005 was not to re-create anyone’s work, but to generate OPS+ for the Can-Am League, where park effects seemed to be even more pronounced than in MLB.  What I needed to know was how the ballpark factor is calculated.  The link provided here just treats it as a given (which I suppose is reasonable since B-R lists the BPF for MLB parks).  This illustrates again the frustration I felt and the point I was trying (but apparently failed) to make in my original post.  Not everyone is just pulling stuff off retrosheet.

As for the “evidence,” I no longer have the emails from five years ago.  If you say you were not haughty then I cannot argue with that (given your patience in The Church of Baseball Part 2, I’m inclined to believe you).  I remember not getting a real answer from anyone, but the only guys I remember being real aholes were those at FireJoeMorgan.

Brad @5, I didn’t say they were identical.  Hopefully you’ll agree there at least some similarities (i.e. the ones I mentioned)

Look, I was not trying to “accuse” anyone of anything.  Colin, I usually read most of the stuff on THT every day; sorry I missed your work.  The larger purpose of my post, beyond questioning any particular metric, was to explore what I perceived to be a very wide canyon between sabermetrics’ sense of itself and the ideas that the average fan may have about it.

@Church #4: I agree that there are perhaps superficial similarities, but the 2008 crisis was thoroughly widespread and—as I noted before—a beast of a different species: investors quit looking for new information. In the saber community, I don’t think there is a widespread abandonment of reason amidst desire for greater complexity. I just don’t see it, at least.

Oops, that should read:

@Church #11

"The larger purpose of my post, beyond questioning any particular metric, was to explore what I perceived to be a very wide canyon between sabermetrics’ sense of itself and the ideas that the average fan may have about it. “

Why should there be anything less than a wide canyon between sabermetrics and what the average fan thinks?  Who said that sabermetrics has anything to do with the average fan?  It doesn’t. Some people and some aspects of it do, and occasionally (very rarely) I write something for the average fan.  Some people seem to think that sabermetrics is the science of trying to explain the statistical analysis of baseball to the average fan, or trying to convert the average fan to a different way of thinking.  It ain’t. In fact, nothing could be further from the truth, at least for me.

What you just said is like someone criticizing nuclear physics or cardiology because of the chasm or disconnect between the science and the general public.  There are hundreds of thousands of articles written about cardiology or even nuclear physics that are designed for the average person to understand and learn from, some good and some bad.  But, the average person would have no idea what a biventricular axillofemoral atrial fibrilation is or a Tashkent geodynamic polygon.  And nor should they.

Sabermetrics is an applied science.  As I said, sometimes there are aspects of it designed or written for the general public.  Some of these writings are clear and easy to understand and some of them are not.  I did not in any way shape or form design UZR to be understood by anyone in particular (although it is NOT rocket science - not even close).  Having a description of it on Primer or Fangraphs is a courtesy.

Attacking things with a wide brush is usually an exercise in futility.  Attacking things with a wide brush and a host of misleading and incorrect information and assumptions is reckless.  Which is pretty much what I am doing right now.

I tend to agree with MGL.  It seems as though it’s become the goal of many saberists to try to educate the masses.  But why should we care what the average fan thinks?

The things that’s missing with regard to transparency for the fan, at least with regard for defensive stats, is we can not see how the defensive stats change day to day. 

For example, say Jacoby has a UZR of -5 on June 1.  Over a 1 week period say he makes a series of great plays, and the fan would like to see how his great play affected his UZR, if it goes from -5 to -6 on a day he seemed to play very well, as a fan I might question the stat (might even rewatch the game on FF to confirm my memory and see if any balls dropped in due to poor jumps).  If it went from -5 to -4 that might validate the stat for me (at least for that game if free of bad jumps). 

But, you can’t check because unlike most of the offensive stats, you can see how they change day to day.  There is no game log, it’s not in a box score, etc.

Furthermore, I don’t trust UZR or even FB+ for OF’ers at Fenway, especially RF and LF due to the wall and lack of foul territory.  Why can’t we see home vs away splits for UZR and Plus/Minus?. 
Surely there is an interest in these splits.  UZR I forgive in this regard, it seems to be free, but the FB+ makes you pay for their stuff (BJ site or the book).

This is the lack of transparency that most talk about when it comes to defensive stats, not the lack of a formula or whatever. 

We are given a number at the end of the year, or a number at a given point in the season, and have a choice to believe in it or not, Amen.  There is no way to check it against our own observations over a short period like we can with other stats, or home vs away (a big difference might expose a flaw in the stat, or a small difference may make the stat more credible to the fan).

None of these requests seem to pose any technological hurdles, yet they are not available. 

Now, perhaps those who design the stats could care less.  That’s fine. But some might think the lack of transparency is because providing said information would invalidate the stat.

And lets not pretend that the saber/stat business is not generating any revenue.  Maybe not for all, but web sites get sold, books get sold, advertisements, subscriptions, etc. 

As for the comments on the science of sabermetrics.  One of the differences that I see is sabermetrics seems to make little attempt in estimating the uncertainty of their theoretical estimators and numbers.

So we get numbers like 5.2 WAR for Player A which means (to those who live in a world where uncertainty does not exist) he is better than Player B at 4.8 WAR, whereas in science they would say Player A is 5.2 +/-1 WAR and Player B is 4.8 +/-1 WAR and given the uncertainty, there is no statistical difference between the 2 players. 

After adjustments for park, league, era, and position, you probably get numbers that can pretty much be used to justify anything, or discredit it, depending on your biases, and nobody can prove or disprove it.

As one mathematical physicist said, give me 4 free parameters and I cam model what an elephant does, give me a 5th free parameter and my elephant can fly.  In business I was always wary about adjusted numbers and estimators.  You can’t avoid them but they are a potential source of great mischief. 

Financial models, climate models, and any kind of model are only to be trusted if they can be validated, or if their predictions are shown to be accurate over time. 

Nobody likes to admit uncertainty unless they have too, so unless you need to get published in a peer reviewed journal, no need to bother.

The numbers are fun though, but I see a danger where folks get too caught up in defending their positions and no longer question their beliefs.  One of the great obstacles to the truth is the illusion you already have it.  This is a problem in science too.  Old paradigms die hard and give birth to new ones, usually only when the believers of the old paradigm die off.

None of these requests seem to pose any technological hurdles, yet they are not available.

Now, perhaps those who design the stats could care less.  That’s fine. But some might think the lack of transparency is because providing said information would invalidate the stat.

The hubris is tough to swallow, but I’ll wade in and try.

MGL is the one who designed the stat, and he has in fact given away his formula for free, so you can get the information you want as long as you are willing to pay BIS or STATS for the batted ball info.  What, you don’t have 10 grand lying around?  Well, neither do most of the rest of us.

So you can do what, for example, Colin Wyers is doing and make your own adaptation based on the ideas that MGL has given away FOR FREE, by downloading and using the batted ball stats that MLBAM is providing you FOR FREE.

But that would take a lot of grunt work, and maybe you want other people to do that for you FOR FREE?

You might want to consider a thank you to folks like MGL and David Appelman.  I know I appreciate them and the work they have shared.

Throwing around accusations without any evidence isn’t very helpful.  You’re welcome to dig in and help advance the frontiers of defensive analysis.  There are a lot of people working on it, and we’d love to have another capable person on board.

There is no conspiracy out there that I am aware of to withhold information regarding some of the metrics.  There might be some that are proprietary for whatever reasons, in which the authors are not willing to share some of their work.  That’s fine of course.  Business is business.

BTW, there will be some improvements on UZR this year which I will explain when they come out.  As well, Fangraphs will likely start showing H/R UZR breakdowns this year, so that people can stop whining about Fenway and some of the other parks.

I am not directing this at pft above (I didn’t even read his post), but it is ALWAYS, ALWAYS, ALWAYS (and ALWAYS) people who are ignorant about something that criticize it with unfounded assertions and ridiculous claims.  That goes for everything - global warming, health care, evolution, etc. - not just baseball stats.  If you don’t understand something, it is OK to ask questions of course, but you want to do a full on critique on something, like we see and hear from the media (and bloggers) regarding sabermetrics 100 times a month?  You damn well better be an expert in that field otherwise I’m not going to listen to you, and no one else should take you seriously either.  But, alas, the world is full of experts and geniuses…

Daring to speak for MGL, I would say that the main reason UZR splits aren’t displayed is that people would be bound to abuse them.  Oh, Jason Bay’s 2009 month by month splits showed a slight trend upward?  That must mean that he was improving and learning Fenway and we should expect him to be much better next year.  Matt Holliday’s UZR was worse on the road than at home?  That must mean Coors field overrates fielders.  You get the point…

Insinuating that the only reason that UZR splits aren’t displayed is that MGL or Appelman feels that they would “invalidate the stat” couldn’t be further from the truth.  MGL has preached not to take explicit UZR numbers to seriously, and only appreciate it in it’s aggregate value and use it as a measure for inference.  He is far more aware of the measurement error in UZR than most other fans who quote the stat all of the time.

The numbers are fun though, but I see a danger where folks get too caught up in defending their positions and no longer question their beliefs.  One of the great obstacles to the truth is the illusion you already have it.  This is a problem in science too.  Old paradigms die hard and give birth to new ones, usually only when the believers of the old paradigm die off.

I have no idea what this means.  What is with everyone and the mediphores recently?

MGL/18, last paragraph: ditto!