Pitch Sequencing
Send Felix to the Rock-Paper-Scissors tourney


One cubicle left, SABRMatt has a thread going.  In it, he perceives (and argues) a thing that is a "meta" aspect of the game.  Neo would never have been able to grok Agent Smith, if Smith hadn't volunteered to tip his hand.  Chaos theory, baby.

Matty writes, concerning Arneson's Theorem 1,


1) Sabermetrics uses a coding language that structurally misses the most important part of the game - pitch sequencing 

OK - this is a major...MAJOR rebel yell for me. I worked in the game for a year doing nothing but getting data out of a relational database and putting it into R to look for patterns. And my number one frustration with SQL and R is that it is prohibitively difficult to look at sequenced events in either language. It can be done in R with some thought, but it is not easy. Designers of SQL - could you *please* work on some commands that sequence data sets in case an analyst doing data mining would like to study sequenced events...cause that's a big thing in a lot of database-driven fields, and your language STINKS at it.

And I can tell you with absolute certainty that Ken is right that the game ignores sequencing, even on the inside. When I made my pitch for employment with Theo Epstein's Cubs, I spent half of the interview convincing his head of analytics that Felix Hernandez is more than the sum of his pitches as though they were random - that none of his "stuff" should be THIS effective in a vacuum, and that it had to be in the sequencing that he was getting such good results. They went right to their in house relationally grouped SQL-produced graphics and said "wow...Felix is this good because his change-up is off the charts" - they missed the point. WHY is his change-up off the charts? He has good arm action but a very readable release point. His change doesn't break that much more than normal (it sinks with some armside run, but it's not like a Nintendo pitch). The velocity differential is laughably small. His change-up isn't just effective because it's a good change up (though it is)...it's effective because he knows when to deploy it.


Do you want to truly appreciate why Matt is right?  (No, wait, that's a catastrophic lead-in.  See, I'm two sentences in and my post is already off the rails.)  

... Do you want to know why rookies get so consistently abused by MLB(TM) veteran pitchers?  We'll tell ya!  FREE!

There, that'll get most of our readers to the first picture.


To learn extremely much about baseball and life, it will cost you something, however.  It will cost you about 50-100 turns at this little Rock-Paper-Scissors game.  But that's only if you want to forever understand, at a cellular level, why Brad Miller suffered the way he did during parts of 2014.


In my own little game, I triumphantly ripped around the scoreboard like a Safeco Hydro with its tail on fire, and lunged out to a 9-4 lead (with 7 draws):

But right around turn 35, the computer went on a tremendous tear, winning 6 of 7 very routinely.  From there, it thrummed me by a score that no random computer would ever be able to dream of.  

Let me read that sentence again.  You did grok that, right?  That a R-P-S computer can beat you much worse than a random computer can?

Don't go swimming for at least one hour.  But then, dive back in.


The computer almost contemptuously strolled back into a 15-14 lead with 14 ties.  The next 50 "throws" would have been about 30-10-10 to the computer, if not worse.  (Yes, I've done this before.  If it is something that nobody cool would be caught dead doing, obviously I will have done it before, many times.)

Find out why.  What was the computer thinking when it won round 44, for its 4th straight win?  It said, "I found another time when your threw PAPER, ROCK, SCISSORS, PAPER, and I noted your next throw was SCISSORS.  I will search for other times you threw PAPER, ROCK, SCISSORS, PAPER."  It scrolled down and showed me several times I'd thrown that 4-pitch sequence :- ) and then informed me of what pitch 5 had been, to follow.

It doesn't take much imagination to understand that by turn 10,000, a computer could easily be running up 98% scores against you.  And this is just a simple little NYT algorithm - it is limited to 4-throw sequences in analyzing your game.  (And here, Bobby Fischer thought he could draw God at chess.  If he had White, of course; Bobby wasn't crazy, you know.)

Anyway.  The little logic string above?  That's what the COMPUTER was thinking.  

What was DR DETECTO thinking?   Nothing!  "Oh come on I've got to be able to win ONE ... maybe rock will save me ... NOOOoooooooo!   OK, maybe rock three times in a row now!" ... NOOOOooooo..."   

Sure, I was thinking for a while.  Right up until I started getting my brains beat in.  Then panic took over.  That's what the human brain does when it gets humiliated.  It gets desperate.  Desperation is not a state of mind you want your dentist, or your starting shortstop, to be in.  Desperation is a state of mind that Lloyd McClendon is well able to grok.  How well WAR groks it, Dr. D is less sure.

This is exactly what Felix Hernandez does to a Brad Miller type... granted, Felix uses intuition rather than a granular 200,000-iteration database.  But, gentlemen, the human brain is also rather noted for its complexity and power.  How many pitches, exactly, has Felix Hernandez thrown with his pitch repertoire?  For how long has Felix been "intuiting" what a batter will be looking for?


Thirty-odd years ago, there was some post-game interview where the rookie Dave Henderson described a pitch-by-pitch at bat.  "OK, the first pitch was too tough to bunt.  So I decided to take one the other way, and leaned outside a little bit ... he busted me on the hands.  I couldn't believe it.  It was like he knew what I was thinking."

Um, yeah.

It might never have occurred to you that you have zero (0) chance against the world champion of Rock-Paper Scissors.  Nor that Mike Zunino has a real good blinkin' excuse for batting .199.  You think Logan Morrison and Dustin Ackley and Zunino are better-equipped to hit next year, than last?


There's a corollary here.  Even the most open-minded sabermetricians, including the Founding Father, argue that pitcher-batter splits mean either (1) nothing or (2) almost nothing.  On this point, kiddies, they are mistaken.  

Jamie Moyer gave up like a 1200 OPS to Bernie Williams, and if there had been another 100 AB's between the two, it woulda happened again.  Some guys, you're just comfortable playing rock-paper-scissors against.

I don't think there is anybody in uniform, not one person, who would deny that these splits are important.  "Reggie can't hit Catfish with a paddle," they'll tell you.  They're not going off the stats.  They're going off the blizzards of garbage swings.

You gotta USE the pitcher-batter splits.  But you gotta use them judiciously.


If Brad Miller could solve all that, he'd still have the problem of the James Paxton [curve-high FB] whipsaw -- the way one pitch can make another "play up."  Or the Felix sinker-changeup whipsaw.  

Or!  the FLINCH that a batter develops against such macabre torment ... and how a refusal to throw the feared whipsaw can tie a batter up worse than ever.

Life in the big city, baby.  No use losing interest in Brad Miller yet :- )

You've heard about the 1,600 ML at-bats...


You gotta PLAY the little NYT rock-paper-scissors computer.  Then, and only then, will you grok the fact that Felix can be (correction, IS) better than he would be, if his pitches were random.  You'll be able to grant that Matt was absolutely right about something, and can go right back to thrumming him as though he were a RPS customer and you the NYT computer.  Hey, what are Yankee$ employees good for, if not for that.

Chris Young is up there pitching to a rookie, and Young is thinking.  The rookie ain't.  That's all.  Comes the day the kid isn't a rookie any more.


Dr D




But ... were you playing the advanced computer or the novice?  ... the novice ID's *your* patterns only; the "expert" uses a DB of 200,000 games by all participants, so if you're an unusual player, you have a big edge early :- )


...so maybe that explains the difference? I went 150 rounds in total, started 34-19-16, and finished 46-55-49 (that would be 12-36-27 after it pegged me)

Add comment

Filtered HTML

  • Web page addresses and e-mail addresses turn into links automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <blockquote> <code> <ul> <ol> <li> <dl> <dt> <dd><p><br>
  • Lines and paragraphs break automatically.

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.


  • Allowed HTML tags: <a> <em> <strong> <cite> <blockquote> <code> <ul> <ol> <li> <dl> <dt> <dd>
  • Lines and paragraphs break automatically.
  • Web page addresses and e-mail addresses turn into links automatically.