THE BOOK--Playing The Percentages In Baseball

You make a very good point, but there is plenty of uncertainty abound in basketball statistics. When you live in Berri’s world of wins produced, though, you tend to overlook that sort of thing.

I’m no expert in baseball, but my sense is that the pitcher/batter matchup is easier to predict than opposing lineups in basketball.

I agree with the general point, but is there any objective way to determine that a 48 minute basketball game is “longer” than a 9 inning baseball game?  That is, other than measuring the inconsistency of statistics, which, as commenters to Berri’s blog suggest, may be partly caused by other factors.

Holy cow, thank you. It drives me batty on that site when he constantly refers to basketball players as “more consistent” than football players… as though there’s some difference in the type of person who plays each sport… when instead it’s simply the nature of the sport and the sample size of the events that are being recorded and measured.  I comment this point to him every time he does it, but get no response and he continues to write it.

I think MGL is fond of pointing this out - the more accurate a measurement is, the more variance will be observed (all else being equal). We see a smaller spread in observed talent in Range Factor than we do in UZR, not because RF is better than UZR but because it isn’t.

Berri seems to be confusing issues with his measurement system with the underlying data. It’s obvious that the play-by-play data we get from Retrosheet is vastly preferable to the box score data he deals with in basketball.

(It’s possible Berri acquires this attitude from defending his Wins Produced system; I’m no expert on basketball but everything I’ve learned about regression models in baseball tells me that they’re tools of last resort.)

And Jim, sure there is. Look at FGA versus PA - that’s our fundamental unit of observation in each sport.

Jim, yes I had a long thread on this a few years ago.  Do a search for
True Talent Sports Leagues
and you’ll find it.

IIRC, a 32-game season for basketball tells you as much as an 82-game season for hockey, 162 for baseball, and 28 for football.  Something like that.

Pretty fascinating stuff, and I’m surprised it doesn’t get more play.

***

I don’t follow Berri too much.  I get the impression he loves to tout his stuff.  I prefer that an honest pro/con be presented on anything created.  We can discuss all the shortcomings of UZR or wOBA or WOWY or FIP or Marcel here, and I’d he happy to do that.  It doesn’t seem that that’s the case in too many other places.

The creators of their metrics must realize how foolish they look when they present a one-sided view.  The discerning reader doesn’t put up with b.s., and it’s the discerning reader that should be their lifeblood.  They’re the only ones keeping them honest.  Anyway, I don’t want to sound all high and mighty about it.  A little honesty is all I expect from the other metric-creators.

This looks like the thread:

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

Too many equations for me right now, but it looks interesting.  I’ll look through it tomorrow

Tango, #5:

My impression of Berri matches yours.  I used to read him more regularly, until I became a little more familiar with other basketball analysts (Pomeroy, Hollinger, etc), though I’m only about 5% as knowledgeable about advanced basketball analysis as I am about baseball and football analysis.

I can’t recall an instance of Berri ever being surprised at the results of his metrics.  Or acknowledging that it may underrate or overrate certain players or types of players.  It’s always that he’s right and that anyone whose analysis disagrees, whether it’s a fan, fellow analyst, or GM, is incorrect.  It’s caused me to only check his blog every few weeks, as opposed to a more daily basis.

Off topic, but you just fixed NBA basketball.  Only play halves.  If a team is ahead after two quarters, they win.  You can even have doubleheaders, play one two quarter game, then take a break, then a second game.

My own idea was to determine who won by quarters, if a team won three quarters they won the game, with games lasting three, four, or five quarters depending on what was necessary.  But this idea is simpler, and better.

Ed, what a fantastic idea you have of my idea!

Every game is a double-header!  I love it!  LOVE it.

Understanding that there’s end-of-game strategies that come into play, can someone figure out the win% presuming the game restarted at each half?

You can even stretch it out to think of it like tennis.  Each half is a “set”, and first one to win two sets wins the match.

The third set is 5-minute OT.

So, if each team wins one set (they each win a half), then they play OT to see who wins the match.

Those are great ideas, especially the two-set variation. You could also make each quarter a set with the first team to 3 winning. I wonder if you would need to make other adjustments like the number timeouts and fouls.

One of the things that drives me nuts about basketball (and there are many) is that there is no point in watching a game until the final 2 or 5 minutes. There is just so much scoring and the action so repetitive that it becomes mind numbing. Anything that adds value to the earlier part of the game, especially the first half, would make it dramatically more appealing to me.

There would be some interesting strategy in play.  If you are losing the first quarter say by 9 points with 2 minutes to go, the losing team may sit all their starters and play their bench.  The leading team will be deathly afraid of blowing it, so they will keep their starters on the court.

Now when the second quarter starts, the losing team will be fresh.  A best 3 out of 5 periods, with the 5th period being half as long (say 6 minutes) would be pretty exciting.