Oh, joy, rapture!
My latest attempt at faking math:
Tuesday night, mostly out of idle curiosity (and procrastinating on the weekly feature and the weekly Power Rankings), I tried running some numbers with Bill James' Pythagorean method. James had the idea that a baseball team's expected winning percentage comes out very close to its runs scored, squared, divided by the sum of that number plus runs-allowed-squared. [RS^2/(RS^2+RA^2), if that helps you visualize. I needed finger puppets.] Over the years, much better mathematical minds than mine (real ones, in other words) have refined and honed it to make it more closely mirror real life.
Better hockey minds than mine found that something similar happens in hockey, substituting goals for runs. But the shootout and the point-for-an-OTL mess it all up: bonus points galore. And I'm nowhere near good enough to refine, develop or hone it.
So, I got the noise out of the way, sorting out each team's shootout record and taking those goals out of the equation. Then, I reverted each team's record back to good ol' fashioned W-L-T (using OTLs as L's and shootouts of any kind as a T*).
So the first set of numbers generated a Pythagorean projection based on the teams' GF/GA. And then the second set of numbers yielded a "real winning percentage," free of bonus-point influence**.
Now, all of this is a still a small sample, but from what the small sample kicks out, updated now through Friday's games.... The most striking team right away is Hamilton, whose 103 "real" goals for and 73 "real" goals against put them behind only Wilkes-Barre and Chicago in expected winning percentage. The Bulldogs have actually moved up to sixth or seventh overall in actual performance, depending on your measure, as they climb out of their early hole.
The top five (predicted winning percentage/"actual winning percentage"):
1) WBS .700/.706
2) CHI .680/.710
3) HAM .666/.581
4) HER .648/.700
5) NOR .646/.672
Which teams were "overachieving" the most? Subtracting "winning percentage" from Pythagorean expectation, here's the top five:
1) Peoria +.097
2) Springfield +.079
3) Rochester +.063
4) Hershey +.052
5) Omaha +.050
Which only caught my eye because of who's in those cities. The downside is the math says they're due for a slump, then... But even Peoria's plus-percentage, like Hamilton's minus-percentage of .085, isn't even a six-point swing to date.
And interestingly, after Friday's games, the two teams closest to the GF/GA relationship: Worcester (98 GF, 98 GA, 12-12-5) and Bridgeport (79 GF, 88 GA, 10-13-5, .446 all around).
*-And we've discussed the danger of this assumption, because bonus points change the way teams play the game, blah blah blah, don't step on a butterfly in the past, blah blah blah.
**-See above, blah blah blah, tie games would cause totalitarian government and funky spelling and maybe donuts falling from the sky, blah blah blah...***
***-"This is indeed a disturbing universe." (2F03)
Tuesday night, mostly out of idle curiosity (and procrastinating on the weekly feature and the weekly Power Rankings), I tried running some numbers with Bill James' Pythagorean method. James had the idea that a baseball team's expected winning percentage comes out very close to its runs scored, squared, divided by the sum of that number plus runs-allowed-squared. [RS^2/(RS^2+RA^2), if that helps you visualize. I needed finger puppets.] Over the years, much better mathematical minds than mine (real ones, in other words) have refined and honed it to make it more closely mirror real life.
Better hockey minds than mine found that something similar happens in hockey, substituting goals for runs. But the shootout and the point-for-an-OTL mess it all up: bonus points galore. And I'm nowhere near good enough to refine, develop or hone it.
So, I got the noise out of the way, sorting out each team's shootout record and taking those goals out of the equation. Then, I reverted each team's record back to good ol' fashioned W-L-T (using OTLs as L's and shootouts of any kind as a T*).
So the first set of numbers generated a Pythagorean projection based on the teams' GF/GA. And then the second set of numbers yielded a "real winning percentage," free of bonus-point influence**.
Now, all of this is a still a small sample, but from what the small sample kicks out, updated now through Friday's games.... The most striking team right away is Hamilton, whose 103 "real" goals for and 73 "real" goals against put them behind only Wilkes-Barre and Chicago in expected winning percentage. The Bulldogs have actually moved up to sixth or seventh overall in actual performance, depending on your measure, as they climb out of their early hole.
The top five (predicted winning percentage/"actual winning percentage"):
1) WBS .700/.706
2) CHI .680/.710
3) HAM .666/.581
4) HER .648/.700
5) NOR .646/.672
Which teams were "overachieving" the most? Subtracting "winning percentage" from Pythagorean expectation, here's the top five:
1) Peoria +.097
2) Springfield +.079
3) Rochester +.063
4) Hershey +.052
5) Omaha +.050
Which only caught my eye because of who's in those cities. The downside is the math says they're due for a slump, then... But even Peoria's plus-percentage, like Hamilton's minus-percentage of .085, isn't even a six-point swing to date.
And interestingly, after Friday's games, the two teams closest to the GF/GA relationship: Worcester (98 GF, 98 GA, 12-12-5) and Bridgeport (79 GF, 88 GA, 10-13-5, .446 all around).
*-And we've discussed the danger of this assumption, because bonus points change the way teams play the game, blah blah blah, don't step on a butterfly in the past, blah blah blah.
**-See above, blah blah blah, tie games would cause totalitarian government and funky spelling and maybe donuts falling from the sky, blah blah blah...***
***-"This is indeed a disturbing universe." (2F03)
posted by Fornabaio at
5:43 AM
0 Comments:
Post a Comment
<< Home