THE BOOK--Playing The Percentages In Baseball

Friday, June 11, 2010

People who like Anthony Giacalone and Joe Hamrahi also like…

I am a numbers whore.  There’s literally no time that I would look at a number, and not think of doing something with that number.  I’ll even count the number of seconds a traffic light will stay red, or how many paces from my car to the office door.  It’s something I do, because numbers make sense to me, they give me comfort, it puts a level of order in my head.

A few weeks ago, I wrote a quick article on ESPN about instant run-off, a voting system used for the Oscars.

The run-off system works like this. You have a group of say five candidates, and whichever one gets the fewest first place votes ends up bowing out of the race. Then people who voted for that candidate shift their support to one of the remaining four. And the process repeats itself until someone gets more than 50 percent of the vote. Basically, an instant run-off system simply means you have to rank your favorites beforehand, so that the whole process can move along faster. Rather than deciding after the fact which politician you will support after your favorite has been removed, you decide before.

I applied it to the NL Cy Young, where Wainwright got 12 first place votes, Lincecum 11, and Carpenter had 9.  With 17 needed to win, you lop off Carpenter’s 9 votes, and go to those voters’ 2nd place votes.  Seven went to Lincecum, and that’s all she wrote.  So you can see that those who liked Carpenter really preferred Lincecum to Wainwright.

Amazon and other merchants also do the “customers who purchased ... also purchased ...”.  It’s all about trying to find out profiles of people.

Anyway, SABR released their voting and they used the preferential system, which is the same as above.  After the write-in votes, the first person lopped off was Anthony.  A disproportionate number of his 128 votes went to Joe.  So, you can see that people who like Anthony also like Joe.  Joe was the second guy lopped off.  Right away, you can see that there’s a faction of people who are simply Anthony/Joe fans, basically the new kids on the block, those who we know via the internet.  The remaining people on the ballot I have never heard of.  Which ones might we internetters go for?

Well, we can see where the Anthony/Joe votes went.  After those two guys were lopped off, this is where their 260 votes went:
67 Leslie Heaphy
48 Tom Hufford (incumbent)
44 Steve Krevisky
44 Paul Hirsch (incumbent)
35 Bruce Brown
22 (Discarded)

We see there’s really no incumbent-bias.  And Leslie Heaphy would likely have been the one to appeal to us, but there was not really that strong a preference.  It worked out fairly random.

Then Steve Krevisky’s 217 votes were redistributed, and it was again the same story:
53 Leslie Heaphy
43 Bruce Brown
40 Paul Hirsch (incumbent)
38 Tom Hufford (incumbent)
43 (Discarded)

Paul Hirsch’s 316 votes were almost evenly split between the three remaining (and discarded).  Tom Hufford’s 406 were evenly split between the final two.

You can also see a problem with the quota, which was fixed at 732 (half of 1464) even though there were dicarded ballots piling up (people did not list all members in the list).  When you are down to the final 2, well, you can already declare a winner.

Then it gets interesting, as this was an “elect 2”.  So, the winner’s original 294 votes disappear, and are redistributed to everyone else.  And again you see the split, with three candidates getting 56-64 votes, and Joe/Anthony getting 61 combined.

(There also seems to be a problem, as cell AK37 should show, at least, 4 exhausted ballots, the carryover.  Unless, somehow, people left the first entry for the elect-2 blank, and voted for the 2nd person, which is kind of strange, but I guess it could have happened.)

I was hoping to play around with the IBA voting in the same manner, and if someone at BPro can help out, it’d be much appreciated.