Saturday, 28 January 2006
Why Computer Polls are Better Than Human Polls
The Courier-Journal's Jerry Brewer gets to vote in the college basketball AP poll. So, I was reading his explanation for how he voted in this week's ballot, and came across this interesting bit.
For the first time this season, I voted a team other than Duke No. 1. Connecticut is the new standard bearer. The Huskies have been lurking near the top all year, and recent victories over Syracuse and Louisville have shown they're improving right on schedule.
I moved Memphis up to No. 2 - the Tigers are 11-2 versus top 100 RPI teams this season - and slid Duke to No. 3.
After reading this I didn't have a problem with him voting Connecticut number one, but I wasa bit intrigued by him putting Memphis ahead of Duke. Especially since a Duke team beat Memphis in a head to head game on a neutral floor.
So my next thought was well, Memphis was 11-2 versus top 100 RPI teams. Maybe Duke hasn't played that good of a schedule. So I went to look up the current RPI Ratings and was a bit suprised with what I found.
| Duke | Memphis | |
| Overall Record | 18-1 | 18-2 |
| vs Top 50 | 7-1 | 6-2 |
| vs Top 100 | 12-1 | 10-2 |
| Strength Schedule Rank | 2 | 3 |
So even though Duke has one less loss, leads Memphis in every major RPI category, and beat Memphis in a head to head match up, Jerry thinks Memphis should be ranked ahead of Duke. There is no justification for that ranking. None! The only reason he made that vote was a regional bias. If he can screw up with the top teams, how accurate can his 10-25 rankings be.
This is why computer rankings are always superior to human rankings. Computers look only at the facts and don't render opinions or regional biases. Computers can consume more data than humans which lead to more informed judgments of how teams compare. Remember this when someone tries to tell you that computer rankings are bogus. No it is the human rankings that are bogus.
