What Inspires Human Polls? | Syracusefan.com

What Inspires Human Polls?

Quazzum69

Stable Genius
Joined
Oct 27, 2012
Messages
3,921
Like
6,107
upload_2014-1-20_14-5-13.png


Here is this week's AP poll and a comparison to three popular ranking system: Sagarin, Sagarin-Elo, Pomeroy and an average of all three. I suspect most (credible) ranking systems are fairly close to these three so it should be fairly close to the true disconnect between human polls and computer rankings.

In my opinion, if a voter is not close to an objective system, it is a clear indication of bias and/or conspiracy to maximize profit via top-25 matchups. There is no stronger evidence for this than the fact that Louisville, Duke, Florida and Kentucky all get a pretty significant bump from both the average and median of these three computer systems.

As can be seen from the R2 values, the predictive systems (Pomeroy and Sagarin) are slightly closer to the AP poll than the Elo, which is interesting because Elo is based on the actual outcome of the games. Since there's no real accountability, a voter can pick and choose which system (predictive vs what a team has accomplished) to move teams around but still have an objective system to use as evidence for their ranking. So a voter could put Duke at 15 because they are very efficient (high in Pomeroy, but low in Elo) but put UMass at 20 because they are not as efficient as Duke even if they have a better resume.

It is quite ironic that the numerous objective ranking systems originally inspired to combat subjective human pollsters may be actually be increasing subjectivity by giving voters such a wide range of empirical evidence to draw from.

The most overranked teams include:

Louisville (12), Duke (18) - 5 spots higher than 3-system average
Florida (6), San Diego State (7), UMass (13)- 8 spots higher
Kansas State (22) - 9 spots higher
Memphis (23) - 10 spots higher
Kentucky (14) - 11 spots higher
Baylor (24) - 20 spots higher

Some underranked teams are Iowa(10) at four spots lower, Pittsburgh (20) at seven spots lower and Michigan (21) also seven spots lower.

It makes you wonder why Baylor is still clinging to a top-25 ranking (maybe the game with Kansas tonight ... more eyeballs for a "top-25" matchup?) even though they are in the 40s for each system. Obviously the big-time basketball programs get a bump (Louisville, Duke, Florida, Memphis, Kentucky).

Even something has seemingly inconsequential as a weekly poll in college basketball manages to get corrupted by money.
 
The "I watch basketball" is good for the top 5 and better in terms of the way people want to frame rankings.

The "I watch basketball" rankings start to sputter by around 7 or 8, has almost failed by #15, and is utterly useless by #20. People may able to assess some teams outside of the top 10, but they don't watch enough of every team to sort the teams properly and are unable to avoid ore-concieved biases related to conference association.

And for teams in that 25-50 range (which is relevant for tourney level teams and how they are tracking)
- We have no AP ranking
- People have not seen enough of most teams, so "I watch basketball" is useless. "I watch basketball" folk assess these teams typically on only a few results
- Teams are highly assessed by bias (conferences, prior year teams)

So if you want to try to measure those teams you need to use an unbiased ranking system. What the are you going to do.
 

Forum statistics

Threads
170,113
Messages
4,871,836
Members
5,988
Latest member
kyle42

Online statistics

Members online
109
Guests online
1,201
Total visitors
1,310


...
Top Bottom