Computer Ranking System

Attached to this post is an adjusted Top 25 ranking set based on a BCS-style evaluation of competitive data as well as human polling.

For reference purposes, I have used the following calculations to arrive at the final average by which the teams are rank-ordered.

Initial Computer Average:
Weights winning percentage based on combined strength and average margin of victory. Adds winning percentage for previous season's national championship tournament to account for some element of past performance.

CPU:
Standardizes Initial Computer Average based on the mean and standard deviation of that data for all teams on which data is available.

POLL:
Standardizes vote totals from both the most recent Power Poll and Prognosticator rankings for each program based on the mean, standard deviation, and given the large amount of zeroes in that data set, skewness for all teams on which data is available. For the purposes of the Prognosticator ranking, the team's ranking in the Fine Fifty is reverse-ordered and translated into a vote (1st place = 50 votes, 2nd place = 49 votes, etc.)

Final Average:
Sum of standardization of CPU and POLL data. Prior to National Championship tournament, this number will be adjusted to compensate for strength of tournaments attended by each team. These calculations have been withheld currently due to the need to adjust numbers based on continuing performance by all participating programs.

All statistical data that underlies these rankings has been compiled based on publicly posted tab summaries (via the AMTA website and Deliberation Room postings) that include all the relevant numbers. As far as I am aware, this requirement has excluded the results of the 1st Annual Norse Invitational as well as, regrettably, the NYU Downtown Invitational. If any individuals have tab summaries for those or other invitationals not available at those locations, please forward them to me so I can update the rankings for comprehensiveness prior to the resumption of the season in January.

Comments and questions regarding rankings and methodologies are welcome, as I hope to improve the accuracy and validity of the rankings as much as possible as the season goes forward, and in particular looking ahead to future seasons.DR_Initial.xlsx DR_Initial.xls

- Kevin Harrison

Tags: polls, rankings

18 Comments

Comment by Abhi Nemani on December 15, 2008 at 2:55pm: This is great stuff. Was it hard to track/compute the individual teams' records?

Comment by dork on December 15, 2008 at 3:22pm: very cool. I am sure this was hard work- and nerd's like me appreciate it!

Comment by dork on December 15, 2008 at 3:51pm: i waqs adding this to my running rankings and noticed that #14 Vandy has not recieved a single ranking, or even a SINGLE VOTE in any poll all year long. Wow. Talk about flying under the radar.

Comment by The Deliberation Room on December 15, 2008 at 3:57pm: DR Computer Rankings 12/15/08 by Kevin Harrison

Comment by Larry "The Sledge" Rosen on December 15, 2008 at 4:28pm: why stop at 25, and if you wanted to throw the subjectivity out of it, (obviously the "top" the schools don't) you could always do a straight numbers game, where wins at invitationals are assigned "quality win points", then win percentage and points for winning an invitational are assessed.

Comment by Kevin Harrison on December 15, 2008 at 11:22pm: The difficulty of it lied not so much in calculating the individual records for the teams, but compiling the CS and point differential data. I have a new-found respect for the creators of tab summaries that already provide that data without requiring the viewer to calculate it themselves.

With regard to stopping at 25, I have data on every single team from every single program that has competed this year at tournaments for which tab summaries have been made publicly available. I may expand to 50 programs for the next update of the rankings if people would prefer it. The idea of the CPU calculation is to make it a more straight numbers game, but I wanted to weight the numbers and the human element equally in this poll. I plan to re-weight either just before the National Championship tournament, or during winter break with an RPI-style calculation for tournament strength that would weight the 'quality' of ballots. My initial idea for this is to find the average amount of ballots won by all programs that attended each tournament to make some sort of multiplier or average, though I have yet to decide whether this would represent a separate one-third of the final average of would be included in the CPU calculation.

Thanks for the suggestion though!

Comment by Abhi Nemani on December 16, 2008 at 6:22am: Are there/should there be automated ways to calculate CS? (I'm largely unfamiliar with the magic of the tab room.)

Comment by Daniel Vaughan on December 16, 2008 at 10:36pm: Question, are you talking Vandy, MTSU or some other team in that 14th spot? The comments could indicate several teams, GA Tech, MTSU, and UAB all had better records than Vandy at MTSU... No disrespect to Vandy, but they hardly ruled the MTSU tournament. MTSU won the tournament and GA Tech had 1st and 2nd place teams (in the divisions).

Oh, and to answer the question in the Georgetown comment box, its because they are that good. Top 10 team for sure. Their Duke showing was impressive.

Overall though, nice poll. I like this a lot better than the other polls which are subject to a lot of bias. I love the fact that CS and PD were factored in. Very good job.

Comment by Kevin Harrison on December 17, 2008 at 3:04am: To clarify I'll have to re-post an updated rank set. A bad keystroke on Vandy's ballot left their C team out of the calculations, which when corrected confirms my revisiting of the tab summary. As a result, Vandy slips to 31st.

Thanks for the catch!

Comment by Kevin Harrison on December 17, 2008 at 3:16am: Rankings have been double-checked and revised for accuracy. DR_Initial-Revised.xls

Computer Ranking System

You need to be a member of The Deliberation Room to add comments!

Search