DPPI - Top 50 Teams (Through Sunday's Games)

For comparison, here is the composite ranking of 61 computers as compiles by Ken Massey on his site.

Using this 61 computer composite, KU would still be a No. 2 seed. They would replace Purdue with Michigan St. as the 4th No. 1 seed. The tourney purposely does not use the computer rankings as a primary. The lean heavily on a formula of quality wins and bad losses. KU has the most wins agianst the top 10, top 25 and top 50. That is why they are projected to be a No. 1 seed.

CJ - what criteria does your DPPI use that causes Wichita State to rank at #17, while Cincy is #7? (Both American Conf teams - WSU has slightly better SOS). The top 3 Big 12 teams fall out as one would expect - KU, TT, WVU.

JR, First of all, I have Cincy 7th and Wichita St. 17th. The composite of all computers has Cincy 8th and WSU 13th so pretty close. And if you look at the actual rating you can see they are pretty close. A point or 2 per game can move a team as much as 10 places. Given that the average standard deviation in performance is about 10 points per game you can see that these teams are really pretty close. As for the DPPI, it is similar to most of the others in that it starts by taking the average scoring margin. Than it adjusts that by taking the average scoring margin of all opponents. Then I do that three more times to take the average scoring margin of your opponents opponents opponents. This is where the various computer algorithms start to differ. In general, they try to adjust for anything that is validly identifiable as a trend. For instance, the one common to all algorithms is venue. Every algorithm adjusts for home, away, neutral. There are different ways to do this but all make some sort of adjustment. I use a number of subroutine to adjust for how well a team performs against highly rated versus low rated teams (I call this a mental toughness adjustment but that is a bad name). I use a diminishing factor for older games as a sort of trend adjustment. In other words, the most recent game always is weighted more heavily than a game in November. Some people weight games against stronger competition more heavily though I do not. After I make all of my adjustments I calculate the consistency factor (i.e. standard deviation of a teams relative performance) and their raw rating. These two numbers form the mean and standard deviation for a probability distribution for victory. Then I run a complete 351 team round robin tournament including an adjustment for "mental toughness". So my rating is actually a measure of how many games a team would win probabilistically if they played all 350 other D1 teams. Then, just to nerd out one final time I have another probability distribution I reverse back to an average score for all 350 games so the rating is actually tied to a probably score more or less. Whew! If you follow all of that than I will be coming to you asking for help.

CJ - Thanks for your detailed reply. First of all, I believe we agree as to where your DPPI has Cincy and Wichita St - I just said WSU first. I guess the key as to differences is average scoring margin. WSU has a habit of having a 20+ point lead, then putting in the third team and losing some of it.
I like your use of diminishing factor for older games. As for me understanding much else of it, sorry, you are on your own.
I do appreciate you taking the time to explain it though.

A lot of potential change particularly in the top ten since these ratings were through Sunday.

Puke and University of Non-Compliance both lost since Sunday and they face each other one more time this week. Plus Cincy plays at Wichita State this weekend and WSU bet them last week in Cincy.

KU needs to take care of their own business to move up.

I'm a statistician and I can follow what you are doing. But I remain puzzled why Cincy is sho high. They have the 100th toughest schedule and against the Top 50 they are 1-4. They seem to be rewarded too much for not losing to the cupcakes they played.