College Basketball Peer Rankings
The rankings are a relative look at what peer teams think of each other. In many ways this relative ranking is similar to other opponent, or opponent’s opponents ranking/rating systems. However the granularity of the comparison is not solely based upon margin of victory. The margins of victory are sorted into bins so that a particular team can “judge” its opponents. It removes the pure linear, or even logarithmic relationship of margin of victory. The relative relationships per team do NOT consider the team that is “judging.” Once all of the relative relationships per team are computed, an iterative computation of rankings is calculated until it is stable.
The binning of results emphasizes the importance of close games and winning; and it makes the results of lopsided wins and losses less important as a discriminator between teams. That is, say Team A beats Team B by 25 points, and Team C by 35 points. A simple margin of victory analysis says that Team B is 10 points better than Team C. But the Peer Analysis discounts the large margin of victory and says that Team B and Team C are equals according to Team A. And so let’s also say that Team D loses to Team B by 5 points and beats Team C by 5 points. And so Peer Analysis shows that according to Team D, Team B is indeed 10 points better than Team C.
There is an adjustment to the results based upon the game’s location – home, road or neutral site. The adjustment is one bin for home/road and no adjustment for neutral.
There is also an adjustment to each result based upon pace of play. For example, a 10 point differential in a 100-90 game is far different than a 10 point differential in a 50-40 game. The pacing of the game shows that 10 points isn’t the same effort in both cases.
The results of these games are compiled for all such possibilities, or permutations. These permutations are run through a mathematical model that computes the relative strengths of all teams each day with a random start value. The repeat-ability of these relationships (final rankings) is very important, and should be consistent – and they generally are a couple of weeks into the season.
College Basketball Peer Voting
The Peer Voting method sorts each game into “bins” according to the margin of victory and location of the game. Each team keep track of its own opponents, and provides a vote of what it thinks of its own opponents. This vote is provided based upon the bins. And it cannot vote for itself. The number of points assigned to the vote is based upon the number of games it has played. For example a team that has played 9 games, it assigns 9 points to its top ranked team, 8 to the second, etc. A maximum of 25 teams ranked by each team. So later in the season, the lower ranked teams on the list will not get points.
A compilation of all teams’ votes is made and that is what is shown.
The College Basketball Peer Ratings Evaluation Data is a product that categorizes the wins and losses each team has during a season. This categorization gives another perspectives into how many quality wins and bad losses each team has, and generally serves as a way to distinguish between teams.
Various rating/ranking systems have done this for a while, including this NCAA. This has traditionally been in the form of win/losses against teams grouped by ranking – i.e. 1-25, or 26-50, etc. There was no distinction whether it was a home/neutral/road game. This was problematic – a home win against a team ranked 26-50 is much different than a road win against the same ranked team.
The NCAA this year (2017-18) has started making Quadrants (four of them), which groups wins by ranking and game location. Other ranking systems have been doing this for a while.
I have done the same thing here. I’ve made four Tiers and then the catchall BodyBag games. It also distinguished between the ranking of the teams and the game location. Each Tier has a home, road and neutral game component. For Tier I, home wins against teams ranked 1-25 are counted, along with neutral games against teams ranked 1-35, and road games against teams ranked 1-60.
I have done a detailed analysis to find the breaks between Tiers based upon game location. I have determined the best break point so that the relative difficulty to get a win based upon game location is the same within each Tier. So in Tier I the home 1-25, neutral 1-35 and road 1-60 are all about the same difficulty. The emphasis for this Tier system is to evaluate the Top 75 teams.
Here is the various Tiers and the break points for each:
Credit: Massey’s Ratings for the game data.