About System Accuracy

A positioning or ranking system is inaccurate when a team with a lower ranking has defeated a team with a higher ranking. Each instance of this is called a "Ranking Violation" in Massey's listing (www.mratings.com), and an "upset" in GridMap. The fewer the number of RV's in a system, the more actual game results the system correctly reflects, and thus, the more accurate the system. Massey tracks these RV's for each system, and reports them as a percent of all games played. For any system to be 100% accurate would be an extremely rare situation. The simplest way to demonstrate why this is so, is to consider the following example: team A defeats team B, team B defeats team C, and team C defeats team A. This situation is what GridMap calls a “loop”. Note that any hierarchical ordering of these teams will include, perforce, one inaccurate game result, where a lower ranked team defeated a higher ranked team. (Though it is possible in this simple situation to assign all 3 teams the same ranking, which Massey's statistic does not count as an RV, this occurs infrequently in subjective systems, and rarely in quantitive systems. And though it has occurred in Gridmap, GridMap counts such a result as an upset, and reports it to Massey as such.) During a typical season, there are usually dozens of these loops. They are most obvious within conferences, but are common in interconference play as well, often involving 10 or more teams. GridMap's accuracy is based on identifying these loops, and "resolving" them (determining which team will be positioned at the top of the loop) in the most logical manner. Another feature of GridMap is that it identifies and displays every loop and it's resolution, a practice that assures no hidden basis for, or bias in, the resolution of loops. No other system identifies its RV's, so the basis or bias in their ordering of teams involved in loops cannot be easily determined or challenged.