Quote:
Originally Posted by Coach Tom
Sorry for the simple question, but how is OPR calculated? As an example, I took the sum of 1114's Waterloo qualification round scores and divided that by the number of matches and got 80.6 -- yet their OPR for that event was 71.88. What is the algorithm? Do you count the minibot?
Others who are not familiar with this metric might benefit from knowing!
|
The simplest way to explain OPR (Offensive Power Ranking) is that it is the solution to a linear systems. The equation for each qualifying match is a sum of the OPR variable for each alliance partner equals the match score (minibots and penalties included) like this:
(OPR of Team AAAA) + (OPR of Team BBBB) + (OPR of Team CCCC) = (Match score for that Alliance)
Once you you play enough matches, and teams play with different partners (connecting a graph) you can solve this linear system for the OPR variable of each team. This is actually done in the program OPRNet using LU decomposition.
OPR seeks to find the "value" of each individual robot in a value that is similar to points but actual points. Negative OPRs occur in every event calculation, which indicates generally infective and penalty-prone robots. Average qualification score is not a good metric because it measure on how good your randomly alliances are not your individual robot. The team with the highest OPR is the "MVP of the Quals". Much like Ernie Banks, this MVP could always be on a below average randomly paired alliance (producing points when paired with low OPR partners helps your OPR).
As I said before OPRNet doesn't use Elim matches in the qualifications. You actually cant solve for OPR with just elimination matches, since alliances are always paired together (no inter pairing between robots on different alliances to make the system solvable). Also there is nothing mathematically that links different events together. Any comparison between OPRs at different events is interesting to look at, but should be taken with a grain of salt.