Originally posted by panther17
Record is not factored into this ranking.
Think of it more as who is the most dangerous teams to play against and not as a ranking.
In other words, it is a "Power Ranking".
Wins and losses do an excellent job of ranking how many games a team has won and lost. They tend not to account for strength of opponents, or general roster strength. Therefore, they are somewhat inadequate as predictors of future results. This is where Power Rankings come in; Power Rankings are alternate rating methods that operate independent of win/loss records, and attempt to evaluate the overall quality of a team at the current moment (rather than measuring their past performance). Power Rankings can take many forms, and effective ones can prove highly valuable as predictive devices.
In this case, we happen to have a game where every player comes pre-rated with an individual measure of power: level. I have developed a Power Ranking formula that compares the levels of all the starting players on all the teams in a conference (or league) and uses them to provide a Power Rating for each team. I refer to this rating as the "Roster Rating", and it fluctuates continually as players level up or boost, and as teams edit their starting lineups. The Roster Rating is nice, but it obviously doesn't provide a complete picture. It ignores backup players and depth. It ignores chemistry and energy, it doesn't account for unusually good or bad player builds (not every level 13 player is equally good), and it ignores the effect of good or bad game planning. Therefore, my rating system has a second component, the "Upset Adjustment". In an effort to account for all the intangibles above, I keep track of every game over the course of the season in which the higher rated team loses to the lower rated team, based on their raw Roster Rating at the time of the game. Any time this happens I take it as an indication that the Roster Ratings of those two teams was flawed, so I deem the game an "Upset" and calculate a rating bonus to be added to the lower rated team's rating, while equally penalizing the higher rated team. The size of the bonus/penalty depends on how large the rating gap between the two teams was, and it is ALWAYS based on the raw Roster Ratings at the time (ignoring any previous adjustments the teams involved may have earned).
The Official Ratings that I post every two days, in the hours before the next game kicks off, are a combination of the Roster Rating at that time, and the accumulated Upset Adjustments a team has earned over the course of the season. Obviously, it takes time for the ratings to come together, but the assumption is that if a team is better than they look on paper, they will over time earn positive upset adjustments, while a "paper tiger" team that is weaker than they look, will over time earn negative Upset Adjustments. By the end of the season, the combination of Upset Adjustment points, and Roster Rating at that time, will hopefully provide an accurate picture of how dangerous each team is at that time. I look forward to testing this theory by seeing how well my rankings predict playoff results.
Until then, yes, the rankings are flawed, though not for the reason zgreech claimed. All ranking systems are flawed. Hopefully though, you can still glean some useful or interesting information from them, and enjoy them for what they are, which is merely a means of adding depth to this game we all love.
(PS - I appreciate getting this question over and over, since it gives me the opportunity to continually refine my response. Since I'm planning on posting my ranking formula on the GLB boards before season 3, once I've refined it with a season of testing, I'm eventually going to have to write a FAQ. This is certainly the most frequently asked question, so it's good to practice responding to it. What do you guys think? How was this version of the answer?)