yeah I found some small bugs in the system. that's what I get for trying to get it up and running right before going to bed. oh well I think it is fixed now, and it doesn't really affect the top 15 teams or so.
The fans demand explanation: Ga Tech (3-0) plays UNC (3-0) -- GaTech has a .99 win% and UNC has 80%? Is this from week 3 or 4? Is this predictive or descriptive? 
This includes all stats through week 3. The win% is based on playing an average opponent. So GaTech's .99 win% means that at the end of the season (assuming from week 4 on they play an average opponent AND they maintain their current efficiency) they will win 99% of thier games.
Obviously GaTech will not play average opponents the rest of way, but this stat gives us a way to measure the strength of each team without reference to future unplayed opponents while still including the strength of the teams they have. So, by itself each teams win% is not incredibly useful other than to rank teams.
BUT if you take the win% of team A and win% of team B you can predict the probability of team A beating team B:
P(W) = (A - A * B) / (A + B - 2*A*B),
A is the win% for team A and B is the win% for team B. So for GaTech vs UNC, as of this week, the probability of GaTech beating UNC is
P = (0.998999-0.998999*0.803846)/(0.998999+0.803846-2*0.998999*0.803846) = 0.995911
Of course here is where human thought has to come into play. As previously mentioned, the SOS component in this system has not converged yet due to the lack of meaningful games played. As a result teams like GaTech who racked up huge efficiency numbers against KU is overrated quite a bit, and while WE know KU is not very good (in fact pretty terrible), the SOS can't yet compensate for the tremendous numbers GaTech put up.
So I would only take this (somewhat) seriously at this point in the season if I knew the two teams I was comparing had reasonably similar strengths of schedule up to this point in time and didn't have some crazy fluke game stats.
