There are three questions which I am asked frequently about certain decisions that were made in assembling the winning percentage database:
The NCAA's official method of computing winning percentage is, in essence, to count a tie as half-a-win plus half-a-loss.
If a team has W wins,
L losses, and T ties, their winning percentage is defined by
the NCAA as {(W+[T/2])/(W+L+T)}.
A record of 6-1-3 is therefore
scored as "{(6+[3/2])/(6+1+3)}"...
or as if it were 7.5 wins out of
10 games.
Thus, a record of 6-1-3 is a
winning percentage of 0.750
(7.5/10).
It is not 0.857 (6-1), as
if ties were dropped from the computation entirely.
It is not 0.600 (6-4), as
if ties counted as losses.
When I started this database, I didn't set out to exactly match official NCAA records. Instead, I was trying to make a set of data that was more clean and consistent than the official NCAA numbers. Nebraska identified some 'exhibition' games in the middle of their season and dropped the wins from the official numbers; I felt that other teams counted equivalently strange opponents from around the same timeframe, and therefore it was inconsistent that Nebraska didn't. Houston agreed to forfeit a game to Temple, and Temple now counts the game as a win, but Houston still counts the game as a win for themselves; I felt that forfeits are not handled very well, or very consistently.
However, over time, people have come to use my data to settle arguments and for similar purposes, and it has become very confusing to people that my numbers diverge from the official numbers in some cases. As a result -- and to reduce my workload, because quality-control over the data is really more than I have time for these days -- I'm changing over to just doing my best to match official NCAA numbers.