KU Season Projection

I have initiated my DPPI Rating system. I had to include some code to shadow link the teams since they aren't fully linked at this point. It is not particularly accurate at this point and won't be until at least 7 or 8 games have been played. Statistically they should be fully linked after 9 games. But this isn't completely without merit either. The good news it projects KU to win 28.5 games through the conference regular season (31 games). I fear that is optimistic given that teams like West Virginia and Texas Tech are off to rough starts and are ranked lower than is probably what they will ultimately be. Based on the DPPI, the next two games will be the toughest of the season. Agains, I'm not sure that will hold up but clearly, if we can win the next two games we will be in excellent shape.

Yeah, the DPPI after only three games played is pretty sketchy. The math is valid but the statistical validity is sketchy. Utah state is an example. They have won 4 games by big margins and are currently projecting as the 2nd best team i the country. But their SOS is so-so at best and since those poorer teams are not yet connected fully, they get way more credit for those big margins against weak competition than they should. It works the other way as well. West Virginia is in the crapper right now because of their 2 losses. As for the 100% for Louisiana, that is actually rounded up. It is 99.999%. That is because of the standard deviation adjustment. Right now, KU has a standard deviation of about 1.35. That skews the probability curve pretty dramatically. In other words, as the projected scoring margin increases the probability increases exponentially. Louisiana also has a low standard deviation. So even though the projected margin was 14.53 points, with standard deviations so small it projects that the team will not play too far from their projection. That is why the projected margin can be identical for different teams but the probability can be quite different. Even if the ratings don't change as the season goes along, if KU has some up and down games and the standard deviation increases, all the probabilities will decrease. Some of the other RPI guys make early season adjustments to the standard deviation calculation based on past seasons. I intend to someday when I get around to it but right now I don't. It will all work out by mid-December on its own so I don't worry about it too much. Although now that you made me think about it, I may add a module to do that. I do an adjustment to shadow link all the teams based on last season's rankings and that affects the rating. Changing the standard deviation would not affect the rating but it would affect the probability of winning. Probably more than 99% of people want to know.