Projected Top Four Seeds

espola · Nov 19, 2012

scoachd1 said: ↑

Every system will have situations where it well perform less well that its counterparts since they are all simplifying the real world. However, these minor in magnitude compared to RPI. From what I understand almost every other mainstream ranking system will use a network to relate each team to every other team. In contrast the RPI doesn't do that. This is a fatal flaw This is why if you look at the RPI explained thread, I showed how the west could have the top 20 teams in the country, win every game from outside the region and still have teams from other regions that could rank higher than one or two of them.

The fact that the underlying algorithms are kept secret have little to do with validity. You can test the validity of a system by comparing its outputs with the expected outputs.
Click to expand...

"Expected outputs" implies another algorithm.

espola · Nov 19, 2012

scoachd1 said: ↑

The issue isn't in flaws but instead the level of the flaws are a different order of magnitude(s). Clearly there are not enough samples compared to the level of chance to get anything close to perfect ranking for soccer. However, even with something as limited as a soccer season you can see the large statistically significant error in the RPI. Just look at the difference in results between men's and women's soccer between RPI and Massey. If you have a reasonable understanding of statistics it is easy to see the difference between the two are due to the flaws in the RPI calculations. You can prove it to yourself with a deck of cards which is much simpler than soccer. If RPI fails miserably at ranking cards, you can be sure it is even worse for ranking soccer teams.
Click to expand...

"Error" compared to what?

Sandon Mibut · Nov 19, 2012