In what universe is the win/loss record of a team not the most fundamental statistic? And where averages are not considered a statistic? Does your metric exclude those 3 RPI elements? I think the word "joke" might apply there. They may not be sufficient, but they are significant. You start with W/L record. In a closed system (where everyone plays everyone) that is sufficient and how post season play is determined. In an open system, where there are way more teams that can be played you have to figure out a way to give the teams a relative rank. You quickly see W/L record is not sufficient. You have to include the strength of the teams that the W/L record was achieved under. But, then you realize that you need the secondary impact of strength of schedule or someone playing a team like Grand Canyon will boost your strength of schedule without considering who Grand Canyon has played. Do you recall UNC starting off the year 17 - 0? What did everybody know? That it was against a weak schedule. And sure enough, they finished the season 15-19 when they started to play better competition. If there is a subjective component in the RPI it would be the weighting of the 3 components. You could stress test that by changing the weightings and seeing how the teams shift. But, for sure, all 3 components of the RPI are important. After that, you can add in scoring margins. You can see the importance of that. And it goes from there. Can you share one of the linear regression models? Just dependent and independent variables, not coefficients. Yes, a grad student could come up with models. But, you have to be able to communicate them. Coaches need to be able to have a clear understanding of what drives the metric so they can consider that in scheduling. And fans and analysts need to be able to discuss them. One positive of the RPI is that it is easy to understand. Yes, the NET is a more robust model. But, you know, that metric is NOT the sole determinant of NCAA tournament seedings and selections. It's an input.