I think by looking at just VA data really minimizes your sample of players and recruiting rankings. You spent a lot of time on this, and I think if you got some more critical feedback from some ont you trust you could turn it into something that has real value. Your conclusion should say "Virginia HS recruiting rankings are not reliable..." because that is all the data you have - Also I still dont think they are unreliable due to attrition - thats the nature of the business/sport. Decising to offer a kid includes both the ranking and the character assessment and you cant predict injuries at all. The ranking is and should be skill based only and part of the decision to offer. I think the grey area between 4 and 3 stars you are seeing is an artifact of 1) data response "contribution" watering down any relationship and seperation (for example do a chart of Avg Games Started (Y-axis) and Stars (X-axis) by year (one trend per year) - how many years show negative trend? That is your reliability metric for those that did not attrite 2) Add in attrition and do the same thing as above - now you get a model with attrition included. 3) Look at national recruiting and see where teams were ranked in year 1 and where they finished year 1, year 2, year 3, and year 4 - one of those will have the highest correlation and be a new predictor of the value of rankings and getting high star players on your team. 4) NIL and Portaling really messes up the modern day analysis by adding another layer of data to your HS data - but if you can build the HS data then rest and add in the portal data/rankings too... Again - thanks for debating and responding - Im retired and a degreed stat guy and took courses in stats in sports over covid - this is a fun convo and your energy on this is awesome!