Gophergrandpa
Well-known member
- Joined
- Oct 17, 2018
- Messages
- 4,308
- Reaction score
- 5,379
- Points
- 113
You would also see that the cloud of data points was quite dispersed rather than tightly located along the trend line. Correlation, yes; predictability, 50/50; conclusiveness, other than at the most positive and negative outlier data points, no.If you had some reasonable measure of "goodness in college" (however it makes sense to measure that) on the Y-axis, and the star rating on the X-axis, and then did a best fit of a line to the "cloud" of datapoints on that plot, you would see that the slope was positive.
IE the basic trend is that higher star rating correlates with higher "goodness in college".
But beyond that, it is tough to infer much.
Remember that regression analysis—which is used to calculate the trend line—doesn’t purport to find the “right” answer; it uses mathematical analysis of data to find the “least wrong” answer assuming that all data inputs are correct—e.g. assuming that Antoine Winfield is correctly rated less than 0.83 and that is Mo Ibrahim is correctly rated less than 0.84, etc. Because there are so many instances of initial incorrect data entry (proved by later success or failure of a player), the cloud of data might be more tightly wound around the trend line if a service were to re-rate players in their third or fourth year of college eligibility (where Antoine and Mo would be counted as solid four stars in the 0.95 range).
Here’s how PJ has managed to improve recruiting markedly (judged by product on the field) without moving the aggregate preliminary numbers as greatly as those enthralled by preliminary numbers would like: he uses the eye test to find athletes that are underrated, under-analyzed or not being given full credit for their development potential … and players whose heart and drive are not reflected in numerical rankings. He doesn’t have to fight the helmet schools for these kids, and fewer of them turn out to be prima donnas. It is IMHO the best formula a coach can use in a non/helmet school setting.
Last edited: