Legit / useful exercise with RPM?

[Crow]

This flows out of JE's improving spm thread and has a lot in common what JE's own spm improvement proposal but to avoid disruption I'll make it a separate thread. if you found RPM for all players then took another pass where you used those values for the opponents of a team and solved again for the team itself (completely fresh or with use of the original run rpm as the prior or an spm)? What if you compared the 2nd run values to original rpm how much difference would you see and would seeing it raise useful questions, doubts or greater confidence? Is there room for legitimate fine tuning thru some further iterative process? Could you learn more about collinearity issues by doing this for runs for all data without a player on the court for all possible pairs with a given player and compare that to the run with everyone and every collinearity included? If perfect knowledge is not achievable isn't this plus what folks normally a richer pile of data results to ponder and make judgments from? Maybe equivalent chance to make misjugments but I'd lean toward at least looking at it, maybe trying it. To get an edge, some extra work and risk might be worth accepting.

[Crow]

Maybe you just get the same team player estimates under you make other changes in approach. Just searching for any way to get better estimates and encouraging those with greater understanding, experience and skill to continue that effort. RPM has advanced and probably still can further.