New Blog: Basketball Intelligence

[rlee]

Today's Top NBA Stories: http://basketballintelligence.wordpress ... a-coaches/

[rlee]

Today's Top NBA Stories: http://basketballintelligence.wordpress ... -schedule/

[rlee]

Today's top NBA Stories: http://basketballintelligence.wordpress ... eke-evans/

[rlee]

http://basketballintelligence.wordpress ... -team-usa/

[rlee]

Today's Top NBA Stories: http://basketballintelligence.wordpress ... k-bkn-mem/

[rlee]

Today's Top NBA Stories: http://basketballintelligence.wordpress ... ecbaelgin/

[rlee]

Today's Best NBA Stories: http://basketballintelligence.wordpress ... in-baylor/

[rlee]

Today's Top NBA Stories: http://basketballintelligence.wordpress ... -nba-sked/

[rlee]

Today's Basketball Intelligence: http://basketballintelligence.wordpress ... ee-agency/

[rlee]

Today's Top NBA Stories: http://basketballintelligence.wordpress ... phxindhou/

[rlee]

Today's Basketball Intelligence: http://basketballintelligence.wordpress ... -randolph/

[rlee]

Today's Top NBA Stories: http://basketballintelligence.wordpress ... u25scamps/

[rlee]

Today's Top NBA Stories: http://basketballintelligence.wordpress ... miaatlbos/

[Crow]

Last comment / endorsement in latest bi link of Tom Ziller's comment "subjectively created" is fair enough; but is your extension that advanced metrics are "typically lacking reliability or validity testing or correlation to winning" fair given the extensive efforts made by Jerry and Daniel, among the best and most prominent current era metric creators, to conduct these requested model tests and achieve best possible results they can? I think it is a broad brush criticism and not accurate for the named two (and at least some others). You can be dissatisfied with the results or the quality of the work or simply not want to use or see it, but it is not consistent with what I understand to say that these particular works are lacking such testing efforts. Based on my reading of the source materials more advanced metrics make these tests than don't.

[Crow]

ASPM and RAPM are estimates with expected errors in the estimates. Pretty rigorous efforts are made to minimize them. What usually gets lost in this kind of discussion (I think) is that pure boxscore metrics and even simple discrete stats have "error" in them too in the sense that there is a certain level of noise in these numbers too in terms of being an expression of true ability or future performance. Shooting percentages have noise in relation to sample size, so do simple rebounding rates given noise in opportunities, same with turnovers, steals , etc.

I am fine with being somewhat skeptical of advanced metrics and all metrics and stats; but I don't think it is even-handed to be troubled by the issues in one class to the point of significantly downgrading them or totally dismissing them and apparently not doing it with the rest as much or at all.