Read on Twitter
. @tableaupublic authors should viz this month's #SportsVizSunday data, all NBA shot locations since 1997.
BUT @SethPartnow showed me some data pitfalls
Briefly:
- "Action Type" is less accurate on misses than makes.
- You should filter out pre-2000 seasons.
Details in thread...
BUT @SethPartnow showed me some data pitfalls
Briefly:
- "Action Type" is less accurate on misses than makes.
- You should filter out pre-2000 seasons.
Details in thread...
"Action Type" describes type of shot. Stat-keeper's dilemma: How do you gauge intent to bank a shot, particularly if it missed?
@SethPartnow: "Action type, especially in earlier years.... is biased towards the “flavor” descriptors being include much more frequently on makes"
@SethPartnow: "Action type, especially in earlier years.... is biased towards the “flavor” descriptors being include much more frequently on makes"
Wish I had realized the problem with Action Type in these #sportsvizsunday data before spending an hour on the following rough dashboard draft (annotated to prevent out-of-context sharing and warn others against making a similar mistake):
The issue with earlier seasons, per @SethPartnow "the location data is sketchy until about 2001-02, with 97-98 being borderline unusable."
cc @JSBaucke @sportschord @SimonBeaumont04 who run #SportsVizSunday
Those who published dashboards already should probably re-examine them.
cc @JSBaucke @sportschord @SimonBeaumont04 who run #SportsVizSunday
Those who published dashboards already should probably re-examine them.
