by Nicholas Beaudrot of Electoral Math
Comparing the top performers in WP-48, which is a rate statistic, to the top-PER performers, which is dependent upon playing time (as well as how many opportunities your coach and teammates give you handle or shoot the ball), is definitely an apples-to-oranges comparison, roughly akin to comparing total rushing yards with yards-per-carry. A better comparison would be the All-PER team to the All-PAWS team, which consists of either Jason Kidd, Kobe Bryant, Kevin Garnett, Dirk Nowitski, and Marcus Camby; or Kidd, Kobe, Shawn Marion, Garnett, and Camby (Dirk is the only "power forward" in the top 30). Just eyeballing things, the all-PER team scores more (except at the 2-guard), while the all-PAWS team does more "other stuff" (rebounding, fewer turnovers, etc.) while still scoring at a healthy clip. Update: a commenter points out that PER is in fact a rate stat. Surprising, because I've rarely seen PER formula put a high-quality bench player was as effective as many starters (Zach Randolph in '02-03 is one of those cases). PER seems to downgrade David Lee and his off-the-chart field goal percentage because he only scores 17.5 or so points per 48 minutes, compared to 27 for Garnett and 36 for Yao Ming (!).
As for game-to-game consistency, there's always the old "hot hand" paper, and the fact that a single game's sample size is too low to glean any real information. Given 18 shots in a game, it's impossible to tell the difference between someone who shoots 45% and someone who shoots 47%. It's highly unlikely, therefore, that shooting performance in game n has any meaningful correlation with shooting percentage in game n+1, though obviously Real Data would be nice.
