Those #s are based on Adjusted +/- as opposed to Traditional Old-Skool +/-. The Old Skool version is exactly what you’re talking about: if Telfair is replaced by someone awful then he’d look great, but if he was replaced by someone amazing, he’d look sucky. Which makes the Old Skool +/- pretty limited in value. However Adjusted +/- compares the player to the “average” player not their backup. Meaning, the “Telfair In” #s would stay the same even if his backup was Chris Paul. Thus, except for Kevin Love, all four guys performed above average when Telfair was in. Considering how many games the Twolves lost last year, they clearly didn’t consistently play above average the entire game (&every game) or they’d have won a bunch more. So your comments that Telfair’s backups sucked is definitely true, as proved by the the across-the-board negative #s when he was out. However, if the team had the completely average backup, the #s with him off-the-court would be zero. So the key isn’t to compare the #s btw him in and out, but rather that when he was in 4 out of 5 of those players had positive #s. Rather than saying Foye was 7.5 pts better with Telfair instead of his lame backups, it’s more useful to look and see that Foye was 3.2pts better with Telfair playing instead of the average player out there.
-Alex

I don’t know if you ever compute the player triplet for a perimeter or a front line but I think the subunit data for perimeter-interior or backcourt-frontcourt combinations would perhaps be among the most meaningful groups to look at, though any player and the PG or the star would be important to.

Cheers!

(With the “without” being total raw +/- at 82 games minus the with Telfair data)