Duo Q with a ~1550 and a ~1700
Win => +14 +11
Win => +13 +11
Loss => -14 -16 =>Queued with a 2100 and a 2k player against ~1750's. The 2k+ players were BAD and caused the loss, and the 2100 player lost 16 rating.
Win => +6 +6 => Elo range was fairly even and we barely pulled off a win.
Win => +12 +11 => Was queued with a 3-man at 15-1550 elo. Didn't check the other team, but they were pretty easy.
Loss=>-12 -13 => MM stuck us with a 2050, 1940, and a 1665. Enemy team was 3 19-2k and 2 ~1550.
It should be noted that I hard randomed about 1/2 the time and we more or less ban random champions, just so long as the other team doesn't get to first pick rammus. Queue times were very short tonight, and I can't recall spending more than 1-2 minutes in queue for any single match that didn't involve 5+ Q dodges.
As you can see, having a lower Elo gave me larger increases for victories and smaller decreases for losses than my Q buddy; however, there's not really enough data here for much conjecture beyond that. Unfortunately, matchmaking is still throwing together players with wildly different Elos, and in minimal time compared to my usual 5-15 minute queue times.