The math does not convince me on either side of the pool-hopping arguments. BOTH sides ignore the stochasticity inherent in the block-finding behaviors. (!!!)
In my post a few pages back, I put up my math showing how pool hoppers don't impact the BTC/minute for a round in any of 3 scenarios: Pool hoppers there the whole round, part of a round, or not at all. My math was wrong because it was looking at a single round worth of information. If you look back at the post, and COMBINE the two rounds you will get the following:
In scenario 1 (with hoppers) "rest of the pool" will get
45 + 48.34 = 93.34 BTC
in 5 + 32.22 = 37.22 min,
resulting in 2.51 BTC/min
In scenario 2 (without hoppers) "rest of the pool" will get
50 + 50 = 100 BTC
in 5.55 + 33.33 = 38.88 min,
resulting in 2.57 BTC/min
These were using fairly simplified numbers, and assuming hoppers were only 100 GH/s rather than 200+ which I now know they were. In those scenarios (a 500k and a 3m round), the net effect was 0.06 BTC/minute drop if hoppers were part of the pool during the start of rounds, or 2.3%. Now my math used was very simple, but the theory holds true for actual pool hopping.