[...] or weight based shares" (slush's pool) which favors faster cards and devalue's shares over time after they've been solved.
If you are going to make such an accusation, can you please show any mathematical/statistical proof of this?
Yes, the variance is slightly larger and the fewer shares/proofs of work you have submitted the more likely that variance significantly affects you, but it doesn't take long at all for you to converge towards the expected average level of generation/payout. It only takes a CPU Miner around 1 week to generate 1000 proofs of work and as is plainly seen in the graph below, the difference in variance between Share based and Score based at 1000 shares/proofs submitted over the entire lifetime of the miner is essentially non-existent. And again, this only takes ~1 week for a CPU miner.
