Yeah except that graph is yet again another graph that doesn't understand or consider statistics.
Like for example, the pool that graph comes from has only had only 90% of expected luck over the last month.
So anyone mining there has been paid only something like 90% of PPS, which ... oddly enough ... doesn't match that graph either
Anyone using a historical hash rate to decide on a pools performance, needs to include:
the block luck if the time period is not long
explain why luck isn't 100% if the time period is long
Long means many months at 1% of the network and probably a couple of months at 10% of the network.
Let me expand on that ...
Ignoring luck, a PPLNS pool that charges a 0.9% fee will expect to average 99.1% of PPS ...
If it isn't, after removing luck from the calculation, then there's something wrong with the pool software, or the miner has been hopping off the pool during the PPLNS payout times.
A PPS pool with a 4% fee will expect to averate 96% of PPS always
There have been a few people running pool comparisons lately and I've yet to see one of them that understands statistics ...
They have all been based on the Gamber's Fallacy: historical luck will match future luck.
But luck is only a measure of the past, and applying it to the future means you don't understand what luck is.