This is also affected by increases in our Hashrate. If our hashrate goes up your return will go down. But, it should be offset by increasing the number of blocks we find. You really can't tell much looking at each individual block, your graph 7 day average is a better indicater.
OK, so you're saying it was a coincidence that this hashrate increase, at the apex of my going from >expected to
How does the hashrate relate to the shares per block being different? I can't see how that works.
If the pool hash rate stays the same while the network hash rate increases, the pool's chance of finding a block decreases, i.e., it will take longer on average for the pool to find a block. Similarly, if the pool hash rate is increasing while the network hash rate stays the same, the pool's chance of finding a block increases, resulting in shorter round times.
Now, if your hash rate stays the same while the pool's hash rate increases, your share of the pot decreases because your payout is based in part on your_shares_submitted_during_round divided by pool_shares_submitted_during_round. Your lower payout will be offset a bit if the pool has an increasing hash rate because the pool will find on average, more blocks.
I understand that your_shares_submitted_during_round / pool_shares_submitted_during_round is what is affected by respective varying hashrates. As such, an increase in the pool's total hashrate explains why I would get a smaller proportion of the total shares for a block. I know that here is added variance to do with how close your last share in the block is to the end of the block, which would explain why the shares proportion doesn't translate perfectly to my earnt BTC as with a proportional earning pool. What I didn't get was why my earnings might have been more than expected more often than less than expected, until I stopped mining for 8 hours, then started again, with more 'balanced' expected earnings.
To be sure, I plotted the numbers for before and after the outage, though not really enough points to confirm anything. I'm guessing that I was just having a lucky run before the outage. If that's right, then there's no long-term penalty for disconnecting other than the current round. Is that right?