Does anyone have a feel for how double geometric scoring affects round size based variance?
Out of interest, have you compared your average payout/share to a theoretical prop payout/share for the same round?
If anyone's not sure what I mean then just pick a previous long round and divide payout by your submitted shares. Then divide the Difficulty for that round by the total shares for that round (theoretical proportional pool payout). Compare and contrast. If you can be bothered I'd like to see the results.
You might need to compare a few sequential rounds rather than just one round. To do this, divide your actual total payout by the total shares you submitted. Then (assuming Difficulty didn't change between the rounds you are looking at) compare with (Difficulty*number of rounds)/(total shares in rounds).