This is very interesting approach and very elegant notebook, I would like to read more, but it seems like paragraph (3) cuts off in the middle of a sentence:
I got two poorly formulated question:
1: How does CO-CM (4) react to groups with differing numbers of donations? The scenario I am thinking of is two pairs of voters:
- A and B that cast a single vote for the same grant.
- C and D that make 99 similar votes. Then D makes one extra vote.
I think naive cluster profile (3) as described would penalize A/B while treating C/D as distinct voters?
The behaviour I would find desirable is to penalize pair C/D more as such overlap is less likely to be result of random chance.
2: Regarding strategy. I might be wrong here, but with current QF in grants sometimes winning move is not to play (project reaches matching cap before I even voted => my vote would carry 0 additional subsidy => bankroll money wait for next round).
Is there ever a scenario under mechanism (2) and (4) where a winning move (i.e. one that increases net profit for project) could be to dimnish my donation to my true preference just so that collusion resistance penalizes me less? Hm… a lot of people are donating only for Rotki so I will donate 99$ to true preference and 1$ to climate stuff just so that I am not lumped with those folks.
.