I suggested a method that is hop-proof, by always offering the same expected return for the next share contributed, and riskless for the pool-operator. It's what (a couple messages after initial proposal) I labeled 'Pay-Per-Last-N-Shares' (PPLNS). It's not the same as proportional or straight PPS.

You suggested PPLNS was unfair but didn't explain why/how.

It's simple to implement, clear, hop-proof, and operator-friendly in the amount of risk (none) and running state required to calculate payout (very little).

If you make N small, some shares contributed wind up earning nothing.. but that's totally unpredictable and affects all shares equally, so there's no tilt towards any miner timing or scale there. It's perfectly fair in those dimensions. By making N arbitrarily large (but still constant and pre-advertised), you could make the likelihood any share pays zero arbitrarily small (if you think fairness requires every share get some token payment no matter how far removed from success).

I'm pretty sure PPLNS provides a disincentive to the 'withholding sabotage' as well.. anyone who's been mining for a while makes marginally more by submitting a found share, compared to waiting for someone else to.

And it doesn't require the operator holding a bunch of 'earnings' in reserve, in the meantime paying some amount that seems to be roughly minimum(pps(miner),proportional(miner)).

I could be wrong, but if so you haven't explained the problems.

And of course it's your pool, your rules. I'm just providing an honest assessment of how MaxPPS looks from the outside.