Eclipse I think is different?
Eclipse uses c=0.01, o=0.99, f = -0.0101...
just want to voice my opinion here, I mined my very first Bitcoin on your pool (and sold it for 0.85 cents! ) and mined probably in total 200 coins with you.
I personally think DGM sucks and will become less popular as the difficulty continues to go up.
Yes it's pool hop proof and "fair" but only really fair to a person if they want to dedicate themselves to mining with that pool and that pool only 24/7
I like your modified proportional score system - a lot of people do and I think you should stay with it , and if you really have to change, change it to a pure PPS like BTCGuild is doing.
thank you for listening
Did you take the time to understand how DGM works and then judged it based on its merits, or did you just arbitrarily decide that it penalizes intermittent miners? It does no such thing. Mining with multiple pools will not reduce your rewards (and could be a good way to reduce your variance), and if you're disconnected you will only lose the worth of the work you could have done during that time. For every share you submit you get on average (Block Reward / Difficulty), that's all there is to it.
For a long time slush's method has been falsely accused of penalizing intermittent miners, so in a way it's kind of refreshing to hear you acquit it of this crime. But it looks like you've just found a new scapegoat in DGM.
slush's method is high variance, less hoppable than proportional, but still significantly hoppable.
Geometric method is similar to slush's method, also high variance, but completely hopping-proof.
DGM is also hopping-proof, but much lower variance.
In fact with DGM you can have the variance as low as you want, much lower than in proportional for example - it's all in the parameters. The easy way to reduce the variance is by increasing maturity time, which I think slush can pull off because it's a big pool, so even the increased maturity time will be pretty short.
PPLNS, when properly implemented, is also hopping-proof and almost as low-variance as DGM. It can be used if DGM is too complicated. In particular, I think shift-PPLNS could work better with slush's parallel architecture than DGM.
PPS is definitely a possibility, and I think PPS proxy pools are the way of the future. But until the substrate for it is developed, slush would have to take great risks, which he may not want to do without significantly raising the fees, reducing the long-term gains of miners.