Quote from: CoinBuzz on March 25, 2014, 05:25:05 PM

@coinshift
For example, now that i am writing this post, pool is dedicated 96.5% on AUR and 0.73% on DOGE and remains on others.
Since pool has 2.5GHash, 0.73% will be about 18Mhash. It is obvious that 18Mhash cant find any DOGE block. So i think the pool algorithm should consider coin's difficulty into it's dedication's amount too (i like to call it hashrate efficiency).
It is wiser that pool assign those 18Mhash to RDD or GDC since it can be more efficient on those coins difficulty [OR] assign enough hashrate to DOGE side.

What do you think?

That's a common question. Since finding a block is an independent random chance, you shouldn't slice up the hashrate and try to look at pieces independently across small time periods (in this case, the instantaneous DOGE hashrate). Statistically there is no benefit to maintaining some "minimum" instantaneous hashrate on a single coin.

We do manage the pool's "hashrate efficiency", but it matters only in aggregate across all the coins we're mining and across a longer time period. In other words, you should look at the weighted average of the coin difficulties, not any one coin. For example, imagine we had 2000 altcoins, all with the same difficulty as DOGE. We then allocate 0.05% to each. If you look at an individual coin you would think the probability of finding a block would be quite low. However, in aggregate, we'll actually find blocks quite often -- this is because the risk/reward ratio for each unit of hashrate is profitable.

Another way to think about it is thinking of it over a longer timescale. Instantaneously at some points it was .73% DOGE and others it was 80% DOGE, but over the course of the day it averaged out to 700MH/sec.