I don't have sim figures for slice vs standard yet but if you work it out scheduling, number of pools and pool size choice affect efficiency and variance as follows:
- Standard scheduling has better efficiency but higher variance than slice
- Larger pools have lower variance than smaller pools
- Larger number of hoppable pools available lead to higher efficiency.
So a couple of choices are:
- Have lots of pools including the smaller ones and use slicing to reduce variance
- Only mine larger pools for better variance and use OldDefaultScheduler for best efficiency
Anyone else spot other tactics?
"Standard scheduling has better efficiency but
higher variance than slice" - I think that's a misguided assumption there.
You are already greatly reducing your variance by spreading your love over many different pools. Even a 100Ghash pool is going to be solving a block in less than 24 hours on average; five that size, and you are averaging 5 smaller-payout blocks a day vs. one pool's variance. Add in hopping a "top five" pool and your daily payments are clockwork. The old scheduler
wants to submit the same percentage of shares to every pool, and the same number of shares to every pool round.
Decreasing variance at the cost of mining efficiency shouldn't enter into any consideration you make. Instantaneous share value quickly drops from 13x to 1x within difficulty N/.43 shares, and a +28% per pool return estimate only happens when you start mining right at new round start. A pool has a five minute round, and now you've only put in half as many shares as you could have by sub-optimal slicing.
If there was a time-slicing option to be made at the expense of profitability (which, only as a side effect, would reduce variance), it would be to spread random shares around to all the pools you are checking for stats and LP pushes when no pool is