The background you've posted is wrong. Eligius has always paid out immediately (or as close to it as possible); while there were some delays in SMPPS due to the variance of block finding times, the maximum rewards ever took was a mere 3 days since I made sure to send before any balance was waiting that long.
Well, it was nice of you to dip into your own pocket to pay miners, that's not a part of the SMPPS protocol.
Obviously these payouts came from the pool wallet, which gets filled by the SMPPS buffering and when miners haven't earned enough to achieve a reasonable payout yet.
But the pool wallet had insufficient funds to pay extra credit - isn't that why there was a back log?
Your blog misrepresents extra credit as "owed", which it is not and never has been.
This is something you've mentioned many times, and I don't understand it. If extra credit is not owed, why pay it? If you didn't pay it, would miners still earn 100% of PPS?
It's paid as part of the reward system rules, as an incentive for miners to keep mining even on long rounds when the buffer is empty. No reward system can pay 100% PPS, with or without extra credit. SMPPS and CPPSRB just do the best they can to try to achieve that.
Do you mean irl or in theory? Theoretically, any provable fair reward method without a fee can pay 100% PPS. If you mean irl, then ok, but the extra credit is an integral part of SMPPS and if the extra credit isn't guaranteed to be paid even in theory, then it even theoretically it may not pay 100% PPS. I understand this never happened for SMPPS, but I would like to understand your thoughts on the matter better.
This is why I thought discriminating between a "buffer" and "extra credit" was wrong. The point of extra credit is to enable fair rewards, and to achieve PPS.
Most other reward systems (PPLNS, proportional, DGM, etc) never attempt to track or pay extra credit at all - under those systems, miners never get this. It is therefore unfair and possibly dishonest to count extra credit against "time to maturity".
I don't understand this. There's no need for other reward methods to track extra credit, since they're provable fair reward methods that have variance due to pool luck (which miners on Eligius do not experience). So what "extra credit is there to track?
A DGM pool with a week of bad luck, will never try to make it up to miners who mined only for that week.
I see where you're coming from, but I think variance is well understood by most miners now. Whether you're mining DGM or PPLNS, there will be variance in rewards - larger than PPS payments some weeks, smaller than average others. Just as there's no reason to track and rectify the losses a miner incurs in an unlucky week, there's also no reason to track and remove the bonus from a lucky week at a DGM or PPLNS pool. Over time the reward will be fair. SMPPS and (I think CPPSRB) won't ever pay more than PPS, hence the necessity of tracking extra credit.
CPPSRB also tracks a value similar to "extra credit", but the same still applies. With CPPSRB, extra credit is actually less likely to ever be paid due to how LIFO works. On the upside, CPPSRB gives actual rewards much more like proportional, so they fit with the actual block payout availability much better without any manual sends; in normal operation, no payout should be delayed more than a single block, and most (at least on Eligius) will be paid out in the same block they are earned in.
I read the info on the Eligius site for the first time, and it's quite an interesting reward method. By assuming that 100% of PPS value won't be paid (am I right in that?) you're able to make certain that reward are paid in a timely manner. So time to maturity is reduced to 100 network block confirmations, with a trade off of increased earning variance. I think most miners would be happy earning a little less if they didn't have to be worried about when it may get paid.
Have you derived (or otherwise calculated) the expected PPS miners will earn under the new method?
With Eligius's no-fee implementation, shares are still paid 100% PPS value, but the unavoidable loss is in the backlog of shares that inevitably never get paid. I think it's easiest to explain if you compare it to PPLNS; the main differences there are just that shares are never paid twice (which results in older shares getting paid on lucky rounds). While PPLNS effectively forces a pool to compromise between variance (low N, like <=expected) and reward delays (high N, like expected*
, CPPSRB's rewarding older shares rather than doubling rewards gives it a low variance with instant rewards.
Nice analogy - that makes it clearer, thanks. However it seems that there is a possibility that in the long term not all shares would be paid.
I don't have a problem with this - as long as a reward method is fair and is unlikely to cause problems for a pool then it's a metter of miner choice. For example, I mined at Eligius for a time (in the proportional days) mostly because I liked not having to register. I still think that's an important option. Miners may choose to use Eligius for reasons other than simply the expected reward.
However, informed miners are happier miners. Including, for example, how many shares on average won't be paid in a given time period - a month in your explanation may help. Or the probability that any given share won't be paid in a given time period. It may also be possible that some shares will never be paid - and if so miners should know the probability of that, too - if it's possible to derive.
It may be that very few shares won't be paid long term and have no significant impact on miners, but I think it would be good for miners to know what that might be - even if it's just the result of a simulation rather than a derivation. Have you managed anything like that?