For those interested, I've written up a more correct explanation (I hope) on Maximum PPS.

I'm still open to other ideas, if they can be demonstrated to be better. Please write up a wiki page and nominate them on IRC.

"HanSolo", your presence on IRC to discuss PPLNS has been requested by multiple people, if you have time.

The error is in assuming either 'value' or 'earnings' are owed to the miner. Until the miner does the actual work (collected in 'value') and earned the pool enough funds (collected in 'earnings'), nothing is due under MaxPPS. For an ordinary miner, in an ordinary scenario, doing the work should also earn the funds required. There are two scenarios where this is not the case:

• someone withholding valid blocks -- in this case, under MaxPPS, he harms the overall pool income, and inevitably everyone is hit with a loss in 'earnings' -- with straight PPS, the pool operator takes 100% of the risk for this, even if he charges a large fee (both DeepBit and BitPenny charged 10%, and the latter had to close because it was losing too much money even still)
• pool hoppers -- in this case, under MaxPPS, pool hoppers might continue to work only on short blocks, and always participate in earning the pool the appropriate funds to pay for the work they do, but they are never paid more than the work they do, so there is no incentive to hop -- under straight Proportional, these miners would only participate when there was a chance at getting paid more than the work they put into the pool, and inevitably the pool would be left empty if everyone did it

You can't fairly disregard fees here. Straight PPS is impractical even with a 10% fee. MaxPPS should pay out more than that remaining 90% in theory.

The mistake here is in assuming the pool actually has funds to pay everyone's "value"; if that is the case, it also means everyone has the "earnings" too. The only time these are different are in excessive periods of bad luck (possibly induced artificially by a cheater) when everyone shares the "hit", or excessive periods of luck where everyone is paid in full for the work they do (and all build a reserve to cover the risk of future bad luck).

If you know of a better way, feel free to suggest it. Every other way I've seen is unfair to some normal miners.

No, and I'm tired of refuting these ridiculous misexplanations that seem to assume some few elite should pool hop while everyone else shouldn't.

Straight PPS is vulnerable to withholding attack due to infinite risk for the pool operator.

Already thought of this. Someone pointed out it's even less fair than the score-based approaches.

MaxPPS is like this, but makes the "reserve" per-user so they can't cheat the pool by withholding blocks.

It's not supposed to punish pool hoppers (it's their right to hop-- so punishing them would be wrong). They simply don't earn any more than someone mining normally, so there's no incentive to do it.

The "big upgrade" I'm working on will be able to combine the pools to work together. The minimum payout will also likely be reduced.

This is planned for the permanent version of MaxPPS.

It was running out of space fast, and I had the opportunity to move it somewhere relatively quickly (~12 hours). All real-time discussion, including what's going on with the pools, takes place in our IRC channel.

The Europe server is far more powerful and has far more space than the US server. I'm not worried about it.

Does this support free Linux (ie, no fglrx)?

*** The channel topic is "http://eligius.st/ | Status: USA down for maintenance. Somehow it still found a block, though... | If a pool goes down for more than a minute or two, call Luke-Jr (whois eligi.us for number) | Plz donate: 1RNUbHZwo2PmrEQiuX5ascLEXmtcFpooL | TradeHill referral code: TH-R12147".

Shortly after the USA server is back online.

If you don't have a real reason to be using the US pool, please switch to the Europe pool. This can be done by using the mining.eligius.st hostname.

If you have a reason to stick to the USA pool, you will need to make sure you are using us.mining.eligius.st explicitly.

Edit: Nevermind this post...

I would appreciate it if anyone who can switch to Europe without significant loss did so... at the current rate USA will run out of disk space within 48 hours, and it's highly unlikely I'll get the new version done by then.

Edit: To force Europe, use eu.mining.eligius.st for hostname

People weren't happy with proportional either. "One-sided PPS" is only a practical reality right now-- the actual policy is straight MaxPPS. The 40% implemented is the bare minimum needed to make the adjustment retroactive.

As the donations have been much less than even covering the costs of running the pool, let alone my expenses of time spent writing code, I cannot guarantee any ETA. I'm doing this basically free of charge, so for the most part it has to live in my "free time". If you want to change that, feel free to help the donation address grow to a reasonable size to cover my expenses, and then I can justify spending more "time I'm doing paid work" on it. My guess as to completion, is maybe 30 hours of work. If this is done in my spare time, it might be 1-2 weeks. In reality, USA will probably run out of disk space within a few days, so I need to hurry regardless.

With MaxPPS, the overall payout should be the same as proportional to legitimate miners.

You're welcome to use the Europe pool (which remains proportional) in the meantime.

To be fair, one of the ideals I've tried to make Eligius meet from the start has been not needing to trust me (immediate payout in the block, third party auditing, etc), but there's a compromise that needs to be made for certain functionality (minimum payouts, anti-pool hopping, etc). I'm not too happy with the current one-sided MaxPPS myself, and I can understand why someone wouldn't want to use it until it's finished if they don't especially trust me. Hence why Europe remains as-is.

Europe will probably remain as-is until the big upgrade. At which time, I plan to have all the servers unified working together on a single logical pool.

That's beneficial with proportional, but not so much with MaxPPS :p

Well, 1) there were a lot of people complaining about the pool hoppers, 2) I figured about a month was plenty of time for people to object to MaxPPS and/or suggest other alternatives, and 3) I realized that while I cannot retroactively change the short blocks, I can retroactively make up underpayments on long blocks.
In hindsight, I also observe another reason: 4) If it turns out to really be a problem, I can still undo the 40% implementation.

Underpayments on the long blocks will be made up retroactively regardless of whether a miner has continued on the pool or not.

That's the variation of mining; one purpose of pools is to eliminate variation. Maybe it's possible to make variation tolerance configurable. I have enough to implement without trying to consider the implications of changes right now. If you want a change made, document it on the wiki and get support for it.

You're describing pool hopping in a roundabout way. Preventing it from earning more is the whole point.

The mistake is in assuming your reward is equal to the "value" or "earnings" numbers. "Value" is basically identical to flat PPS (without fees), and "earnings" is identical to "proportional". But under MaxPPS, your reward is considered to be the lower of the two. Contrasted with proportional, the "value" removes the advantage of pool hopping (which everyone seems to agree is a good thing to prevent). Contrasted with PPS, the "earnings" removes the vulnerability of the miner who always withholds the good shares. An honest miner, in theory, will maintain "value" and "earnings" that are approximately equal.

Since the terminology used seems to be confusing people, perhaps I should rename them to "x" and "y" or something?

This does explain your confusion a bit. At no time are either "value" or "earnings" reset. They both continue to accumulate. If you miss a round, you just pick up where you left off whenever you want.

Perhaps my mistake was using English words for technical details...

Actually, maybe.... I did ban at least China a while back from my personal webserver (which happens to run USA for now).

The statistical data isn't updated yet, so Artefact2 will just show you what you'd actually get if we found a block at this instant. When the upgrade is complete, it will look nicer.

Again, the concept of the new payout system: Maximum PPS

Unfortunately, we found 2 very short blocks just now, so it's looking "ugly" to some folks who don't get the concept. Basically, while you were just paid 1/4 and 1/8 less than you would have gotten under the old system, when those "ugh, no block all day" blocks hit, you'll get reimbursed for these short ones. Doing it this way eliminates the benefits of pool hopping.

Also note that only USA is on the new system for now. If you want to use the old system for any reason, you can still mine on the Europe pool.

Due to concerns of pool hopping, I have implemented part of MaxPPS early. Miners will no longer be paid more than the PPS-value of their shares in short blocks. This means that when the block is found early, Eligius acts as a PPS pool (except without the huge fees usually associated with PPS). This also means there is a large buffer of extra mining coins in each block. To conclude the poll, I have decided to compromise the minimum payout: it will be gradually reduced to 0.33 554432 BTC. Please note that long blocks are still on the old proportional payment schedule until the "big upgrade". I am aware that over the long run, this current payment schedule can result in insane benefit to the pool at the expense of the miners, but I hope to wrap up the "big upgrade" before it has any practical effects in this way, and at the time of the big upgrade, I plan to have the MaxPPS retroactively be applied to the present so anyone mining on the long blocks will have their work on short blocks reimbursed properly.

This rambling is probably very confusing, so please ask questions if you need clarification on anything.

I can't say it isn't the miner. It has no real reason to run out of work and idle, even if the pool takes a few seconds to respond.
Usually Phoenix problems are high invalid rates, though.