Which is the best pool for mining? - A guide for choosing the right pool

[DeathAndTaxes]

There is no disagreement mate. We all know that PPLNS / Prop suck and that PPS is really the only viable pool system. Only solo is better when you have the power. People using Prop / PPLNS deserve to be cheater by pool hoppers, have a ton of variance and cheated by pool operator. Sheep do your research !

Indeed if PPLNS can be abused by pool hoppers I certainly need to do some research and switch to a better reward system. After all, stopping pool hoppers was the reason I went with PPLNS.

Could you please point me to some evidence or explanation of these facts so I can do my research properly? Or perhaps you can explain how it works yourself.

Within a difficulty "set" PPLNS can't be hopped.  When difficulty changes if the change is very large (30%+) there is a small window where the pool could be hopped in theory.  However there are two real world issues that make this useless:
a) difficulty changes only occur once every 14 days
b) the window isn't always present and last only till block is solved at most

When I pool hopped I had some PPLNS pools setup in my config.  The software never found a viable hop point.  Even on difficulty changes the PPLNS pools were either <100% efficiency or when they did present a "bonus" other pools were worth even more.

To "solve" even this small (and mostly academic flaw) a PPLNS pool could record the difficulty of each share submitted.  The split would then be weighted by difficulty.

1) look at all shares in last N
2) multiply each share by difficulty when submitted
3) take product of number of shares * difficulty
4) total is miner's weighted share count
5) do the same for whole pool
6) miner share = (miner's weighted share count) / (pool's weighted share count)

You could normalize the numbers by dividing difficulty by difficulty of first share.  This has no effect on the math it simply makes the numbers smaller and more easily understood.

For example:
In last 2 million shares
you submitted 100,000 shares at difficulty of 1,378,309 worth 1.000 shares each.
you submitted  80,000 shares at difficulty of 1,202,890 worth 0.873 shares each.
Total weighted shares submitted: 169,818.31 shares

[kislam]

you don't have to, we have ppl around that can simulate any situation over a long period of time...

Simulations are one thing, real life data is quite another. I am disappointed by the (lack of) response from the PPLNS proponents/supporters, though. This test could potential prove their very important point that PPLNS reward method does not negatively impact the intermittent miner over the long run, compared to other methods.

No takers?

[DeathAndTaxes]

you don't have to, we have ppl around that can simulate any situation over a long period of time...

Simulations are one thing, real life data is quite another. I am disappointed by the (lack of) response from the PPLNS proponents/supporters, though. This test could potential prove their very important point that PPLNS reward method does not negatively impact the intermittent miner over the long run, compared to other methods.

No takers?

I am not sure what you are asking.  A good PPLNS pool to use for a test?  Try BitMinter.Com

How long do you need to test for to compensate for variance (luck) between pools?
You will want to submit a lot of shares to each pool.  Honestly I would look at using at least 1 million shares per pool.  Longer is always better.  You could always report back at 2M, 5M, 10M shares.

The length of time of each "switch", the randomness, the "time away" doesn't matter.  Just make sure you record the number of shares submitted to each pool, revenue generated and you can get the revenue per share.

[DrHaribo]

I am getting a new 5830 in the next few days. I propose to set up the card to mine round-robin at PPLNS, PPS and Prop pools over an extended period, just to see how the payments (after adjusting for fees if the pool charges any) stack up against each other.

Interesting idea. When choosing a proportional pool it may be best to choose one that isn't being pool-hopped, otherwise all it shows is how badly pool-hopping affects other miners. Your ideas for this test look fine, the only difficult thing is the time frame. Maybe report at intervals like DAT suggested.

* Basis of comparing the earnings - OTTOMH i think it might be ok to compare average reward per share from each pool, but pls suggest better ideas

Yes, comparing payout per accepted proof of work (share) is the way to do it, I think. Disregard stales. Other reward models will lag behind PPS, so make sure the current block at the proportional pool has completed, and all your accepted proofs of work are no longer eligible for pay at the PPLNS pool. After that, either wait for blocks to confirm, or count in the unconfirmed balance.

this should be of interest for you:
https://bitcointalk.org/index.php?topic=39832.0

That's what I read before I implemented PPLNS in my pool. Meni Rosenfeld does some excellent work. Just like he suggests, I use a score for each proof of work equal to 1 divided by the difficulty at the time it is submitted. This is what I meant with "properly implemented" PPLNS earlier in this thread, just forgot to write that this time.

To "solve" even this small (and mostly academic flaw) a PPLNS pool could record the difficulty of each share submitted.  The split would then be weighted by difficulty.

Yes, that is what I do. Each accepted proof of work gets a score=1/difficulty. I should have said weighted PPLNS is hopper-proof.

[DrHaribo]

The length of time of each "switch", the randomness, the "time away" doesn't matter.  Just make sure you record the number of shares submitted to each pool, revenue generated and you can get the revenue per share.

Yeah, but make sure each reward system is tested under the same difficulty, otherwise the shares are not directly comparable.

Round-robin with cgminer sounds like a good solution.

[Inaba]

Can you add in DGM to your test?  I would like to know how my pool compares in a real world scenario.

[bulanula]

Can you add in DGM to your test?  I would like to know how my pool compares in a real world scenario.

LOL. Never heard of double geometry stuff till now. How does it work ?

[urstroyer]

Can you add in DGM to your test?  I would like to know how my pool compares in a real world scenario.

LOL. Never heard of double geometry stuff till now. How does it work ?

Double Geometric Method was born here:
https://bitcointalk.org/index.php?topic=39497.0

And here is some FAQ content:
https://yourbtc.net/faq-page

Inaba uses different parameter on eclipse than me on yourbtc. Maybe both could be considered in a test. Definitely interested too.

[teukon]

you don't have to, we have ppl around that can simulate any situation over a long period of time...

Simulations are one thing, real life data is quite another. I am disappointed by the (lack of) response from the PPLNS proponents/supporters, though. This test could potential prove their very important point that PPLNS reward method does not negatively impact the intermittent miner over the long run, compared to other methods.

No takers?

The simulations are really only suitable for illustrating a particular reward system, it is mathematics that establishes whether or not a reward system impacts intermittent miners.

A real data test will be difficult to perform for a number of reasons.  I fear all that could be gleaned from such a test is which pool "currently" performs the best and say nothing conclusive about future performance or about reward systems.

[bulanula]

you don't have to, we have ppl around that can simulate any situation over a long period of time...

Simulations are one thing, real life data is quite another. I am disappointed by the (lack of) response from the PPLNS proponents/supporters, though. This test could potential prove their very important point that PPLNS reward method does not negatively impact the intermittent miner over the long run, compared to other methods.

No takers?

The simulations are really only suitable for illustrating a particular reward system, it is mathematics that establishes whether or not a reward system impacts intermittent miners.

A real data test will be difficult to perform for a number of reasons.  I fear all that could be gleaned from such a test is which pool "currently" performs the best and say nothing conclusive about future performance or about reward systems.

I really don't know why anyone would bother with anything but a PPS pool ...

Want to be cheated by operator, long rounds, pool hoppers, variance then please go ahead and worship deepbit / slush and other such botnetmasters of pools.

[kano]

you don't have to, we have ppl around that can simulate any situation over a long period of time...

Simulations are one thing, real life data is quite another. I am disappointed by the (lack of) response from the PPLNS proponents/supporters, though. This test could potential prove their very important point that PPLNS reward method does not negatively impact the intermittent miner over the long run, compared to other methods.

No takers?

The simulations are really only suitable for illustrating a particular reward system, it is mathematics that establishes whether or not a reward system impacts intermittent miners.

A real data test will be difficult to perform for a number of reasons.  I fear all that could be gleaned from such a test is which pool "currently" performs the best and say nothing conclusive about future performance or about reward systems.

I really don't know why anyone would bother with anything but a PPS pool ...

Want to be cheated by operator, long rounds, pool hoppers, variance then please go ahead and worship deepbit / slush and other such botnetmasters of pools.

Firstly, you have only stated 3 reasons there not 4.
long rounds=variance
And in that case, as long as the pools luck - %fee is less than the PPS pools %fee you will be ahead.

So you are left with trusting the pool operator and hoping that hoppers are not taking more than a small % of the pools total (again less than the % difference charge between the pools) ...

[DeathAndTaxes]

you don't have to, we have ppl around that can simulate any situation over a long period of time...

Simulations are one thing, real life data is quite another. I am disappointed by the (lack of) response from the PPLNS proponents/supporters, though. This test could potential prove their very important point that PPLNS reward method does not negatively impact the intermittent miner over the long run, compared to other methods.

No takers?

The simulations are really only suitable for illustrating a particular reward system, it is mathematics that establishes whether or not a reward system impacts intermittent miners.

A real data test will be difficult to perform for a number of reasons.  I fear all that could be gleaned from such a test is which pool "currently" performs the best and say nothing conclusive about future performance or about reward systems.

I really don't know why anyone would bother with anything but a PPS pool ...

Want to be cheated by operator, long rounds, pool hoppers, variance then please go ahead and worship deepbit / slush and other such botnetmasters of pools.

Firstly, you have only stated 3 reasons there not 4.
long rounds=variance
And in that case, as long as the pools luck - %fee is less than the PPS pools %fee you will be ahead.

So you are left with trusting the pool operator and hoping that hoppers are not taking more than a small % of the pools total (again less than the % difference charge between the pools) ...

Actually it is more like 1: Variance.

Pool hopping is a non issue as PPLNS, SMPPS, and double geo can't be hopped.

Trusting Pool operator never goes away.  A PPS pool operator can cheat by either running away with the money or "stale share shaving" (reporting some % of valid shares as stale to reduce payout).  If you don't trust your pool operator well you are fucked either way.

So that leaves variance.  Given the 5%-10% fee on most PPS pools I will gladly eat some variance to get 5% or more extra by not paying fees.  The hourly variance may be significant but I don't care how much I make per hour.  I care how much a make per month (when my powerbill comes).  Variance is a non-issue on a month long timeframe. 

If someone thinks less daily variance is worth it then they should pay the extra fee but thinking they "got" anything other than reduced payout for reduced variance is just denial.

[Inaba]

Much more concise and elegantly put than I could have... but that's exactly my thoughts, Death

[bulanula]

you don't have to, we have ppl around that can simulate any situation over a long period of time...

Simulations are one thing, real life data is quite another. I am disappointed by the (lack of) response from the PPLNS proponents/supporters, though. This test could potential prove their very important point that PPLNS reward method does not negatively impact the intermittent miner over the long run, compared to other methods.

No takers?

The simulations are really only suitable for illustrating a particular reward system, it is mathematics that establishes whether or not a reward system impacts intermittent miners.

A real data test will be difficult to perform for a number of reasons.  I fear all that could be gleaned from such a test is which pool "currently" performs the best and say nothing conclusive about future performance or about reward systems.

I really don't know why anyone would bother with anything but a PPS pool ...

Want to be cheated by operator, long rounds, pool hoppers, variance then please go ahead and worship deepbit / slush and other such botnetmasters of pools.

Firstly, you have only stated 3 reasons there not 4.
long rounds=variance
And in that case, as long as the pools luck - %fee is less than the PPS pools %fee you will be ahead.

So you are left with trusting the pool operator and hoping that hoppers are not taking more than a small % of the pools total (again less than the % difference charge between the pools) ...

Actually it is more like 1: Variance.

Pool hopping is a non issue as PPLNS, SMPPS, and double geo can't be hopped.

Trusting Pool operator never goes away.  A PPS pool operator can cheat by either running away with the money or "stale share shaving" (reporting some % of valid shares as stale to reduce payout).  If you don't trust your pool operator well you are fucked either way.

So that leaves variance.  Given the 5%-10% fee on most PPS pools I will gladly eat some variance to get 5% or more extra by not paying fees.  The hourly variance may be significant but I don't care how much I make per hour.  I care how much a make per month (when my powerbill comes).  Variance is a non-issue on a month long timeframe. 

If someone thinks less daily variance is worth it then they should pay the extra fee but thinking they "got" anything other than reduced payout for reduced variance is just denial.

If variance is not important then why bother to mine in a pool at all

[DeathAndTaxes]

If I had 50GH/s I wouldn't.
Likewise if my electrical bill came once a year I wouldn't.

However I don't have 50GH/s and my electrical bill comes once a month.  As I state I could care less about hourly or daily or even weekly variance.  However I would like to get revenue each month to pay the electric bill.

There is no significant variance over a 30+ day period when using a pool.  A pool also has the advantage of merged mining, and ability to consolidate hashing power towards a single wallet.  If pools didn't exist I likely would use an internal pool for my rigs anyways.  Given the number of free, donation, or low cost pools I simply don't see the benefit of doing all that work myself.

[bulanula]

If I had 50GH/s I wouldn't.
Likewise if my electrical bill came once a year I wouldn't.

However I don't have 50GH/s and my electrical bill comes once a month.  As I state I could care less about hourly or daily or even weekly variance.  However I would like to get revenue each month to pay the electric bill.

There is no significant variance over a 30+ day period when using a pool.  A pool also has the advantage of merged mining, and ability to consolidate hashing power towards a single wallet.  If pools didn't exist I likely would use an internal pool for my rigs anyways.  Given the number of free, donation, or low cost pools I simply don't see the benefit of doing all that work myself.

I think one can also solo MM but I don't know the exact steps to get that working 100%.

[organofcorti]

There is no significant variance over a 30+ day period when using a pool. 

There is a significant amount of variance based on the pool hashrate though. What pool hashrate is giving you insignificant variance after only 30 days? Actually it probably makes more sense for me to ask what number of round it takes for you to notice variance being insignificant, I'd be interested to know if it tallies with my own observations.

As far as SMPPS hopping goes, I've had average maturity time decreased by half a round on average. Not helpful on a big pool, but very handy on a small one.

[DrHaribo]

Simulations are one thing, real life data is quite another.

I added a new statistics page at my pool showing real life data: http://bitminter.com/stats/rewards

You can see how actual payouts compare to the expected average income per share (basically 0% fee PPS). I can probably improve this with some stats for "all time average" and "average last X time", but I think it is already useful. There are also charts of the pool's luck here: http://bitminter.com/stats/luck

Quick way to get some data without taking months to run a test.

[cablepair]

very nice

[kislam]

Simulations are one thing, real life data is quite another.

I added a new statistics page at my pool showing real life data: http://bitminter.com/stats/rewards

You can see how actual payouts compare to the expected average income per share (basically 0% fee PPS). I can probably improve this with some stats for "all time average" and "average last X time", but I think it is already useful. There are also charts of the pool's luck here: http://bitminter.com/stats/luck

Quick way to get some data without taking months to run a test.

Please add a cumulative graph comparing total earning over time for both methods, or at least a weighted average trend line for the PPLNS reward. Otherwise it is not readily apparent how the rewards from two methods compare over time...