yes but if the PPLNS N is changed that it is around 7 days (according to the diff rules of p2pool), then its not day nor time based.
You mean set n to ~ 9 x Difficulty? (that's seven days shares at the moment)
Topic: [1500 TH] p2pool: Decentralized, DoS-resistant, Hop-Proof pool - page 585. (Read 2591920 times)
March 11, 2013, 09:30:46 AM
You mean set n to ~ 9 x Difficulty? (that's seven days shares at the moment)
March 11, 2013, 09:13:36 AM
yes but if the PPLNS N is changed that it is around 7 days (according to the diff rules of p2pool), then its not day nor time based.
You mean set n to ~ 9 x Difficulty? (that's seven days shares at the moment)
March 11, 2013, 09:03:45 AM
yes but if the PPLNS N is changed that it is around 7 days (according to the diff rules of p2pool), then its not day nor time based.
March 11, 2013, 09:00:54 AM
the PPLNS system should be changed to the last 7 days.
Also I think about increasing difficulty. Current share time avg. 10sec is generating 10-15% DOA/orphans. That's the main reason why people don't use p2pool - no point to do that when you see miner stats with 10% rejected rate and loosing money.
I am voting to increase share time to 30 or 60 seconds and PPLNS time to 7 days. This would decrease stale ratio and attract more miners.
This is why PPLNS should always be in terms of the last n shares, not time. Difficulty and pool hashrate do not have an effect on the score variance if there is no termporal term in the scoring function.
I know what you mean, and it would help in the short term. But id n was set to (for example) 2 x current Difficulty, then you'd never have to change it. The PPLNS shouldn't be any number of days.
What would happen if the pool's percentage of the hashrate increased significantly? You'd want to change the number of days back down in order to reduce the time it took to recieve payment.
It might seem unfair, but on average it makes no difference. Smaller miners will experience variance more on this pool especially since the share difficulty is so high for them.
PPLNS set to shares rather than hours will result in reduced variance for all miners.
March 11, 2013, 08:57:37 AM
the PPLNS system should be changed to the last 7 days.
Also I think about increasing difficulty. Current share time avg. 10sec is generating 10-15% DOA/orphans. That's the main reason why people don't use p2pool - no point to do that when you see miner stats with 10% rejected rate and loosing money.
I am voting to increase share time to 30 or 60 seconds and PPLNS time to 7 days. This would decrease stale ratio and attract more miners.
This is why PPLNS should always be in terms of the last n shares, not time. Difficulty and pool hashrate do not have an effect on the score variance if there is no termporal term in the scoring function.
I know what you mean, and it would help in the short term. But id n was set to (for example) 2 x current Difficulty, then you'd never have to change it. The PPLNS shouldn't be any number of days.
What would happen if the pool's percentage of the hashrate increased significantly? You'd want to change the number of days back down in order to reduce the time it took to recieve payment.
March 11, 2013, 08:55:23 AM
the PPLNS system should be changed to the last 7 days.
Also I think about increasing difficulty. Current share time avg. 10sec is generating 10-15% DOA/orphans. That's the main reason why people don't use p2pool - no point to do that when you see miner stats with 10% rejected rate and loosing money.
I am voting to increase share time to 30 or 60 seconds and PPLNS time to 7 days. This would decrease stale ratio and attract more miners.
This is why PPLNS should always be in terms of the last n shares, not time. Difficulty and pool hashrate do not have an effect on the score variance if there is no termporal term in the scoring function.
I know what you mean, and it would help in the short term. But id n was set to (for example) 2 x current Difficulty, then you'd never have to change it. The PPLNS shouldn't be any number of days.
What would happen if the pool's percentage of the hashrate increased significantly? You'd want to change the number of days back down in order to reduce the time it took to recieve payment.
March 11, 2013, 08:50:55 AM
the PPLNS system should be changed to the last 7 days.
Also I think about increasing difficulty. Current share time avg. 10sec is generating 10-15% DOA/orphans. That's the main reason why people don't use p2pool - no point to do that when you see miner stats with 10% rejected rate and loosing money.
I am voting to increase share time to 30 or 60 seconds and PPLNS time to 7 days. This would decrease stale ratio and attract more miners.
This is why PPLNS should always be in terms of the last n shares, not time. Difficulty and pool hashrate do not have an effect on the score variance if there is no termporal term in the scoring function.
March 11, 2013, 08:50:14 AM
the PPLNS system should be changed to the last 7 days.
Also I think about increasing difficulty. Current share time avg. 10sec is generating 10-15% DOA/orphans. That's the main reason why people don't use p2pool - no point to do that when you see miner stats with 10% rejected rate and loosing money.
I am voting to increase share time to 30 or 60 seconds and PPLNS time to 7 days. This would decrease stale ratio and attract more miners.
March 11, 2013, 08:49:57 AM
the PPLNS system should be changed to the last 7 days.
Also I think about increasing difficulty. Current share time avg. 10sec is generating 10-15% DOA/orphans. That's the main reason why people don't use p2pool - no point to do that when you see miner stats with 10% rejected rate and loosing money.
I am voting to increase share time to 30 or 60 seconds and PPLNS time to 7 days. This would decrease stale ratio and attract more miners.
This is why PPLNS should always be in terms of the last n shares, not time. Difficulty and pool hashrate do not have an effect on the score variance if there is no termporal term in the scoring function.
March 11, 2013, 08:46:20 AM
the PPLNS system should be changed to the last 7 days.
Also I think about increasing difficulty. Current share time avg. 10sec is generating 10-15% DOA/orphans. That's the main reason why people don't use p2pool - no point to do that when you see miner stats with 10% rejected rate and loosing money.
I am voting to increase share time to 30 or 60 seconds and PPLNS time to 7 days. This would decrease stale ratio and attract more miners.
March 11, 2013, 08:34:10 AM
the PPLNS system should be changed to the last 7 days.
March 11, 2013, 07:52:59 AM
Block ETA is now 19hrs and rising, when we go over 24hrs (pool rate under 200GH) mining will become totally unprofitable...
We need more power!
We need more power!
That would be difficult.
If I'm new to p2pool and i see something like this:
Pool Hashrate: 267.3 GH/sEstimated Time to Block: 19h29m Current Round: 1d11h5m
Pool L* (7 days, 30 days, 90 days): 50.4% 77.1% 90.1%
Why should i point my GH's to p2pool? at this time every share count and every PPS Pool is better at this moment.
Why goes the L* so drastical down this time?
Variance is always higher when a pool's hashrate is a smaller proportion of the network's hashrate. p2Pool is just as likely to be extremely lucky.
March 11, 2013, 06:57:58 AM
Block ETA is now 19hrs and rising, when we go over 24hrs (pool rate under 200GH) mining will become totally unprofitable...
We need more power!
We need more power!
That would be difficult.
If I'm new to p2pool and i see something like this:
Pool Hashrate: 267.3 GH/sEstimated Time to Block: 19h29m Current Round: 1d11h5m
Pool L* (7 days, 30 days, 90 days): 50.4% 77.1% 90.1%
Why should i point my GH's to p2pool? at this time every share count and every PPS Pool is better at this moment.
Why goes the L* so drastical down this time?
March 11, 2013, 05:11:52 AM
Block ETA is now 19hrs and rising, when we go over 24hrs (pool rate under 200GH) mining will become totally unprofitable...
We need more power!
We need more power!
Can you please explain in more technical way?
March 11, 2013, 04:55:38 AM
Block ETA is now 19hrs and rising, when we go over 24hrs (pool rate under 200GH) mining will become totally unprofitable...
We need more power!
We need more power!
March 06, 2013, 12:19:38 PM
It would be true, if we hash 10 times/day. But we have 300`000`000`000 hashes / second, so we CAN use it as fair number of tries and it "have" to balance .... in some point 
The past has no outcome on future results. So, right now the current block has a 50% chance of taking between 3.4 to 4 days to find, and 50% of taking more than 4 days to find. (Median and mean are not the same in exponential probability distributions). The problem with p2pool now is that the relative variance is 27% greater than it was before ASICS (from 14 blocks per week +/- 3.7 to 9 blocks a week +/-3, poisson distribution)
Perhaps you can help with a conundrum with which I've wrestled for a while.
The number of shares required to solve n blocks can be described by a (shifted) negative binomial CDF, where size = number of rounds and p = 1/Difficulty.
As an example, the upper tail probability of more than 1.6 x 9 x Difficulty shares being submitted in 9 rounds is 0.05.
If the 10th round is more than 1.6 x Difficulty shares, the probability of this number of shares occurring in ten rounds is less than 0.04, and if the 10th round is less than 1.6 x Difficulty shares, the probability of this number of shares occurring in ten rounds is more than 0.04.
This relates back to rav3n_pl's comment, since it does imply that for an arbitrary number of runs of n rounds where the total shares submitted for the first (n-1) rounds is very unlikely, there will be more runs where CDF for the total of n rounds is closer to the median than further from it.
This means the nth round is more likely to be shorter than longer. Not shorter than average, but just shorter, and therefore luckier.
So I think I'm misunderstanding something about the way this "paradox" should be interpreted. Where am I going wrong?
Good question, heading out but want to think about it a bit more later. I think it may has to do with conditional probability, which ties into gambler's fallacy (unless I'm reading this too quickly). While it is far more likely to have 1 lucky round and 9 unlucky rounds than it is to have 10 unlucky rounds, if you are sitting at the tenth round it is equally probably to have a lucky or unlucky round regardless of what happened in the past. You can also approach this as a conditional probability problem when for isolated random events (P(A given B) = P(A)P(B)/P(B) when P(A) is not affected by event B)
Here is a worked out example showing that although it is far more likely to have a certain outcome in n-tries, the probabillity of an event occuring on the nth try is independent of the previous trials.
http://en.wikipedia.org/wiki/Gambler%27s_fallacy#Explaining_why_the_probability_is_1.2F2_for_a_fair_coin
http://en.wikipedia.org/wiki/Gambler%27s_fallacy#Explaining_why_the_probability_is_1.2F2_for_a_fair_coin
March 06, 2013, 11:22:20 AM
Perhaps p2pools problem is that there are too many people observing it constantly and thus are altering the outcome.
/Schrödinger
/Schrödinger
March 06, 2013, 01:34:19 AM
It would be true, if we hash 10 times/day. But we have 300`000`000`000 hashes / second, so we CAN use it as fair number of tries and it "have" to balance .... in some point
The past has no outcome on future results. So, right now the current block has a 50% chance of taking between 3.4 to 4 days to find, and 50% of taking more than 4 days to find. (Median and mean are not the same in exponential probability distributions). The problem with p2pool now is that the relative variance is 27% greater than it was before ASICS (from 14 blocks per week +/- 3.7 to 9 blocks a week +/-3, poisson distribution)
Perhaps you can help with a conundrum with which I've wrestled for a while.
The number of shares required to solve n blocks can be described by a (shifted) negative binomial CDF, where size = number of rounds and p = 1/Difficulty.
As an example, the upper tail probability of more than 1.6 x 9 x Difficulty shares being submitted in 9 rounds is 0.05.
If the 10th round is more than 1.6 x Difficulty shares, the probability of this number of shares occurring in ten rounds is less than 0.04, and if the 10th round is less than 1.6 x Difficulty shares, the probability of this number of shares occurring in ten rounds is more than 0.04.
This relates back to rav3n_pl's comment, since it does imply that for an arbitrary number of runs of n rounds where the total shares submitted for the first (n-1) rounds is very unlikely, there will be more runs where CDF for the total of n rounds is closer to the median than further from it.
This means the nth round is more likely to be shorter than longer. Not shorter than average, but just shorter, and therefore luckier.
So I think I'm misunderstanding something about the way this "paradox" should be interpreted. Where am I going wrong?
March 05, 2013, 05:36:00 PM
And we've got a winner
March 05, 2013, 04:06:34 PM
It would be true, if we hash 10 times/day. But we have 300`000`000`000 hashes / second, so we CAN use it as fair number of tries and it "have" to balance .... in some point
The past has no outcome on future results. So, right now the current block has a 50% chance of taking between 3.4 to 4 days to find, and 50% of taking more than 4 days to find. (Median and mean are not the same in exponential probability distributions). The problem with p2pool now is that the relative variance is 27% greater than it was before ASICS (from 14 blocks per week +/- 3.7 to 9 blocks a week +/-3, poisson distribution)
Jump to: