We should all remember, when pool luck is down we all lose. Eleuthria loses too. If there was anything he could think of to fix this he would do it. He is probably pulling his hair out like the rest of us. he is also very good at his job and most likely understands this stuff better than we do.
This is not to say that people shouldn't keep coming up with ideas. Personally, i liked the underperforming hardware idea. a lot of the miners have controllers on their boards that may have a problem with numbers that are too large. There is always the possibility that one of us may come up with something that no one else thought of. Brainstorming is good.
But just yelling that the system is broke and to fix it will not accomplish anything. Even if we did determine that some evil organization is interfering or cheating in some way, it is quite likely that nothing could be done about it.
I understand the randomness of Luck and that it is what it is, but i too "feel" that there has to be something causing this but, untill someone can prove their theory we just have to accept things the way they are. nobody complained when btcguild luck was over 100% for months on end. now it has swung the other way. At least until someone can prove otherwise.
Topic: [CLOSED] BTC Guild - Pays TxFees+NMC, Stratum, VarDiff, Private Servers - page 84. (Read 903163 times)
May 06, 2014, 01:26:05 PM
May 06, 2014, 12:43:08 PM
BTC Guild is smaller (relative to network size) than it was months ago. The days of a public facing pool being 25%+ of the network are over, unless they have an obscenely large private farm propping it up. Private mining entities now make up ~40% of the network, leaving only 60% left to fight over. As a result, shifts are no longer setup around 15-20x difficulty, they're setup to be roughly 10x difficulty. This means bad luck shifts are no longer 8-10 blocks. A neutral shift right now is 8.75 blocks (meaning 9 blocks or more in a shift is positive luck).
I understand and agree that BTCGuild's size is decreasing in terms of network percentage, as are the other public facing pools. However, moving the average luck down to 8.75 b.p.h doesn't introduce a large margin of error on the values I stated in my earlier post.
The bottom line is the luck is shit, and has been long enough that it's actually dragged BTCGuild's "all-time" luck sub 100. I recognize that luck is out of our collective hands, so I've not come in here pissing and moaning about it. However, the sustained streak is starting to raise flags, and I find it interesting that Eligius and BTCGuild are experiencing similar trends extending from a similar point in time. I don't blame you for the "luck" or for nefarious acts, and in the event that it's actually bad hardware I fail to see any way to counter it.
May 06, 2014, 11:49:05 AM
2) There have been no backend changes for ScryptGuild. There have been no changes to the backends period in the last 9 months. 6 of those months were positive on luck. The last 2 were not. The 9th month isn't available since luck only started being tracked across multiple difficulties 8 months ago (prior to that it only showed the most recent shifts of the current difficulty).
eleuthria, you should just eliminate the luck graph on the site entirely. Even though the metric is measurable, it cannot be acted upon, so, in this instance, the measurement is not valuable. And if I have to read the "bad luck" posts for 4 more months (until we are even again), I'm going to kill myself.
P.S. Yes, I do know that the 4 month number is not really the guaranteed length of time it will take until the variances work themselves out - it was a joke.
Bad luck graph or not. It's pretty fucking obvious when there's a continual stream of >6 blocks found per shift when historically it's been >9.
BTC Guild is smaller (relative to network size) than it was months ago. The days of a public facing pool being 25%+ of the network are over, unless they have an obscenely large private farm propping it up. Private mining entities now make up ~40% of the network, leaving only 60% left to fight over. As a result, shifts are no longer setup around 15-20x difficulty, they're setup to be roughly 10x difficulty. This means bad luck shifts are no longer 8-10 blocks. A neutral shift right now is 8.75 blocks (meaning 9 blocks or more in a shift is positive luck).
May 06, 2014, 09:40:35 AM
2) There have been no backend changes for ScryptGuild. There have been no changes to the backends period in the last 9 months. 6 of those months were positive on luck. The last 2 were not. The 9th month isn't available since luck only started being tracked across multiple difficulties 8 months ago (prior to that it only showed the most recent shifts of the current difficulty).
eleuthria, you should just eliminate the luck graph on the site entirely. Even though the metric is measurable, it cannot be acted upon, so, in this instance, the measurement is not valuable. And if I have to read the "bad luck" posts for 4 more months (until we are even again), I'm going to kill myself.
P.S. Yes, I do know that the 4 month number is not really the guaranteed length of time it will take until the variances work themselves out - it was a joke.
Bad luck graph or not. It's pretty fucking obvious when there's a continual stream of >6 blocks found per shift when historically it's been >9.
May 06, 2014, 09:01:14 AM
2) There have been no backend changes for ScryptGuild. There have been no changes to the backends period in the last 9 months. 6 of those months were positive on luck. The last 2 were not. The 9th month isn't available since luck only started being tracked across multiple difficulties 8 months ago (prior to that it only showed the most recent shifts of the current difficulty).
eleuthria, you should just eliminate the luck graph on the site entirely. Even though the metric is measurable, it cannot be acted upon, so, in this instance, the measurement is not valuable. And if I have to read the "bad luck" posts for 4 more months (until we are even again), I'm going to kill myself.
P.S. Yes, I do know that the 4 month number is not really the guaranteed length of time it will take until the variances work themselves out - it was a joke.
May 06, 2014, 07:43:08 AM
Not to hijack the thread, but relating to luck, the scryptguild pool seems to have not hit any blocks for any coins in hours.
May 06, 2014, 07:38:32 AM
At what point would you consider it to definitely not be a bad luck streak and that something else is wrong?
The longer the time period, the less likely it is that good or bad luck will persist.
For example, it's a lot easier to flip a coin and have it come up heads twice in a row than it is to have it come up heads ten times in a row.
May 06, 2014, 12:43:53 AM
At what point would you consider it to definitely not be a bad luck streak and that something else is wrong?
May 05, 2014, 10:58:00 PM
People keep mentioning "low difficulty" as if this made a difference? I was under the impression that the difficulty of a particular hash does not have bearing on whether it is the "solve" for block hash or not, that even a GPU still has the slightest chance of catching the correct timing and submitting the block solving hash, however unlikely it may be.
I think you might have misunderstood my point. I probably am not expressing myself clearly.I do not mean the difficulty you are referring to.
I do not mean a malicious withholding attack (that is not in the attacker's financial interest).
I am referring to a machine that was made but has errors either in the chip or the firmware that means it gives valid solutions to work UP TO A CERTAIN POINT. So, when you run tests on it on OLD blocks, it solves them. But, it has some flaw that prevents it from solving blocks as difficulty increased. So, it essence it is now USELESS but still produces valid work (but the best share it sends is substantially below the current difficulty -- so it never solves a block). This would be very much like what entropy referred to as a 46 card deck (when you are expecting to see 4 kings every 52 cards).
Is that clearer? So, not some purposed withholding of solutions just incompetence. Think of the people who have made miners (BFL, Avalon, Hashfast, KNC, Cointerra). They have all had huge time constraints and failures. Is it so far fetched that one of them (or someone else) made 3-10 PH/s of miners that "work" but do not solve blocks at these levels?
This is one of the things I've been discussing with the other pool operators. The bad luck trend for BTC Guild and Eligius seemed to start around the time difficulty crossed 2.1b. This is the limit of a signed 32-bit integer, one of the most common numeric variables used for programmers. It's quite rare that people need > 2.1b (4.2b for unsigned). The difficulty, growing exponentially, ramped up to this number extremely quickly. I realized that BTC Guild's internal reporting used a signed 32-bit int for displaying difficulty on a status screen. I wasn't sure if the program would crash or simply loop around, so I did server maintenance around ~1b diff to update it to use a double variable instead.
The problem here is that we've been unable to identify any hardware that is broken in this manner so far. It doesn't make sense that such an error would even happen at the hardware level, and most ASICs use cgminer/bfgminer, which we know is not the source if there is a problem of this nature.
Which if we do, would you share it with the other pools heh
May 05, 2014, 10:54:27 PM
Is that clearer? So, not some purposed withholding of solutions just incompetence. Think of the people who have made miners (BFL, Avalon, Hashfast, KNC, Cointerra). They have all had huge time constraints and failures. Is it so far fetched that one of them (or someone else) made 3-10 PH/s of miners that "work" but do not solve blocks at these levels?
There's nothing different about hashes that end up being high difficulty or low at the hardware level. You have to actively look for them and remove them to differentiate them which is quite doable but is an active act of malice, not incompetence. May 05, 2014, 10:38:18 PM
People keep mentioning "low difficulty" as if this made a difference? I was under the impression that the difficulty of a particular hash does not have bearing on whether it is the "solve" for block hash or not, that even a GPU still has the slightest chance of catching the correct timing and submitting the block solving hash, however unlikely it may be.
I think you might have misunderstood my point. I probably am not expressing myself clearly.I do not mean the difficulty you are referring to.
I do not mean a malicious withholding attack (that is not in the attacker's financial interest).
I am referring to a machine that was made but has errors either in the chip or the firmware that means it gives valid solutions to work UP TO A CERTAIN POINT. So, when you run tests on it on OLD blocks, it solves them. But, it has some flaw that prevents it from solving blocks as difficulty increased. So, it essence it is now USELESS but still produces valid work (but the best share it sends is substantially below the current difficulty -- so it never solves a block). This would be very much like what entropy referred to as a 46 card deck (when you are expecting to see 4 kings every 52 cards).
Is that clearer? So, not some purposed withholding of solutions just incompetence. Think of the people who have made miners (BFL, Avalon, Hashfast, KNC, Cointerra). They have all had huge time constraints and failures. Is it so far fetched that one of them (or someone else) made 3-10 PH/s of miners that "work" but do not solve blocks at these levels?
This is one of the things I've been discussing with the other pool operators. The bad luck trend for BTC Guild and Eligius seemed to start around the time difficulty crossed 2.1b. This is the limit of a signed 32-bit integer, one of the most common numeric variables used for programmers. It's quite rare that people need > 2.1b (4.2b for unsigned). The difficulty, growing exponentially, ramped up to this number extremely quickly. I realized that BTC Guild's internal reporting used a signed 32-bit int for displaying difficulty on a status screen. I wasn't sure if the program would crash or simply loop around, so I did server maintenance around ~1b diff to update it to use a double variable instead.
The problem here is that we've been unable to identify any hardware that is broken in this manner so far. It doesn't make sense that such an error would even happen at the hardware level, and most ASICs use cgminer/bfgminer, which we know is not the source if there is a problem of this nature.
May 05, 2014, 10:19:06 PM
People keep mentioning "low difficulty" as if this made a difference? I was under the impression that the difficulty of a particular hash does not have bearing on whether it is the "solve" for block hash or not, that even a GPU still has the slightest chance of catching the correct timing and submitting the block solving hash, however unlikely it may be.
I think you might have misunderstood my point. I probably am not expressing myself clearly.I do not mean the difficulty you are referring to.
I do not mean a malicious withholding attack (that is not in the attacker's financial interest).
I am referring to a machine that was made but has errors either in the chip or the firmware that means it gives valid solutions to work UP TO A CERTAIN POINT. So, when you run tests on it on OLD blocks, it solves them. But, it has some flaw that prevents it from solving blocks as difficulty increased. So, it essence it is now USELESS but still produces valid work (but the best share it sends is substantially below the current difficulty -- so it never solves a block). This would be very much like what entropy referred to as a 46 card deck (when you are expecting to see 4 kings every 52 cards).
Is that clearer? So, not some purposed withholding of solutions just incompetence. Think of the people who have made miners (BFL, Avalon, Hashfast, KNC, Cointerra). They have all had huge time constraints and failures. Is it so far fetched that one of them (or someone else) made 3-10 PH/s of miners that "work" but do not solve blocks at these levels?
May 05, 2014, 08:55:49 PM
People keep mentioning "low difficulty" as if this made a difference? I was under the impression that the difficulty of a particular hash does not have bearing on whether it is the "solve" for block hash or not, that even a GPU still has the slightest chance of catching the correct timing and submitting the block solving hash, however unlikely it may be.
Don't recall seeing "low difficulty" anywhere. The miner side difficulty has no influence on whether or not their hash will solve a block.
EDIT: NVM, saw it. My point above still stands.
May 05, 2014, 08:49:03 PM
People keep mentioning "low difficulty" as if this made a difference? I was under the impression that the difficulty of a particular hash does not have bearing on whether it is the "solve" for block hash or not, that even a GPU still has the slightest chance of catching the correct timing and submitting the block solving hash, however unlikely it may be.
May 05, 2014, 08:18:37 PM
I am not attacking or bitching or accusing BTCGuild in anyway, I believe it is a group of bad (defective) miners.
Since we solo mined, we know it is not Avalon's, KNC Juipters nor Cointerras as they have all performed to expectations. But that leaves a lot of other hardware out there and also leaves open the possibility of it just being a dark miner who made their own chips. I am sure someone could audit each miner very quickly if they had them available.
It is not likely that every large pool is having "bad luck" without someone being the beneficiary. Some group of miners is running good luck in a large way or else blocks would be averaging 12-13 minutes. Doesn't that mean it is likely some portion of the pool is only solving low difficulty work and can never produce a solution that solves a block?
That would mean the one safe mining would be at small pools (where the free riders could be seen), solo mining (need more than 600TH/s right now) or pools that do some sort of "work audit" on chronically unlucky miners or brand new miners (by sending them recently solved work and see if they send back the correct result).
Sorry for the distraction and this is the last I will post about it here, but that is how I feel.
Since we solo mined, we know it is not Avalon's, KNC Juipters nor Cointerras as they have all performed to expectations. But that leaves a lot of other hardware out there and also leaves open the possibility of it just being a dark miner who made their own chips. I am sure someone could audit each miner very quickly if they had them available.
It is not likely that every large pool is having "bad luck" without someone being the beneficiary. Some group of miners is running good luck in a large way or else blocks would be averaging 12-13 minutes. Doesn't that mean it is likely some portion of the pool is only solving low difficulty work and can never produce a solution that solves a block?
That would mean the one safe mining would be at small pools (where the free riders could be seen), solo mining (need more than 600TH/s right now) or pools that do some sort of "work audit" on chronically unlucky miners or brand new miners (by sending them recently solved work and see if they send back the correct result).
Sorry for the distraction and this is the last I will post about it here, but that is how I feel.
Luck is not a zero sum game. The difficulty can continue to rise even if the entire network was having bad luck, as long as the growth of new speed is greater than the deficit caused by poor luck. Work audits are not possible with stratum, since each user has an independent ExtraNonce1, meaning their nonces will create radically different results than another user's, given the same template and ntime. Additionally, a truly malicious withholding attack of large size is unblockable since they could simply make new accounts. If they have enough speed to cause noticeable harm to a pool's production, and enough money to literally throw it away (since doing a withholding attack against PPLNS is costing them money for every block they withhold) they have more than enough money to proxy that attack through new IP ranges and accounts.
May 05, 2014, 08:10:26 PM
I am not attacking or bitching or accusing BTCGuild in anyway, I believe it is a group of bad (defective) miners.
Since we solo mined, we know it is not Avalon's, KNC Juipters nor Cointerras as they have all performed to expectations. But that leaves a lot of other hardware out there and also leaves open the possibility of it just being a dark miner who made their own chips. I am sure someone could audit each miner very quickly if they had them available.
It is not likely that every large pool is having "bad luck" without someone being the beneficiary. Some group of miners is running good luck in a large way or else blocks would be averaging 12-13 minutes. Doesn't that mean it is likely some portion of the pool is only solving low difficulty work and can never produce a solution that solves a block?
That would mean the one safe mining would be at small pools (where the free riders could be seen), solo mining (need more than 600TH/s right now) or pools that do some sort of "work audit" on chronically unlucky miners or brand new miners (by sending them recently solved work and see if they send back the correct result).
Sorry for the distraction and this is the last I will post about it here, but that is how I feel.
Since we solo mined, we know it is not Avalon's, KNC Juipters nor Cointerras as they have all performed to expectations. But that leaves a lot of other hardware out there and also leaves open the possibility of it just being a dark miner who made their own chips. I am sure someone could audit each miner very quickly if they had them available.
It is not likely that every large pool is having "bad luck" without someone being the beneficiary. Some group of miners is running good luck in a large way or else blocks would be averaging 12-13 minutes. Doesn't that mean it is likely some portion of the pool is only solving low difficulty work and can never produce a solution that solves a block?
That would mean the one safe mining would be at small pools (where the free riders could be seen), solo mining (need more than 600TH/s right now) or pools that do some sort of "work audit" on chronically unlucky miners or brand new miners (by sending them recently solved work and see if they send back the correct result).
Sorry for the distraction and this is the last I will post about it here, but that is how I feel.
May 05, 2014, 08:06:48 PM
Thanks eleuthria. I do think you are an honest pool operator, and hope the bad luck streak turns around soon but understand that it is out of your control.
On a positive note, the next difficulty increase looks like it may be in the single digits for the first time since May of 2013, so that's a very good thing! Perhaps the difficulty is finally starting to level out!
On a positive note, the next difficulty increase looks like it may be in the single digits for the first time since May of 2013, so that's a very good thing! Perhaps the difficulty is finally starting to level out!
May 05, 2014, 08:05:26 PM
My hardware was split between BTC Guild and Ghash evenly. I am way underperforming daily now there as well.
I refrained from mentioning GHash and Discus Fish because they don't have publicly available historical stats, at least that I was able to find. Eligius has a CSV export for it's block history, and p2pool has a 3-month luck chart which is reported in the same way BTC Guild reports luck (% compared to expected earnings, rather than CDF). Whether or not they have had good luck, neutral luck, or similar bad luck is a mystery to me, so I only report on the comparable pools that I'm aware of.
Sorry no offense.. I have taken down power Hungary rigs. Getting some newer stuff hopefully put together in a few days and going to aim it at the pool. Good luck everybody
May 05, 2014, 07:54:34 PM
My hardware was split between BTC Guild and Ghash evenly. I am way underperforming daily now there as well.
I refrained from mentioning GHash and Discus Fish because they don't have publicly available historical stats, at least that I was able to find. Eligius has a CSV export for it's block history, and p2pool has a 3-month luck chart which is reported in the same way BTC Guild reports luck (% compared to expected earnings, rather than CDF). Whether or not they have had good luck, neutral luck, or similar bad luck is a mystery to me, so I only report on the comparable pools that I'm aware of.
May 05, 2014, 07:51:55 PM
My hardware was split between BTC Guild and Ghash evenly. I am way underperforming daily now there as well.
Jump to: