ethpool.org VS ethermine.org

13»

Comments

  • o0ragman0oo0ragman0o Member, Moderator Posts: 1,288 mod
    Hmm been on a call-out and have missed this this 'wonderful' dialog (excuse the sarcasm)

    Might I point out that the reason this debate is so heated is because ethpool's algo does introduce a number of statistical and arbitrary complexities which each give rise to more emergent complex behavior in the leaderboard and the bias of payouts. These complexities present as a significant deviation from the very simplistic Piosson distribution of ethash because the pool additionally bonds miners into additional competition with each other at the top and the bottom of the board.

    There are at least three arbitrary mechanisms, which have absolutely no relationship with the POW designed statistics and which present additional statistical and causal terms within the whole probability:
    1. Once a block is found, credits are deducted from the miner equal to the second miner's current credits. This is a completely artificial mechanism whereby the state of one miner directly effects the state of another miner by a completely random degree.
    2. The payout of Uncles to the top miner without credits adjustment forces the credits required well above network difficulty. While this act's as a novel 'jackpot', it forces the rest of the pool to mine above difficulty for the rest of the time.
    3. The concept of a leaderboard presents an additional competition between miners to not just get to the top, but be at the top when the pool finds a block. Again we have a situation where the state of a miner is artificially determined by the state of another miners.
    These additional dynamics layered over the basic Piosson distribution take the pool's payout behavior well outside of statistics and into the realm of complex or even chaotic behavior in which the exact nature of emergent bias is difficult realise. However there appears a very strong bias toward the pool's highest hashing miners who are paid out at significantly less than difficulty by profiting from the pools lowest hashing miners who are paid out at significantly higher difficulty. This is simply because the big hash has so much more advantage in these additional arbitrary competitions.

    Assessing the largest hash in relation to the above list we have...
    1. When the largest miner attains top, credits required will accelerate away from the rest of the pool until a block is found. When the block is found, the deduction of second placed credits from the top miner which transfers and accumulates that accelaration advantage via a higher reentry into the leaderboard meaning they never have to mine more than the required difficulty.
    2. The Uncle payouts for the top most hashing miners at the top of the leaderboard statistically double the period in which the miner is accelarating away from the rest of the pool and so yield an exponential increase to the bias toward the top miner.
    3. The topmost hashing miner does not have to compete with lesser miners for top placement. Once it is there, it's there and doesn't have to worry about a larger miner jumping ahead to steal it's block.
    Assessing the smallest hash we have..
    1. When the smallest miner finally gets a block, it will never have a large margin between itself and the second place and so is dumped at the bottom of the table every time. It must mine the entire required credits every round.
    2. If the smallest miner attains top and is award an uncle, they will still lose top before the next block is found
    3. The least hashing miner must compete with every other miner in all these additional mechanisms and is the most disadvantaged in all those mechanisms
    So it is no wonder we are seeing a 15GHs miner on the pool. Someone out there (perhaps even another pool proxying through to ethpool) has smarts enough to realise this bias and they are getting paid out well in excess of the Piosson distribution to the detriment of the of every other miner on ethpool.

    (@dr_pra)

  • zorvalthzorvalth Member Posts: 174
    edited March 2016

    Hmm been on a call-out and have missed this this 'wonderful' dialog (excuse the sarcasm)

    Assessing the smallest hash we have..

    1. When the smallest miner finally gets a block, it will never have a large margin between itself and the second place and so is dumped at the bottom of the table every time. It must mine the entire required credits every round.
    2. If the smallest miner attains top and is award an uncle, they will still lose top before the next block is found
    3. The least hashing miner must compete with every other miner in all these additional mechanisms and is the most disadvantaged in all those mechanisms
    So it is no wonder we are seeing a 15GHs miner on the pool. Someone out there (perhaps even another pool proxying through to ethpool) has smarts enough to realise this bias and they are getting paid out well in excess of the Piosson distribution to the detriment of the of every other miner on ethpool.



    Not true :) lets compare 5GH miner and 25mh miner. There is 200 times difference. So its completely normal 5GH miner to start with 200 times more credits after the reward. And if you watch the credit line long enough you will see it will be exactly as this. You start every new block with proportionally the same credits that everyone start. If you start with less with the average that means you took the block earlier if you start with more thats means you got the block a little later then supposed to.

    Also my screenshots from up there prove you wrong. I compared several 3-5GH miners with several 200-500mh miners no difference in the end it will be the same with 20-50mh/ miners. The only problem with the small miners is the one i described earlier with the rapidly increasing the difficulty of the network.
  • o0ragman0oo0ragman0o Member, Moderator Posts: 1,288 mod
    zorvalth said:


    Not true :) lets compare 5GH miner and 25mh miner. There is 200 times difference. So its completely normal 5GH miner to start with 200 times more credits after the reward. And if you watch the credit line long enough you will see it will be exactly as this. You start every new block with proportionally the same credits that everyone start. If you start with less with the average that means you took the block earlier if you start with more thats means you got the block a little later then supposed to.


    @zorvalth, You entirely missed the point of the first and second miner's credits being artificially coupled. Neither do we have access to the entire leader board to actually see what's going on at the bottom.

    The fairer system to prevent coupled miners would be to simply subtract the difficulty of the block found from the miners credits.
    Also my screenshots from up there prove you wrong.
    Your screen shots are of the estimated payout as shown in ethpool which are worked out according to the standard Piosson distribution and therefore fail to take into account the poorly understood biasing mechanisms I've outlined. Those estimates are rendered completely inaccurate as attested to by so many small hash miners on ethpool while the big hash miners are either very happy or very quite. You've simply taken those numbers at face value and have in no way proved me wrong.
  • dr_pradr_pra Member Posts: 445 ✭✭✭
    His screenshots compare predicted earnings based on the miners hashrate and current network difficulty to the actual earnings for each miner. How do you draw the conclusion that the presented estimates calculated similar to how each other mining calculator works are wrong?
  • zorvalthzorvalth Member Posts: 174
    edited March 2016
    First as @dr_pra explained its impossible to deduct the current diff because in case of good luck of the pool will be miners with negative start and they easily can change the payment address and start from 0. Also since the new uncle distribution the connection between the difficulty and credits in ethpool so the idea is impossible.

    Second you obviously didnt take a look in my shots or didnt understand them because they prove that in the end it doesnt matter what everybody think of the system the payments are close to the ether calculator no matter the MH of the miner. And take another look my screenshots shows two important things:
    1.Estimate earnings based on the MH
    2. Actual Ether/h rate based on the payments in the last 24 hours.

    and the very close margin between the 1 and 2 prove you wrong.
  • o0ragman0oo0ragman0o Member, Moderator Posts: 1,288 mod
    zorvalth said:

    First as @dr_pra explained its impossible to deduct the current diff because in case of good luck of the pool will be miners with negative start and they easily can change the payment address and start from 0.

    Subtracting block diff and resetting 0 if the result is negative is hardly an impossible thing. But is does decouple the miners.
    Also since the new uncle distribution the connection between the difficulty and credits in ethpool so the idea is impossible.
    The uncles regime should inflate the required credits above diff to the percent at which they are found. What I was point out is that the Uncles effectively give the top miner double time at the top of the leaderboard and blow out the required credits for everyone else. It presents as highly divergent mechanism from the Poisson distribution which leads to inaccurate and unpredictable estimates. To be a killjoy, I would suggest a fairer mechanism would be to subtract credits proportional to the uncle payout rather than full diff. credits -= block_diff * Uncle_pay/5
    Second you obviously didnt take a look in my shots or didnt understand them because they prove that in the end it doesnt matter what everybody think of the system the payments are close to the ether calculator no matter the MH of the miner. And take another look my screenshots shows two important things:
    1.Estimate earnings based on the MH
    2. Actual Ether/h rate based on the payments in the last 24 hours.
    How can actual Eth in the last 24hs possibly be a useful measure for small hash which might only get paid out once in every 1 or 2 weeks?

    @dr_pra @zorvalth, I shall take another look at the screen shots.
  • zorvalthzorvalth Member Posts: 174
    Thats why i compared 200-500mh vs 3-5GH(10 times bigger) miners, as there is no difference there it wont be with these with 50mh(5 times smaller)
  • o0ragman0oo0ragman0o Member, Moderator Posts: 1,288 mod
    @dr_pra @zorvalth, I'm sorry but the screen caps make no sense to me at all.
    zorvalth said:

    I looked in to it, i got an extension which is calculating the eth based on the payments and based on the hasrate(as the ether calculators). If what you are saying is right big miners should have muvh better rate then the small miners.

    I see the estimates table and I see a 'Rate: xxx E/h'. What is the supposed real number here? You said that you're using some extension and does that look up the actual payments to miners? And how is that going to be an accurate number when small miners might take weeks to get an actual payout? And then bugger off for greener pastures?

    I also object to your classification of 'small miners' none of which in your screen caps had hash less than 250Mhs. I'm running 12 GPU's and mining at 200Mhs. When I say small hash, I'm talking about 1 and 2 GPU rigs, e.g <50Mhs who stand to lose out the most. What you've called small miners, I would class as medium and would hazard to guess that hash own by individual miners across the network would follow a power law.

    Further, your assumption that 50Mhs is simply 1/10 of 500 mh/s and should scale linearly is correct under a piosson distribution, yet I'm describing the emergent non-linearities which predict bias and inefficiency on ethpool so again is an inaccurate statment.


  • blueboxbluebox Member Posts: 181 ✭✭
    I think we've all had quite enough for the time being; almost sorry I brought it up...

    @o0ragman0o
    Thank you for speaking my language, I do follow enough of your explanation, and it covers what I suspect is the flaw — or simply a disadvantage — to small miners (<100MH) versus the 500MH+ miners out there. ethpool presents an "arms race" component in that the bigger you get the better it is for you, to the detriment of the smallest folks. I refuse to join the arms race by buying more and more GPU's after learning the hard way years ago, and would simply like to "participate" in the technology (while making a small buck or two along the way).

    @zorvalth I also understand what you're trying to present, however my suspicion regarding the complexities of the payout algorithm for small miners like me (75MH) on ethpool still remain. My empirical observations follow more in line with ragman's probable explanation. I mined on ethpool before the homestead fork when diff went from 17T to 19T. Time to block was shown as 5 days at the start (after 24hrs) and the target date didn't change after several days, estimated daily payout was shown as 1.1ETH (!?). Mining after the fork on ethermine during the same diff rise (17-19T) and I averaged 40% more over 5 days than ethpool (1.45ETH). Take it as you wish...

    @dr_pra You must have better insight/access to real miner production that we don't have from the website, but I doubt it's worth your time to settle trivial arguments about small miners. ;)
  • zorvalthzorvalth Member Posts: 174
    @o0ragman0o and @bluebox get me few miner address with small hash who kept steady hashrate for a couple weeks and i will show you there actual payment rate.
  • blueboxbluebox Member Posts: 181 ✭✭
    @zorvalth I'm only pointing to ethermine at the moment, and one of my windows boxes decided to reboot itself overnight, even though updates are turned off. Really hate windows, don't like ubuntu any better, I'm the RHEL/CentOS data center type.

    I'm not going to switch back and forth again, perhaps you can point a couple of your spare cards at each during the same time period and prove it to us; I don't have the hash and time to do so again, this is more hobby to me... given 75MH you can tell. :)
  • rismondorismondo Member Posts: 58
    I can try it all, just send me gpus and I will do all kind of testing to my wallet :) Where do you want me to put adrress? haha
  • hasherhasher Member Posts: 642 ✭✭✭
    A little update. My farm is now at 1GH. Should I still be mining at ethermine.org? I have a feeling ethpool will be more profitable for me with the uncles etc..
  • metalorkmetalork Member Posts: 15
    @hasher why dont you try a small pool like ethermine.rocks it pays very well, only 0.1 fee. you should try for a couple fo days. @ethermine_rocks can explain a bit more...
  • hasherhasher Member Posts: 642 ✭✭✭
    Time has passed.. and i'm at 2.4GH now :)
    Still at ethermine, but tempted to jump to ethpool.. should I?
  • RabassoRabasso Member Posts: 151 ✭✭
    @hasher I don't think so....in ethpool you have this whale miners (21 gh/s) that are mining a block every hr, so basically they are making 0.08 more eth by doing it in the pool, that doesn't count the uncle rewards, while pushing everyone else out the list. they should go solo mining. This pools should not accept whale miners, they should have a cap of 5gh/s max, because the fact is that we would make more eth by solomining, and that is what im going to do, stick it to the Men.

    when that miner came online he registered 60 gh/s averaging 1block every 30 minutes, so taking 10% of the total pool profit and just producing a fraction of it. Don't know how he did it but it was a burst that has lasted 4 hrs.
    http://ethpool.org/miners/5247a7b6e8088b83cfb755a5896c114fd24083d9

    @ethpool.pp.ua @EthPool_utocat
  • RabassoRabasso Member Posts: 151 ✭✭
    zorvalth said:

    Hmm been on a call-out and have missed this this 'wonderful' dialog (excuse the sarcasm)

    Assessing the smallest hash we have..

    1. When the smallest miner finally gets a block, it will never have a large margin between itself and the second place and so is dumped at the bottom of the table every time. It must mine the entire required credits every round.
    2. If the smallest miner attains top and is award an uncle, they will still lose top before the next block is found
    3. The least hashing miner must compete with every other miner in all these additional mechanisms and is the most disadvantaged in all those mechanisms
    So it is no wonder we are seeing a 15GHs miner on the pool. Someone out there (perhaps even another pool proxying through to ethpool) has smarts enough to realise this bias and they are getting paid out well in excess of the Piosson distribution to the detriment of the of every other miner on ethpool.



    Not true :) lets compare 5GH miner and 25mh miner. There is 200 times difference. So its completely normal 5GH miner to start with 200 times more credits after the reward. And if you watch the credit line long enough you will see it will be exactly as this. You start every new block with proportionally the same credits that everyone start. If you start with less with the average that means you took the block earlier if you start with more thats means you got the block a little later then supposed to.

    Also my screenshots from up there prove you wrong. I compared several 3-5GH miners with several 200-500mh miners no difference in the end it will be the same with 20-50mh/ miners. The only problem with the small miners is the one i described earlier with the rapidly increasing the difficulty of the network.
    @zorvalth Ok, do compare them now with 23gh/s miner?
  • hasherhasher Member Posts: 642 ✭✭✭
    So just to report back... I have 2.5GH now. I mined at ethpool for 4 days to see what would happen. First payment was with an uncle! But then after that, no more uncles for the remaining days I was there. I actually mined less eth at ethpool than what I would have received at ethermine. Moved all my rigs back to Ethermine, and now i'm mining more eth per day (roughly an extra 2 ETH per day than at ethpool)
  • RabassoRabasso Member Posts: 151 ✭✭
    hasher said:

    So just to report back... I have 2.5GH now. I mined at ethpool for 4 days to see what would happen. First payment was with an uncle! But then after that, no more uncles for the remaining days I was there. I actually mined less eth at ethpool than what I would have received at ethermine. Moved all my rigs back to Ethermine, and now i'm mining more eth per day (roughly an extra 2 ETH per day than at ethpool)

    Yes I also changed to ethermine, that gives me better returns, My first payment at ethpool was also with an uncle but the 2 payments after that averaged 8 hr more to produce so i was losing 1 eth per payment in average, this pool is for whales, not for under 5GH/s. there is a bunch of little guys who should do their math and change pools because their delays are way more than ours.... and loses would be way higher too. all this takes me to the point that we would probably make more money solo mining, you where solo mining @hasher? how did that go? if it was easy we should create pools for no more than 100 miners or so, i mean how hard can it be? hosting servers
  • SamNeilSamNeil Member Posts: 3
    Hi Everyone, I need an urgent assistance. I am currently mining ETH on dwarfpool.com. Lately, the stats are malfunctioning. I would like to mine on ethpool.org. Can someone help me with out with this please? I need the full script to plug in the simplemining. Thank you
  • MiningAmazingMiningAmazing Member Posts: 9
    edited January 18
    Hi,

    Can you explain this?




    Why credits are that high and quite bigger than the credit amount of the 2nd?

    This is having direct impact on the expected time that will take to my rig to reach "the top" (I've only 250 mH/s, upgrading to 400...), this situation is quite disappointing :'(

    Thanks!!!
Sign In or Register to comment.