Topic: [CLOSING] GHash.IO mining pool official page - page 12. (Read 149381 times)

I only ever used this pool and had no idea until reading this, have any recommended alternative?

They've done it so that it can be explained away with random variance. As mentioned above, they never make more than expected in any given day, but often less. I don't know if they  hijack blocks or simply claim to have 25% higher hash rate than they have, but something is terribly wrong here. I understand there are people with more insight into statistics than I have, but mathematically this is very improbable. But I have some experience with companies in a desperate situation, from that perspective this is a no-brainer.

Check the ckpool kano.is thread, one of the users there has put together something I think is mobile friendly. Otherwise you would need to look at something like MultiMiner from nwools but I think that is more of a monitor on the hardware side.

Any pools where you can monitor your miners with your mobile ?

the consensus seems to be leaning that way but they would never admit it.

I don't even think they check this thread anymore as they never responded to the devcoin question either.

is this pool cheating its users? Every few days or so I am making 50% less than what I should. I never have any 150% days to make up for it.

Thing is, this is always hard to grasp at first. As you know it takes a looooong time until your intuitions about mining are close to reality.


Edit:

Here's a plot of the 95% CI lower and upper bounds as a percentage of and a function of blocks solved per time period:

Organorcorti,

I don't think graymatter knows who he's talking statistics with. Maybe point him to your website.

http://organofcorti.blogspot.com/

Although I am surprised its as much as 20% I don't doubt your numbers in the least.

The variance *is* +/- 20%. You can calculate it for yourself if you like:


That is, obtain the chi squared quantiles at 2.5 and 97.5% for your observed number of blocks, and divide by 2. For example, if the network solves 1008 block and GHash.IO solves (as you suggest) 15% of them, then variance would be +/- 17%. You can check this for yourself.


Edit: Use wolfram alpha if you like. the following links calculate the lower and upper bounds for 150 blocks:

Lower:
http://www.wolframalpha.com/input/?i=InverseCDF%5BChiSquareDistribution%5B300%5D%2C+0.025%5D%2F2%2F150
Upper:
http://www.wolframalpha.com/input/?i=InverseCDF%5BChiSquareDistribution%5B302%5D%2C+0.975%5D%2F2%2F150

in response to: More like 9% of the network, which means a weekly variance in blocks of ~ +/- 20%.

my point was the variance shouldn't be + or - 20% of a variance.  The large the pool is, the more blocks it finds, the less that luck plays into a factor.  Its called statistic probability.  The more quarters you flip, the closer to a 50/50 number you will get to.  For example, flip 10 quarters you might get 7 heads 3 tails, but over 100 units, it gets closer to 50%, over 1 million units, you are + or - a fraction of a percent.

The same logic applies to large pools, more hashrate / larger % of the bitcoin network = more stable luck.  My point was, ghash in December was a large percentage of the network, and should have not been so vulnerable to large luck spikes as you suggest, + or minus 20%.  Its less of a statistical anomaly.  

Sorry, I have no idea what you mean. What does "What we found when looking at numbers was it was consistently negative 10-20% week after week." mean in terms of hashrate?

In the second paragraph you're discussing luck - I was only commenting on hashrate.

shutting down their cloud operation probably was a huge hit to their bottom line.
THey obviously didn't reinvest in it. Or it would still be running.

Yes, and to note it was about 15% of the network through December, and while there can be a + or - 16% or so, the variance shouldn't be that large as statistically it should be a + and - effect over multiple weeks.  What we found when looking at numbers was it was consistently negative 10-20% week after week.

In order to accomplish this you're talking about a 1/1000 probability that the luck would be constantly negative over a larger timespan..  There should have at least been one or two weeks over the last couple of months that came out with 'positive luck' this was not the case.  This used to be the case about 3-4 months ago with some daily payouts exceeding expected.  My guess is Ghash is going under and they're doing everything they can to stay afloat.

More like 9% of the network, which means a weekly variance in blocks of ~ +/- 20%.



It's actually around 96 blocks (+/- 20%) for the week.

You used blockchain.info's pie chart and extrapolated from there?

I left quite a while back due to seeing such poor returns for a zero fee pool. I found that it was greater than 6% less than expected and I was monitoring it for months checking it daily at the same time each day.

I'm unable to say whether or not they are cheating anyone but there's something wrong.

Anyone else?

I agree with you, I noticed it but didn't run the numbers.
I'm sure if I did it would match up to yours though.

Well, the really scary thing is to look at a month.

Averaging 9.3 coins per day at most recent difficulty means over 28 days (4 weeks) / two difficulty jumps should have netted

260.3 coins +

I'm looking at the payout for that same 28 day window, and we only got paid out 215.7 coins.

Now the scary thing about this calculation is that 260.3 was assuming at last weeks difficulty, but the window was over the previous 2 difficulty rounds (from December 15th to January 15th) before switching.  Regardless, this is a major variance from expected to received...  I will get an online public Google document in a bit to show the payouts we received on a daily basis vs expected.

My guess is this is not a unique assessment.  I hope a few other people can hop on and verify this.  I know there was a lot of complaints about 7 hour block times, and considering this pool supposedly makes up 16% of the network the luck would be very very low to support multiple time spans like that in a row.  I'm a little upset that the variance is so huge over such a big period.

Also, I would argue, a week is NOT too short of a time span, considering the SIZE of the Ghash and the amount of blocks it finds during a 1 week period (~156 blocks) - variance in payouts should decrease based on the size of the pool.

best I have in an excel breakdown, problem is I need to look at the historic on what the difficulty was during the time I should have received the payouts.   I mean, if anyone else wants to calculate there own payout:

Go to a BTC calculator, and calculate how much you should make per day such as:
http://www.coinwarz.com/calculators/bitcoin-mining-calculator

Then, add up the most recent week on how mcuh you've been paid, and then compare that to the estimated profitability.

Its best to calculate this on the previous difficulty and only count the week before the difficulty jump.

All I know was with the 800 terahash average hashrate the client we support had, should have been recieving 9bitcions a day, his average payout was only 7.3 over the period of a week:

Expected = 64.4 based on difficulty and calculator
received = 51.52 bitcoin over same period.

A pool the size of ghash should not have THAT much variance over the period of a week...  Honestly, I would love a second opinion of someone else checking there own numbers over the last difficulty period....  When we went back to calculate, we realized we were being scammed / screwed by ghash and switched immediately.  I'm now getting expected payouts on time and accurately on antpool.

I think one of the reasons antpool does well is because it doesn't do small shares, it starts at 1024, meaning, old hardware that gives hashes, but probably can't even find a block now times out and doesn't contribute to antpool / rob the pool of shares.  Whatever they're doing it seems to work, and the PPLNS stats, everything looks legit.  Also, let me say, BTC Guild is awesome as well, but I just need the higher hashrate to stabilize the returns for these large investors.  Having a week of 80% is just too much money for them to lose.

Once we reach about 3-4 peta, in a deployment or two, we may solo mine - not sure yet. (block per day average)

put the numbers up, I'm sure there would be lots more interested if this is true.
I thought it was odd myself they hit an 11 hr stretch of no blocks,. Which is I think is very very rare for them with that much of a hashrate.