ASICMINER: Entering the Future of ASIC Mining by Inventing It

[binaryFate]

Regarding variance in general:

Since blocks are randomly discovered, the number of blocks discovered over a given period of time is Poisson distributed.
The expected number of blocks discovered by ASICMINER in a week is: 6*24*7*network_fraction.
Since in a Poisson distribution, the mean is equal to the variance, and the standard deviation is equal to the square root of the variance, we have a standard deviation of
sqrt(6*24*7*0.1), or approximately 10 blocks, assuming ASICMINER controls 10% of the network, or
sqrt(6*24*7*0.05), or approximately 7 blocks, assuming ASICMINER controls 5% of the network.

This means that the standard deviation is approximately 10% the expected number of discovered blocks if ASICMINER controls 10% of the nework, or 14% if ASICMINER controls 5% of the network.
If this is a lot or not is up for debate, but it is not insignificant.

Using the same math, we get a standard deviation of 53% (assuming 10% of network) for a six hour window and 37% for a 12 hour windows. That's quite a lot.

Thanks for the maths 
Not insignificant, but as I wrote, most of the weeks it would be hidden by the variation of the actual hash rate (glitches, problems, tests, etc).

Add on this that hardware sells also adds up to dividends in a variable way... it would be ridiculous to complain about a potential 14% variance, when the actual variance of dividends is much higher for different reasons.

[hcburger]

Not insignificant, but as I wrote, most of the weeks it would be hidden by the variation of the actual hash rate (glitches, problems, tests, etc).

Add on this that hardware sells also adds up to dividends in a variable way... it would be ridiculous to complain about a potential 14% variance, when the actual variance of dividends is much higher for different reasons.

Yes, over a week it is indeed not too important.
However, it would seem that it makes little sense to display 6 hour and 12 hour hash rate estimates on, e.g. http://www.asicminercharts.com/live/ . Even 24 hour estimates are quite unreliable: stddev of 26% assuming 10% network control, 37% assuming 5% control.

Or maybe the charts could be modified to include this standard deviation, so that the level of uncertainty can also be visualized?

[keystroke]

Didn't you see friedcat's post about exponential new equipment being manufactured in September and October?

Yes.  Which has nothing to do with why the existing 50 TH/s network hasn't been able to operate consistently at 50 TH/s. 

Yes, good point. It would be nice to get an explanation for this.

[Rival]

Not insignificant, but as I wrote, most of the weeks it would be hidden by the variation of the actual hash rate (glitches, problems, tests, etc).

Add on this that hardware sells also adds up to dividends in a variable way... it would be ridiculous to complain about a potential 14% variance, when the actual variance of dividends is much higher for different reasons.

Yes, over a week it is indeed not too important.
However, it would seem that it makes little sense to display 6 hour and 12 hour hash rate estimates on, e.g. http://www.asicminercharts.com/live/ . Even 24 hour estimates are quite unreliable: stddev of 26% assuming 10% network control, 37% assuming 5% control.

Or maybe the charts could be modified to include this standard deviation, so that the level of uncertainty can also be visualized?

I would think it would be trivial to include a standard deviation to the charts. A bar graph of 30Ths might benefit from a "+/- 2.43 Thz" disclaimer. If poll takers can do that, why can't we?

[DeathAndTaxes]

Not insignificant, but as I wrote, most of the weeks it would be hidden by the variation of the actual hash rate (glitches, problems, tests, etc).

Add on this that hardware sells also adds up to dividends in a variable way... it would be ridiculous to complain about a potential 14% variance, when the actual variance of dividends is much higher for different reasons.

Yes, over a week it is indeed not too important.
However, it would seem that it makes little sense to display 6 hour and 12 hour hash rate estimates on, e.g. http://www.asicminercharts.com/live/ . Even 24 hour estimates are quite unreliable: stddev of 26% assuming 10% network control, 37% assuming 5% control.

Or maybe the charts could be modified to include this standard deviation, so that the level of uncertainty can also be visualized?

I would think it would be trivial to include a standard deviation to the charts. A bar graph of 30Ths might benefit from a "+/- 2.43 Thz" disclaimer. If poll takers can do that, why can't we?

It would be trivial but then again you don't see a poll which says "52% +/- 45% approve (with 95% confidence interval)".  That is essentially the equivalent of the 6 hour stat.

[Jayjay04]

Why the officcial site '' www.asicminer.co '' isn't just up to date.
Just put the live hashrate, stats on block found and revenues there dammit !

More transparency is getting essential please !!

[eastpk]

Why the officcial site '' www.asicminer.co '' isn't just up to date.
Just put the live hashrate, stats on block found and revenues there dammit !

More transparency is getting essential please !!

lol that isnt the official site....

[hcburger]

It would be trivial but then again you don't see a poll which says "52% +/- 45% approve (with 95% confidence interval)".  That is essentially the equivalent of the 6 hour stat.

Agreed: The 6 and 12 hour estimates are essentially useless.
But maybe we could have the estimated stddevs for the longer-term (maybe 24 hour and above) hash-rate estimates?

[Jayjay04]

Why the officcial site '' www.asicminer.co '' isn't just up to date.
Just put the live hashrate, stats on block found and revenues there dammit !

More transparency is getting essential please !!

lol that isnt the official site....

Well is there one ?!?!

[eastpk]

Why the officcial site '' www.asicminer.co '' isn't just up to date.
Just put the live hashrate, stats on block found and revenues there dammit !

More transparency is getting essential please !!

lol that isnt the official site....

Well is there one ?!?!

Yup, this is the official website.....

https://bitcointalk.org/index.php?action=profile;u=49840;sa=showPosts

[ianp]

Not insignificant, but as I wrote, most of the weeks it would be hidden by the variation of the actual hash rate (glitches, problems, tests, etc).

Add on this that hardware sells also adds up to dividends in a variable way... it would be ridiculous to complain about a potential 14% variance, when the actual variance of dividends is much higher for different reasons.

Yes, over a week it is indeed not too important.
However, it would seem that it makes little sense to display 6 hour and 12 hour hash rate estimates on, e.g. http://www.asicminercharts.com/live/ . Even 24 hour estimates are quite unreliable: stddev of 26% assuming 10% network control, 37% assuming 5% control.

Or maybe the charts could be modified to include this standard deviation, so that the level of uncertainty can also be visualized?

I would think it would be trivial to include a standard deviation to the charts. A bar graph of 30Ths might benefit from a "+/- 2.43 Thz" disclaimer. If poll takers can do that, why can't we?

It would be trivial but then again you don't see a poll which says "52% +/- 45% approve (with 95% confidence interval)".  That is essentially the equivalent of the 6 hour stat.

I just want to add that I believe an accurate prediction of our current hashrate is about 63.5 th/s.

I do not have any reason to believe that they have sold off any hashing power, nor do I have any reason to believe they have increased power over the last four weeks.

http://www.dpcapital.net/blockchain/?hours=672

[runeks]

Not insignificant, but as I wrote, most of the weeks it would be hidden by the variation of the actual hash rate (glitches, problems, tests, etc).

Add on this that hardware sells also adds up to dividends in a variable way... it would be ridiculous to complain about a potential 14% variance, when the actual variance of dividends is much higher for different reasons.

Yes, over a week it is indeed not too important.
However, it would seem that it makes little sense to display 6 hour and 12 hour hash rate estimates on, e.g. http://www.asicminercharts.com/live/ . Even 24 hour estimates are quite unreliable: stddev of 26% assuming 10% network control, 37% assuming 5% control.

Or maybe the charts could be modified to include this standard deviation, so that the level of uncertainty can also be visualized?

I would think it would be trivial to include a standard deviation to the charts. A bar graph of 30Ths might benefit from a "+/- 2.43 Thz" disclaimer. If poll takers can do that, why can't we?

It would be trivial but then again you don't see a poll which says "52% +/- 45% approve (with 95% confidence interval)".  That is essentially the equivalent of the 6 hour stat.

I just want to add that I believe an accurate prediction of our current hashrate is about 63.5 th/s.

I do not have any reason to believe that they have sold off any hashing power, nor do I have any reason to believe they have increased power over the last four weeks.

http://www.dpcapital.net/blockchain/?hours=672

How do you calculate that? I don't think that's accurate. My calculations don't indicate 63.5 TH, not even on the highest spikes of the 24-hour average (which is almost certainly an inaccurate estimate).

[ianp]

Not insignificant, but as I wrote, most of the weeks it would be hidden by the variation of the actual hash rate (glitches, problems, tests, etc).

Add on this that hardware sells also adds up to dividends in a variable way... it would be ridiculous to complain about a potential 14% variance, when the actual variance of dividends is much higher for different reasons.

Yes, over a week it is indeed not too important.
However, it would seem that it makes little sense to display 6 hour and 12 hour hash rate estimates on, e.g. http://www.asicminercharts.com/live/ . Even 24 hour estimates are quite unreliable: stddev of 26% assuming 10% network control, 37% assuming 5% control.

Or maybe the charts could be modified to include this standard deviation, so that the level of uncertainty can also be visualized?

I would think it would be trivial to include a standard deviation to the charts. A bar graph of 30Ths might benefit from a "+/- 2.43 Thz" disclaimer. If poll takers can do that, why can't we?

It would be trivial but then again you don't see a poll which says "52% +/- 45% approve (with 95% confidence interval)".  That is essentially the equivalent of the 6 hour stat.

I just want to add that I believe an accurate prediction of our current hashrate is about 63.5 th/s.

I do not have any reason to believe that they have sold off any hashing power, nor do I have any reason to believe they have increased power over the last four weeks.

http://www.dpcapital.net/blockchain/?hours=672

How do you calculate that? I don't think that's accurate. My calculations don't indicate 63.5 TH, not even on the highest spikes of the 24-hour average (which is almost certainly an inaccurate estimate).

Well -- it is a bit skewed. It only calculates based on the current difficulty, not the past difficulty.

That is to say -- that if difficulty were to rise ten fold right now, the historical hashrate would rise as well because it's basing everything off of current difficulty.

The link I posted produces the average hashrate of the previous 28 days.

[runeks]

Not insignificant, but as I wrote, most of the weeks it would be hidden by the variation of the actual hash rate (glitches, problems, tests, etc).

Add on this that hardware sells also adds up to dividends in a variable way... it would be ridiculous to complain about a potential 14% variance, when the actual variance of dividends is much higher for different reasons.

Yes, over a week it is indeed not too important.
However, it would seem that it makes little sense to display 6 hour and 12 hour hash rate estimates on, e.g. http://www.asicminercharts.com/live/ . Even 24 hour estimates are quite unreliable: stddev of 26% assuming 10% network control, 37% assuming 5% control.

Or maybe the charts could be modified to include this standard deviation, so that the level of uncertainty can also be visualized?

I would think it would be trivial to include a standard deviation to the charts. A bar graph of 30Ths might benefit from a "+/- 2.43 Thz" disclaimer. If poll takers can do that, why can't we?

It would be trivial but then again you don't see a poll which says "52% +/- 45% approve (with 95% confidence interval)".  That is essentially the equivalent of the 6 hour stat.

I just want to add that I believe an accurate prediction of our current hashrate is about 63.5 th/s.

I do not have any reason to believe that they have sold off any hashing power, nor do I have any reason to believe they have increased power over the last four weeks.

http://www.dpcapital.net/blockchain/?hours=672

How do you calculate that? I don't think that's accurate. My calculations don't indicate 63.5 TH, not even on the highest spikes of the 24-hour average (which is almost certainly an inaccurate estimate).

Well -- it is a bit skewed. It only calculates based on the current difficulty, not the past difficulty.

That is to say -- that if difficulty were to rise ten fold right now, the historical hashrate would rise as well because it's basing everything off of current difficulty.

The link I posted produces the average hashrate of the previous 28 days.

Well that's the problem then. It's not just a bit skewed, it's very skewed. The difficulty 28 days ago was 31M - less than half of what is now. So you're over-estimating the hashrate for the first ~4 days by over 110%, and the following 10 days by over 75%.

[ianp]

Not insignificant, but as I wrote, most of the weeks it would be hidden by the variation of the actual hash rate (glitches, problems, tests, etc).

Add on this that hardware sells also adds up to dividends in a variable way... it would be ridiculous to complain about a potential 14% variance, when the actual variance of dividends is much higher for different reasons.

Yes, over a week it is indeed not too important.
However, it would seem that it makes little sense to display 6 hour and 12 hour hash rate estimates on, e.g. http://www.asicminercharts.com/live/ . Even 24 hour estimates are quite unreliable: stddev of 26% assuming 10% network control, 37% assuming 5% control.

Or maybe the charts could be modified to include this standard deviation, so that the level of uncertainty can also be visualized?

I would think it would be trivial to include a standard deviation to the charts. A bar graph of 30Ths might benefit from a "+/- 2.43 Thz" disclaimer. If poll takers can do that, why can't we?

It would be trivial but then again you don't see a poll which says "52% +/- 45% approve (with 95% confidence interval)".  That is essentially the equivalent of the 6 hour stat.

I just want to add that I believe an accurate prediction of our current hashrate is about 63.5 th/s.

I do not have any reason to believe that they have sold off any hashing power, nor do I have any reason to believe they have increased power over the last four weeks.

http://www.dpcapital.net/blockchain/?hours=672

How do you calculate that? I don't think that's accurate. My calculations don't indicate 63.5 TH, not even on the highest spikes of the 24-hour average (which is almost certainly an inaccurate estimate).

Well -- it is a bit skewed. It only calculates based on the current difficulty, not the past difficulty.

That is to say -- that if difficulty were to rise ten fold right now, the historical hashrate would rise as well because it's basing everything off of current difficulty.

The link I posted produces the average hashrate of the previous 28 days.

Well that's the problem then. It's not just a bit skewed, it's very skewed. The difficulty 28 days ago was 31M - less than half of what is now. So you're over-estimating the hashrate for the first ~4 days by over 110%, and the following 10 days by over 75%.

I store the difficulty at the time the block was mined -- I don't calculate it though. One of these days I'll get around to that -- I've just been too busy lately

Edit -- whoops, I do not store difficulty. One of these days I'll add difficulty and go update everything

[conv3rsion]

One thing that everyone should realize that its in Friedcat's interest for the share price to decrease significantly right before rolling out a 2nd generation of hardware or products. Why? Because he is the only person that knows exactly what the company is planning to do AND while he cannot sell his existing shares without those actions being known (via the blockchain), he can buy cheaply from others much more anonymously.

In other words, if you're Friedcat and you know the plans for the next 6 months or so, you can increase your company ownership much easier by first withholding dividends and / or newly purchased hardware for some period of time, then purchasing additional shares (on whatever market) at a reduced price. This is insider trading, but since this is a virtual market, there is no protection against it.

I'm not accusing Friedcat of anything, I'm simply explaining a consequence of the current market. I'm here for the dividends, the share price has lost 50% of its value, and it would take over 18 months for reduced dividends to make up the difference, but at the company's current valuation, I'm much more of a buyer than a seller.

[Rebelution]

One thing that everyone should realize that its in Friedcat's interest for the share price to decrease significantly right before rolling out a 2nd generation of hardware or products. Why? Because he is the only person that knows exactly what the company is planning to do AND while he cannot sell his existing shares without those actions being known (via the blockchain), he can buy cheaply from others much more anonymously.

In other words, if you're Friedcat and you know the plans for the next 6 months or so, you can increase your company ownership much easier by first withholding dividends and / or newly purchased hardware for some period of time, then purchasing additional shares (on whatever market) at a reduced price. This is insider trading, but since this is a virtual market, there is no protection against it.

I'm not accusing Friedcat of anything, I'm simply explaining a consequence of the current market. I'm here for the dividends, the share price has lost 50% of its value, and it would take over 18 months for reduced dividends to make up the difference, but at the company's current valuation, I'm much more of a buyer than a seller.

True, but this is just one of the many, many ways that Friedcat can steal from his shareholders considering its a completely unregulated market, there are no verifiable financials, no verifiable inventories, no way to confirm the miners aren't mining under other addresses, etc.  All of it is based on trust.  If you don't trust Friedcat to not screw us over, then you should not "invest" (gamble really).

[bingjiw]

http://trilema.com/2013/the-analysis-of-a-failure-asicminer/

[GoldenWings91]

http://trilema.com/2013/the-analysis-of-a-failure-asicminer/

Mircea Popescu runs MPEX which is an exchange that ASICMiner refused to be listed on. He's been attacking AM ever since so anything he says should be taken with scepticism.

[murfshake]

http://trilema.com/2013/the-analysis-of-a-failure-asicminer/

Pretty bad article actually.