[4+ EH] Slush Pool (slushpool.com); Overt AsicBoost; World First Mining Pool

[binja9]

Thanks for that - good explanation with solid math.
My following query would be - if 2 blocks are found "at the same time" - how is the block owner\solver determined?
If its exact time as suggested earlier then why would 2 miners think they had solved it ?

If you look at a blockchain record, you will see that there are two time stamps: the time the block was found, as recorded by the finder, and the time it was reported to the blockchain, which is the one that counts (but see the next paragraph).  When a block is solved by two miners almost simultaneously, it can (and does) happen that the first finder loses out through being too slow in reporting their success, which may be caused by their own internal issues or by a delay in the internet.  It appears to happen to most pools sooner or later.  Since everybody will continue to work on a block until the new find has been logged and propagated across the network, occasional duplicates are inevitable.

There is a rare complication: if the miner who loses the race happens to solve and report a second block before the first result has been propagated, his blocks will form the new chain by having the longer fork, and the poor sod who got there first loses out.  This could happen by accident; if intentional - which is theoretically possible - it's called "selfish mining".

Brilliant - thanks for the clear explanation

[1713954889]

1713954889

[1713954889]

1713954889

[1713954889]

1713954889

[1713954889]

1713954889

[1713954889]

1713954889

[gourmet]

My understanding is that when two pools find the 'same' block, they are actually finding different hashes that meet the difficulty requirement (please correct me if I am wrong). In this case, given that the network difficulty makes this so that the expected block length is ten minutes, and invalid blocks are not that common (IIRC, we have seen two in the last month or so), I am quite surprised that there should be such a 'collision' on a two-second block - surely the odds of this happening would be vanishingly small, unless there is something intrinsic that makes the difficulty on certain blocks much lower than it should be.

Not only different hashes; each pool assembles its own block from the available transactions in the transaction pool according to its rules. So the "same" block can potentially differ between the two pools in the sense it contains different transactions.

The last idea is quite interesting. Can someone have enough will and power to check it? :-)
[edit] I'd sure not say vanishingly, but the chance is of course smaller for two pools at once than it is for one pool.

I would have thought that the probability of two people finding a sub 2-second block, at the same time would be the probability of one finding it, squared.
Assuming the block difficulty is correct, at the moment, we are on block 288579. 
I don't know how many 2 -second blocks there have been so to make the numbers easier, lets guess at around 300, which makes the odds of a given block being less than 2 seconds around 1 in 1000, or 0.001.
From the statistics on blockchain.info, it looks like there are around 2.5 orphaned blocks per day. The total number of blocks each day is tuned to be 144 (one every ten minutes), so dividing this by 2.5 gives around 60, or a 1 in 60 chance of there being such a collision on a given block.
if my maths is right, this means there is a a 1 in 1000 chance of each participant hitting a 2 second block, so multiply these together to get 1 in 1000000, and there is a 1 in 60 chance of there being a collision in this way, so the odds of this happening would be roughly 1 in sixty million. To put this in perspective, you have a 1 in 600000 chance of being struck by lightning, and this would appear to be some 100 times less likely!

Now, further up there, I made an assumption - that the block difficulty is 'correct'. As I understand it, the block difficulty essentially defines how many leading zeroes must be in a block's hash for it to be considered 'solved'. For example, the last block (found by us!) had a hash of 0000000000000000d69d39dbe4e025909fc30b4be52abfaccfec9315c497ac6c, the digits after the leading zeroes are irrelevant (which is what leads to block collisions in the first place, as two hashes can be found with the same number of leading zeroes).
The block itself is composed of the newly found hash, the hash from the previous block, the block number, difficulty and time, the pool's address, and all the transactions in that block (this is a simplified description), so my question is this - since the odds of two pools finding a 2 second block at the same time is so low (and I stick by 'vanishingly small' as a good description), is it possible that some blocks have more 'solutions' than are statistically expected - in other words, for some blocks, are there more solutions than expected with that number of leading zeroes?
For example, the hash above has 16 leading zeroes, in binary this is 128 leading zeroes, so if my understanding is correct, the odds of getting this out of any random hash would be 2^128.
Any thoughts?

By extremely short blocks, you can't account for mean orphan probability. By a two-second block, probability of orphan race is extremely high, near to 100% IMHO, as the time differences between the good and the orphaned block normally are of the same size as is the block duration by a 2-second block. So I think you must leave out your "60" multiplier.

[nottm28]

Enter Organofcorti please to clear this up...

[EdB666]

By extremely short blocks, you can't account for mean orphan probability. By a two-second block, probability of orphan race is extremely high, near to 100% IMHO, as the time differences between the good and the orphaned block normally are of the same size as is the block duration by a 2-second block. So I think you must leave out your "60" multiplier.

Good point - the odds still work out at around 1 in 1000000, but as Terry Pratchett once wrote, million-to-one chances happen nine times out of ten.

I still wonder if the contents of some blocks can make the difficulty vary significantly from the assigned difficulty though. Perhaps there is a weakness in (double) SHA256 where hash collisions are more likely with certain values?
Maybe OrganOfCorti can help us out on this one? Any crypto experts around? Where's Bruce Sterling when you need him...

[Trefdraeth]

By extremely short blocks, you can't account for mean orphan probability. By a two-second block, probability of orphan race is extremely high, near to 100% IMHO, as the time differences between the good and the orphaned block normally are of the same size as is the block duration by a 2-second block. So I think you must leave out your "60" multiplier.

Good point - the odds still work out at around 1 in 1000000, but as Terry Pratchett once wrote, million-to-one chances happen nine times out of ten.

I still wonder if the contents of some blocks can make the difficulty vary significantly from the assigned difficulty though. Perhaps there is a weakness in (double) SHA256 where hash collisions are more likely with certain values?
Maybe OrganOfCorti can help us out on this one? Any crypto experts around? Where's Bruce Sterling when you need him...

OOC is the man, he will elucidate....

[organofcorti]

By extremely short blocks, you can't account for mean orphan probability. By a two-second block, probability of orphan race is extremely high, near to 100% IMHO, as the time differences between the good and the orphaned block normally are of the same size as is the block duration by a 2-second block. So I think you must leave out your "60" multiplier.

Good point - the odds still work out at around 1 in 1000000, but as Terry Pratchett once wrote, million-to-one chances happen nine times out of ten.

I still wonder if the contents of some blocks can make the difficulty vary significantly from the assigned difficulty though. Perhaps there is a weakness in (double) SHA256 where hash collisions are more likely with certain values?
Maybe OrganOfCorti can help us out on this one? Any crypto experts around? Where's Bruce Sterling when you need him...

OOC is the man, he will elucidate....

Thanks for the vote of confidence guys, but I'm having trouble following the question and answers. Can someone boil post a precis of the question?

[nottm28]

Thanks for the vote of confidence guys, but I'm having trouble following the question and answers. Can someone boil post a precis of the question?

Holy crap - a 35 second block - that has to be some kind of record!

naw, i saw a 20 second block once.

I remember a 14 second one. And it's not been too long ago.
And a 4 second one last year.

Shortest I ever seen was block 239570 found by slush on 2013-06-03 in just 2 seconds. Unfortunately it was invalid for us...

I guess the question is "are the odds of two pools finding a 2 second block at the same time 'vanishingly small' ?"...

[gourmet]

By extremely short blocks, you can't account for mean orphan probability. By a two-second block, probability of orphan race is extremely high, near to 100% IMHO, as the time differences between the good and the orphaned block normally are of the same size as is the block duration by a 2-second block. So I think you must leave out your "60" multiplier.

Good point - the odds still work out at around 1 in 1000000, but as Terry Pratchett once wrote, million-to-one chances happen nine times out of ten.

I still wonder if the contents of some blocks can make the difficulty vary significantly from the assigned difficulty though. Perhaps there is a weakness in (double) SHA256 where hash collisions are more likely with certain values?
Maybe OrganOfCorti can help us out on this one? Any crypto experts around? Where's Bruce Sterling when you need him...

OOC is the man, he will elucidate....

Thanks for the vote of confidence guys, but I'm having trouble following the question and answers. Can someone boil post a precis of the question?

EdB666 tried to calculate the chance of invalid block (orphan race) by a two-second block.
He started with the chance of two pools finding a two-second block at once, being the chance of one pool finding a two-second block (1/1000) squared, i.e. 1/1000000. For the purpose of finding the chance to one being an orphan, he multiplied this number by the chance for any block to be orphan, which he found to be ~1/60.
I disputed that the last step couldn't be applied for extremely short blocks, as there the chance for orphan is much greater, near to 1, as the block duration approaches the usual time difference between orphan and regular blocks.

[EdB666]

Thanks for the vote of confidence guys, but I'm having trouble following the question and answers. Can someone boil post a precis of the question?

In short:

I noted that we had a 35 second block and suggested that this might be some sort of record - another poster mentioned that we once had a 2 second block, but it was invalid.
This got me thinking that this situation must be extremely unlikely to happen, since the odds of a block lasting 2 seconds are quite short, and the odds of two pools finding a two second block at the same time would be even shorter. A quick estimate of the odds, if my understanding of the maths is right, is that the probability of a block being 2 seconds or less is around 0.001, so two at the same time would be around 0.000001, or 1 in 1 million (as pointed out by another poster, the odds of there being a 'race' between these two would be close to one because the block time is so short, rather than 1 in 60, which seems to be the approximate rate of orphaned blocks in the network)
This led me to wonder whether there is something about the (double) SHA256 algorithm that might lead to hash collisions being more likely for certain blocks, so that these blocks rather than being expected to last on average 10 minutes, would have much lower than expected difficulty. This might shorten the odds of there being an orphaned 2 second block. I don't know how possible this actually is, given that the block contains the address of the pool doing the hashing, so that the actual block being computed by two pools would differ.
So, I suppose what I am asking is whether my estimates seem fair, or if I have got the maths all wrong, and whether it is plausible that the hashing algorithm could lead to collisions in this way, causing the occasional block to be much easier to solve, for certain pools, or the network as a whole. I appreciate that cryptanalysis is far from a simple topic, so the second question is really more rhetorical.

[organofcorti]

Before I get started, was the 2 second block invalid because it lost an orphan race or because it was invalid for some other reason? And was it two seconds after the last solved block for the network or for the pool?

[gourmet]

Holy crap - a 35 second block - that has to be some kind of record!

naw, i saw a 20 second block once.

I remember a 14 second one. And it's not been too long ago.
And a 4 second one last year.

Shortest I ever seen was block 239570 found by slush on 2013-06-03 in just 2 seconds. Unfortunately it was invalid for us...

Blockchain.info says it's main chain...?
Acording to the link, it's ours, and there's no other block at that height.

[wdalessi]

I am curious what effect it could or would have to start mining in a pool after it shows a luck rating of below 80% long term (Monthly), and 60% short term (1 week).

I know this method is used extensively in stock trades. They call it the RSI or Relative Strength Index.

http://www.investopedia.com/terms/r/rsi.asp

Time and price movements are not random, as say hashing is. Luck can be relative, just as Relative Strength is.

Though it is possible to have permanently bad or low luck, it can not happen in nature over extremely long periods of times.

Yes you can flip "Heads" on a coin 100 times in a row, but if starting at toss #1 the laws of probability will prove out in the long run. Aberrations are just "Variance" of the longer term averages.

This can be used to a miners advantage.

If you only start mining on a pool when it has short term (1 to 2 Weeks) luck below  say 60 and switch to Pay Per Share when the pool luck goes above say 120, how would this affect the long term earnings of a miner?

Because just as Bad Luck, never lasts, so to Good Luck can also not last.

Unfortunately my college math skills have long been forgotten.

[ZephramC]

I am curious what effect it could or would have to start mining in a pool after it shows a luck rating of below 80% long term (Monthly), and 60% short term (1 week).

I know this method is used extensively in stock trades. They call it the RSI or Relative Strength Index.

http://www.investopedia.com/terms/r/rsi.asp

Time and price movements are not random, as say hashing is. Luck can be relative, just as Relative Strength is.

Though it is possible to have permanently bad or low luck, it can not happen in nature over extremely long periods of times.

Yes you can flip "Heads" on a coin 100 times in a row, but if starting at toss #1 the laws of probability will prove out in the long run. Aberrations are just "Variance" of the longer term averages.

This can be used to a miners advantage.

If you only start mining on a pool when it has short term (1 to 2 Weeks) luck below  say 60 and switch to Pay Per Share when the pool luck goes above say 120, how would this affect the long term earnings of a miner?

Because just as Bad Luck, never lasts, so to Good Luck can also not last.

Unfortunately my college math skills have long been forgotten.

This is called Gambler's fallacy. In reality for truly random process "bad luck" does not imply "good luck in the future". Even if you toss the coin and wait for 10 heads in a row then the next toss would be equally likely head as tail. There is nothing like "after 10 red in a row in roulette there must be more black than red". The ratio should approach probability, but the difference between "good" and "bad" does not go to zero.

[nottm28]

Before I get started, was the 2 second block invalid because it lost an orphan race or because it was invalid for some other reason? And was it two seconds after the last solved block for the network or for the pool?

Alas I only have these details from the slush api. My simple java app writes the blocks to a file over time.

Here is an extract of the log files:

block=start_time,duration,reward,hashes,luck,difficulty,confirmations
239597=2013-06-04 04:07:32,1:44:35,0.00615060,12426,1.09,12153411,
239581=2013-06-04 02:22:57,1:06:19,0.00568213,12426,1.09,12153411,
239579=2013-06-04 01:16:38,1:19:04,0.00532830,12426,1.09,12153411,
239570=2013-06-03 23:57:34,0:00:02,0.00457616,12426,1.09,12153411,invalid
239557=2013-06-03 22:39:26,1:02:08,0.00631690,12581,1.19,12153411,
239550=2013-06-03 21:37:18,1:47:07,0.00620117,12382,1.24,12153411,
239530=2013-06-03 19:50:11,0:23:53,0.00621462,12406,1.29,12153411,

As you can see, there is no indication from the api as to whether the block is invalid because it's orphaned or some other cause. But the logs show this was not a consecutive block for slush. The previous slush block before the 239570 2 second invalid was 239557...

[edit] btw check out the average time for a found block in them days - max 2 hours - and the difficulty only 12 million, and the entire pool hashrate of ~ 12 TH/s - them were the days...

[MrTeal]

I am curious what effect it could or would have to start mining in a pool after it shows a luck rating of below 80% long term (Monthly), and 60% short term (1 week).

I know this method is used extensively in stock trades. They call it the RSI or Relative Strength Index.

http://www.investopedia.com/terms/r/rsi.asp

Time and price movements are not random, as say hashing is. Luck can be relative, just as Relative Strength is.

Though it is possible to have permanently bad or low luck, it can not happen in nature over extremely long periods of times.

Yes you can flip "Heads" on a coin 100 times in a row, but if starting at toss #1 the laws of probability will prove out in the long run. Aberrations are just "Variance" of the longer term averages.

This can be used to a miners advantage.

If you only start mining on a pool when it has short term (1 to 2 Weeks) luck below  say 60 and switch to Pay Per Share when the pool luck goes above say 120, how would this affect the long term earnings of a miner?

Because just as Bad Luck, never lasts, so to Good Luck can also not last.

Unfortunately my college math skills have long been forgotten.

This is called Gambler's fallacy. In reality for truly random process "bad luck" does not imply "good luck in the future". Even if you toss the coin and wait for 10 heads in a row then the next toss would be equally likely head as tail. There is nothing like "after 10 red in a row in roulette there must be more black than red". The ratio should approach probability, but the difference between "good" and "bad" does not go to zero.

Just to add to that, the reason why the luck will tend to move back to 100% is that the initial variance will get swamped by the new numbers. If you flip a coin 100 times and hit heads 20 times, each of the flips after that still has a 50% chance of coming up heads. So, if you flip it another 900 times you'd expect to hit 450 heads and 450 tails. When you look at the total of 470 heads and 530 tails it might seem like your luck has improved from 20/80, but really that's just where you'd expect it to be for those 900 flips that haven't happened yet.

[MLWhuffo]

I'm having a problem with Vardiff. I'm running a pipelined miner; very perky but it takes a while to sync back up when the pool forces a change. The worst offender is Vardiff; it tries to force me to the appropriate difficulty level (which is fine with me), but when it does that my rig has to reset and resync. By the time it's ready to go, the pool has already decided to change the difficulty to a lower level. Rinse, repeat - what a nuisance. When it finally ends up at difficulty 1, it is allowed the time to get going - and almost immediately Vardiff changes the difficulty level again.

If there was a way to lock the difficulty level in Cgminer, I'd lock it at the appropriate value. But there's not, and it's no longer possible to set a difficulty level at the pool. Slush could fix this by allowing us to set a minimum difficulty level; Vardiff could determine the right one then it could be locked in. Or he could put a longer delay between difficulty changes. As it is, it's very difficult to keep my miner running; it spends far too much time resynchronizing and not enough time hashing.

I can live with the "new block" resets, but the Vardiff stuff is ridiculous.

[RFF2]

I'm using GUIminer v2012-12-03. It ticks away at an embarassingly slow rate, but it's mine own (and I'm getting a better graphics card).

Nothing has been accepted or marked stale for over a month. It used to work. The console shows:

Code:

2014-03-01 14:08:38: Listener for "Graphics": stratum.bitcoin.cz:8332 01/03/2014 14:08:38, checking for stratum...
2014-03-01 14:08:38: Listener for "Graphics": stratum.bitcoin.cz:8332 01/03/2014 14:08:38, started OpenCL miner on platform 0, device 0 (Caicos)
2014-03-01 14:08:38: Listener for "Graphics": stratum.bitcoin.cz:8332 01/03/2014 14:08:38, diverted to stratum on stratum.bitcoin.cz:3333
2014-03-01 14:08:48: Listener for "Graphics": stratum.bitcoin.cz:8332 01/03/2014 14:08:48, Failed to subscribe
2014-03-01 14:08:50: Listener for "Graphics": stratum.bitcoin.cz:8332 01/03/2014 14:08:50, IO errors - 1, tolerance 2

I've done a lot of searching and come to the conclusion that nowhere is the appropriate server and port and an effective date listed.  I've tried stratum.bitcoin.cz and api.bitcoin.cz and ports 3333 and 3332 but the upshot is still the same. Am I using the wrong settings?

[FalconFly]

Straight from the slush pool Main page https://mining.bitcoin.cz/ (sounds like this is the pool you use) :

3. Run miner

Run your own GPU/FPGA/ASIC miner with the worker credentials you gave earlier, and connect it to following URL:
http://api.bitcoin.cz:8332
(Main pool URL)

or
stratum.bitcoin.cz:3333
(If you have Stratum-compatible miner)

I have mine set to stratum.bitcoin.cz:3333 on slush, works like a champ.

PS.
Don't run GPUs on Bitcoin, they're slow as hell, produce no reasonable output (will likely take months just to reach slush minimum pool payout of 0.01BTC) and consume tons of energy which runs up the electrical bill to shocking numbers within short times.
You'd be better off with one of the obsolete Block Erupters (at least they only draw 2W of power)...

[Sir Alan]

... I've done a lot of searching and come to the conclusion that nowhere is the appropriate server and port and an effective date listed.  I've tried stratum.bitcoin.cz and api.bitcoin.cz and ports 3333 and 3332 but the upshot is still the same. Am I using the wrong settings?

Your settings appear to be out of date.

Select the server "slush's pool(IPv6/alt)"  (this will automatically point to the correct url and port)
Make sure you have the correct user name and password in the appropriate boxes
Check that the selected device is correct - i.e. the GPU, not the CPU
Use these extra flags: -v -w 128 -f 10

If it still doesn't connect or mine, there could be a firewall issue.

While not wishing to drive anyone away from this pool, FalconFly is right about the poor return on effort of using a GPU for mining Bitcoin.  You might of course consider this worthwhile as a hobby, but you can earn more with your GPU by Scrypt mining through ScryptGuild.

[J_Dubbs]

Anyone else have low rewards/shares the last few rounds? Looks like my rewards were cut in half but estimated reward is currently on track, confused...