Topic: [1500 TH] p2pool: Decentralized, DoS-resistant, Hop-Proof pool - page 688. (Read 2591920 times)

this ^^^
blockchain.info is notoriously inaccurate - even tho the picture is pretty
plus http://blockorigin.pfoe.be/top.php is actually accurate over 2016 blocks, was put together by kinlo

The force is strong with this one.

Well... not exactly.

A duration time T can *never* be normally distributed, since T can't be zero nor less than zero.

We can better say that the number of trials it takes to solve a block is geometrically distributed:
http://en.wikipedia.org/wiki/Geometric_distribution.

The amount of time required by the solution is the product of the amount of time required by each trial (which in ideal conditions is the same for every given mining device) times the number of trials: so it's also geometrically distributed. We call this geometric distribution "the distribution of the population".

However we can take a different approach: we will run a mining device and write down how much it will take to solve the first block, than the second block, than the third (always with the same difficulty, of course) and so on N times. At the end we will get a set of N values (N durations), but for now we don't have any values, we have a set of... N random variables with the same geometrical distribution of the population! This set of random variables is called "a sample".

Let's sum these random variables, and then divide the sum by their number, N. We get another random variable: the "sample mean". We don't have any number yet, just variables.

Being the "sample mean" a sum of random variables (identically distributed), we can apply to it the central limit theorem: http://en.wikipedia.org/wiki/Central_limit_theorem.

So a "sample mean" (regardless of its population) is always asymptotically bounded by a normal distribution (in practice, and in the worst case, 50 observations are more than enough).

The "sample mean" expected value is the same of the population, and its variance is the population variance divided by N.

I'm sure you know it all, but maybe someone was confused.

HTH.

Thats is less than 50% of deepbit's actual found blocks, blockinfo is just not listing all deepbit under deepbit.

everyone else gets orphans too.

The problem is if you split your rig and half mined p2pool and half mined on a normal PPS pool, the p2pool half would mine significantly less over time.

some days p2pool is awesome and some days it's bad that's normal variance. The problem is long term it doesn't catch up.

Did you guys see the current pool stats on bitcoinwatch.com?



p2pool third, deepbit seventh and almost out of the graph! :-D

Ente

[yes, this is found blocks, not "real" hashrate]

Flaw in pool luck calculations?
"The list below includes all known p2pool blocks. Orphaned blocks appear with strikethrough. Note: The list below may not include all orphaned blocks due to limitations in how the bitcoin client reports them."
See this:
http://blockchain.info/blocks/P2Pool
We have orphaned block 183088 that is not shown on http://p2pool.info/ site.
As s1 said b4 all good and orphaned blocks should be counted to pool luck. If we not see all of them there maybe they are not count? And thats why our long term pool luck looks bad in statistics?

You must have been a business major in college.  Heisenberg's Uncertainty Principle does not apply to normal everyday situations, only to quantum processes at the subatomic level.

Talking about "luck" or looking at block solution times or commenting about them on message boards has no effect on future outcomes.  You're simply engaging in "magical thinking".

90% of the problems humanity has is because someone believes in something that may or may not have anything to really do with the problem at hand.

Quantum physics says that obresvation of experiment can change it resoult.
If we will talk much about L it will fly away, like it always do

"Luck" is a psychological/emotional term with no objective basis in probability.

The amount of time it takes to solve a block is bounded by a normal distribution.  If a block (or series of blocks) is solved in less time than the mean, then people say it's "good luck".  If a series of blocks are solved in more time than the mean then they say it's "bad luck".

In either case the block solution times are bounded by a normal distribution centered about a mean time.

well, technically i don't see why it doesn't. my understanding is it's all down to probability and random numbers, which, in the end, is "luck" if i've ever heard it.

I guess the biggest problem is that people believe that luck actually has something to do with it ...

Yea, talk more bout l*** an we will back to 20hrs blocks... PPL! CMON!

When 90 day luck goes over 95% then I will accept that there is not a problem, until then I will still complain.

Telling people things like "This morning is a lesson in how pool variability can cut both ways." is irrelevant until that 90 day luck goes over 95% (IMO).

Maybe yesterday morning.  Only 2 of those orphaned blocks turned out to be legit.

This morning there were actually 4 solved blocks, only 1 of which was orphaned.

You did notice that 2 of those 3 were orphaned ? :/

This morning is a lesson in how pool variability can cut both ways.  3 blocks in a row solved in as many hours.  Very lucky. 

that... was brilliant.

http://www.youtube.com/watch?v=lldrizLu_d8

Great video.