Pages:
Author

Topic: Eligius: 0% Fee BTC, 105% PPS NMC, No registration, CPPSRB - page 86. (Read 1061485 times)

donator
Activity: 2058
Merit: 1007
Poor impulse control.
Makes some sense.  When I get stats v2 up going I'll be sure to include share counts for orphaned blocks on the API side for this purpose.

Great! Post here when you do? In the meantime I'm using an estimate based on the ratio of block durations.

On the front end side, though, I definitely think the current method is more towards what someone wants to see as a "luck" stat.

For most miners, that's true. However on the odd occasion we have "Vladimir vs BTC Guild" style kerfuffles, you can show someone immediately how likely luck is over any given number of blocks. In fact having the current 'luck' side-by-side with the CDF might not be a bad idea?
legendary
Activity: 1223
Merit: 1006
Makes some sense.  When I get stats v2 up going I'll be sure to include share counts for orphaned blocks on the API side for this purpose.

On the front end side, though, I definitely think the current method is more towards what someone wants to see as a "luck" stat.
donator
Activity: 2058
Merit: 1007
Poor impulse control.
I've always used Luck as a representation of actual earnings vs expected earnings, because that is what most miners are concerned about, and it is easy for them to understand.  If luck is 200%, they made twice what they expected to earn in that period.  If luck is 50%, they earned half.  CDF is nice for explaining the odds of bad rounds, but most miners are only concerned with earnings, and CDF doesn't help with visualizing what they earned compared to what they "should have" earned.

I understand, you're considering "luck" as a measure of profit.

However I think you're under-valuing the CDF. Once they find out how much less than normal they earned, most miners start asked "How likely is that"? Only the CDF can tell you that. It's not just for rounds, but can be used to evaluate the probability of a given mean over many rounds.

legendary
Activity: 1750
Merit: 1007
I've always used Luck as a representation of actual earnings vs expected earnings, because that is what most miners are concerned about, and it is easy for them to understand.  If luck is 200%, they made twice what they expected to earn in that period.  If luck is 50%, they earned half.  CDF is nice for explaining the odds of bad rounds, but most miners are only concerned with earnings, and CDF doesn't help with visualizing what they earned compared to what they "should have" earned.
donator
Activity: 2058
Merit: 1007
Poor impulse control.
I think the method used by Eligius and BTC Guild is a better representation of "luck" since whether or not a block I orphaned is part of luck, IMO.

Perhaps, but most miners do consider them different and tend to subtract their orphan estimate form their luck estimate.

Most importantly though, this sort of luck statistic is not easily analysed. For example, you can't (as far as I know) calculate a CDF for a number of blocks when the mean value (that is the percentage of valid blocks) is also a random variable.

I think we need to think about these things separately and express that clearly. Instead of 'luck':  percentage of expected profit, and percentage of expected shares.

Anyone agree?

Edit: Yes, what Kano said (better than I did)
legendary
Activity: 4592
Merit: 1851
Linux since 1997 RedHat 4
I think thr method used by Eligius and BTC Guild is a better representation of "luck" since whether or not a block I orphaned is part of luck, IMO.
However, the expected luck of the pool is calculated based on difficulty and submitted shares.

Some pools get a lot of orphans, and in that case they need to resolve why and fix it ...

Not reporting the pools orphans anywhere at all would be like giving up on the pools ability to compete favourably with other pools.
donator
Activity: 2058
Merit: 1007
Poor impulse control.
* Your pool luck scores conflate losses of blocks to orphans (which will vary) and poor pool luck (which should in the long term equal 1). This means that any analysis of the accepted / expected share values can only be estimates using the method suggested below.

* Whenever a miner compares your luck scores to that of another pool, your score will always appears lower since you include the shares attributed to the orphaned block.

BTC Guild's luck is reported in the same way.  Orphaned block solves are not included.

It's ok to include orphans as long as the shares submitted to that are attributed to them, not to the next block. For your luck scores, are you ignoring orphaned block shares as well as the block solves?
legendary
Activity: 1223
Merit: 1006
I think thr method used by Eligius and BTC Guild is a better representation of "luck" since whether or not a block I orphaned is part of luck, IMO.
legendary
Activity: 1750
Merit: 1007
* Your pool luck scores conflate losses of blocks to orphans (which will vary) and poor pool luck (which should in the long term equal 1). This means that any analysis of the accepted / expected share values can only be estimates using the method suggested below.

* Whenever a miner compares your luck scores to that of another pool, your score will always appears lower since you include the shares attributed to the orphaned block.

BTC Guild's luck is reported in the same way.  Orphaned block solves are not included.
donator
Activity: 2058
Merit: 1007
Poor impulse control.
My guess is that Eligius does something odd with it's share accounting - maybe adding the orphaned shares to the next valid block? If they're doing that, I can't undo it and determine the actual luck and CDF

For stats and reward purposes at Eligius, orphaned blocks are treated as if they never happened (the round continues, share count continues, the orphaned block's share is just another share in the longer round).  Does this help at all?

Yes it does, and sounds familiar. I think we've had this chat before and I'd forgotten.

The stats don't waste time calculating the accepted share count for an orphaned block,

I understand why the scoring has been set up that way since - in a way and from the miner's POV -   the orphaned block never occurred at all. However doing so means:

* Your pool luck scores conflate losses of blocks to orphans (which will vary) and poor pool luck (which should in the long term equal 1). This means that any analysis of the accepted / expected share values can only be estimates using the method suggested below.

* Whenever a miner compares your luck scores to that of another pool, your score will always appears lower since you include the shares attributed to the orphaned block.

but the time stamps are there for them in the stats and API, and a reasonable approximation could be made I'm sure.

I think so too. Your API is great, has as much information in it as anyone would need.

So, for anyone who wants to recalculate the luck for Eligius, just take the durations for the orphaned blocks and the block immediately following them, split the shares from the second block between the two blocks according to the ratios of the durations, and you're done.

note:  the 'profit percentage' plot on my blog avoids this issue completely by including orphaned blocks.
legendary
Activity: 1223
Merit: 1006
My guess is that Eligius does something odd with it's share accounting - maybe adding the orphaned shares to the next valid block? If they're doing that, I can't undo it and determine the actual luck and CDF

For stats and reward purposes at Eligius, orphaned blocks are treated as if they never happened (the round continues, share count continues, the orphaned block's share is just another share in the longer round).  Does this help at all?

The stats don't waste time calculating the accepted share count for an orphaned block, but the time stamps are there for them in the stats and API, and a reasonable approximation could be made I'm sure.
donator
Activity: 2058
Merit: 1007
Poor impulse control.
Below is a comparison of Eligius' (submitted d1 shares) / (expected d1 shares) to that of bitcoinPool, which had such extraordinarily bad luck that one could be justified in believing that something was badly wrong. Luck might be bad for Eligius recently, but it was pathological for bitcoinPool.

Still, something weird is going on. Luck for the pool's valid blocks is:
Code:
 (submitted d1 shares) / (expected d1 shares) = 1.057674 over 8925 valid blocks. 
The lower tail CDF for this is: 0.99999995. This is so close to 1 that Wolfram Alpha just calculates it as 1:

http://www.wolframalpha.com/input/?i=CDF%5BErlangDistribution%28+8925%2C++8925%29%2C+1.057674%5D

This means that this really is unlikely to ever happen. However I'd like to make it clear that I don't think anything nefarious (on the pool's side) is going on here.

My guess is that Eligius does something odd with it's share accounting - maybe adding the orphaned shares to the next valid block? If they're doing that, I can't undo it and determine the actual luck and CDF.


legendary
Activity: 4592
Merit: 1851
Linux since 1997 RedHat 4
I'll show you a list of links that says your supposition is false:
http://organofcorti.blogspot.com.au/2014/12/december-14th-2014-bitcoin-mining-pool.html
http://organofcorti.blogspot.com.au/2014/12/december-20th-2014-mining-pool.html
http://organofcorti.blogspot.com.au/2015/01/january-11th-2015-mining-pool-statistics.html
http://organofcorti.blogspot.com.au/2015/01/january-18th-2015-mining-pool-statistics.html
http://organofcorti.blogspot.com.au/2015/01/january-25th-2015-mining-pool-statistics.html
http://organofcorti.blogspot.com.au/2015/02/february-1st-2015-mining-pool-statistics_2.html
newbie
Activity: 28
Merit: 0
website is back  Grin
legendary
Activity: 924
Merit: 1000
Dark Passenger Bitcoin miner 2013,Bitcoin node
I have no failover so still hashing away Cheesy
newbie
Activity: 28
Merit: 0
its been like this since about 0330  Huh

hope only the website down. miner still running to this pool.
legendary
Activity: 924
Merit: 1000
Dark Passenger Bitcoin miner 2013,Bitcoin node
its been like this since about 0330  Huh
Pages:
Jump to: