@ckolivas: I can see, for people who want a high variance pool to essentially gamble, that rewarding shares above current difficulty might be part of the attraction. Indeed, having the maximum share payout be (significantly) above 25 BTC would be, too. Given Graet is choosing fairly agressive parameters, I'm assuming that's what he's after?
No he's not. There was never any intention of making this much greater risk for the pool, only more fun for the miners. Higher variance but capping it would achieve that goal. 
Topic: Pay On Target: New High variance payout System Offered by Ozcoin - page 9. (Read 36502 times)
December 18, 2012, 06:11:24 PM
December 18, 2012, 06:08:12 PM
@ckolivas: I can see, for people who want a high variance pool to essentially gamble, that rewarding shares above current difficulty might be part of the attraction. Indeed, having the maximum share payout be (significantly) above 25 BTC would be, too. Given Graet is choosing fairly agressive parameters, I'm assuming that's what he's after?
@kano: that example isn't really that extreme. Assuming it's ballpark 30 times difficulty you'd expect the network as a whole to generate such a block every 30 blocks or every five hours. And you got less than half what you'd got if you'd mined that block yourself. (EDIT: so Ozcoin would expect to get such an event every five days or so - such events will average out nicely.) It's the tail events that pose the risk to the pool operator. (EDIT: eg the 1G share, which Ozcoin will find once every year-and-a-half, and which will cost Graet 400 BTC at a=0.8, or the 10G block that Ozcoin would expect to see once every 10-15 years which would cost Graet 2500 BTC. Hope I got my maths right, that was a very quickback of the envelope back of the bc -l extrapolation and I may well have messed up.)
@kano: that example isn't really that extreme. Assuming it's ballpark 30 times difficulty you'd expect the network as a whole to generate such a block every 30 blocks or every five hours. And you got less than half what you'd got if you'd mined that block yourself. (EDIT: so Ozcoin would expect to get such an event every five days or so - such events will average out nicely.) It's the tail events that pose the risk to the pool operator. (EDIT: eg the 1G share, which Ozcoin will find once every year-and-a-half, and which will cost Graet 400 BTC at a=0.8, or the 10G block that Ozcoin would expect to see once every 10-15 years which would cost Graet 2500 BTC. Hope I got my maths right, that was a very quick
December 18, 2012, 05:51:39 PM
Currently there's no upper bound on the value of a share... If someone finds a share of difficulty several billion, several quadrillion, whatever, it will just bankrupt Graet. It's possible for someone to submit a share that earns a payout of more than the 21,000,000 BTC that will ever be mined.
Would it not be prudent to put a cap on the maximum payout from a share, say BTC 250 or BTC 2500 or whatever Graet feels comfortable with. (This would also have the side effect of making the variance finite. In reality such a payout limit already exists - but better to implement it in the maths than by bankruptcy
EDIT: Of course, this almost certainly makes the analysis a complete bitch.
EDIT: Also of course, a simple cap would make the expected payout dependent on the work difficulty. I imagine there are ways of fixing that, at least if the pool is also willing to impose a maximum work difficulty when using this scheme. (The alternative approach would be a scheme that puts a cap on share difficulty, perhaps at some multiple of the current difficulty.)
roy
Would it not be prudent to put a cap on the maximum payout from a share, say BTC 250 or BTC 2500 or whatever Graet feels comfortable with. (This would also have the side effect of making the variance finite. In reality such a payout limit already exists - but better to implement it in the maths than by bankruptcy
EDIT: Of course, this almost certainly makes the analysis a complete bitch.
EDIT: Also of course, a simple cap would make the expected payout dependent on the work difficulty. I imagine there are ways of fixing that, at least if the pool is also willing to impose a maximum work difficulty when using this scheme. (The alternative approach would be a scheme that puts a cap on share difficulty, perhaps at some multiple of the current difficulty.)
roy
As I mentioned in #cgminer yesterday, my last OzCoin block I got was
[2012-11-28 18:02:55] Accepted 0000002b Diff 99.2M/8 BLOCK! ICA 0 pool 0
Which was also just before halving (block 209939)
So calculating it now, but based on the numbers back then (with a=0.8 and BTC=50)
(8 * (50*.97) / 3438908.9601591) * (1-0.8 ) * (99200000 / 8 )^0.8 = 10.67...BTC
So even with slightly higher than usual blocks it's was already a pretty high payout at 0.8
Recalculating just changing it to 0.5 ...
(8 * (50*.97) / 3438908.9601591) * (1-0.5) * (99200000 / 8 )^0.5 = 0.19865...BTC
Oh well.
Yes I mine at 8 difficulty (even though I'm only around total 2GH/s)
Makes the network a lot quieter for my (old'ish) NetGear PowerLine down to my garage
December 18, 2012, 05:31:49 PM
As an aside, I'm disappointed they didn't adopt the name I chose for this payment scheme: Difficulty Extrapolated Reward Payment, aka DERP.
Graet, I respect you and greatly appreciate all the hard work you've put in, but how could you make a mistake like this?!
December 18, 2012, 05:15:37 PM
To be fair, what I had originally suggested with this scheme was to cap the maximum paid to that of the current block difficulty to prevent the ridiculously high difficulty share from bankrupting the pool.
As an aside, I'm disappointed they didn't adopt the name I chose for this payment scheme: Difficulty Extrapolated Reward Payment, aka DERP. Though I guess one could argue we're now hashing pot.
As an aside, I'm disappointed they didn't adopt the name I chose for this payment scheme: Difficulty Extrapolated Reward Payment, aka DERP. Though I guess one could argue we're now hashing pot.
December 18, 2012, 04:58:28 PM
Currently there's no upper bound on the value of a share... If someone finds a share of difficulty several billion, several quadrillion, whatever, it will just bankrupt Graet. It's possible for someone to submit a share that earns a payout of more than the 21,000,000 BTC that will ever be mined.
Would it not be prudent to put a cap on the maximum payout from a share, say BTC 250 or BTC 2500 or whatever Graet feels comfortable with. (This would also have the side effect of making the variance finite. In reality such a payout limit already exists - but better to implement it in the maths than by bankruptcy
EDIT: Of course, this almost certainly makes the analysis a complete bitch.
EDIT: Also of course, a simple cap would make the expected payout dependent on the work difficulty. I imagine there are ways of fixing that, at least if the pool is also willing to impose a maximum work difficulty when using this scheme. (The alternative approach would be a scheme that puts a cap on share difficulty, perhaps at some multiple of the current difficulty.)
roy
Would it not be prudent to put a cap on the maximum payout from a share, say BTC 250 or BTC 2500 or whatever Graet feels comfortable with. (This would also have the side effect of making the variance finite. In reality such a payout limit already exists - but better to implement it in the maths than by bankruptcy
EDIT: Of course, this almost certainly makes the analysis a complete bitch.
EDIT: Also of course, a simple cap would make the expected payout dependent on the work difficulty. I imagine there are ways of fixing that, at least if the pool is also willing to impose a maximum work difficulty when using this scheme. (The alternative approach would be a scheme that puts a cap on share difficulty, perhaps at some multiple of the current difficulty.)
roy
December 18, 2012, 04:15:13 PM
For now, I don't even care if I make less or not. I just like seeing my average share go up and down.
Also, the OZcoin widget seems to be keeping up with my estimated hashrate, but isn't reporting any shares in the worker section. Also, the Estimated Payout section is having issues. None of this is really unexpected.
Also, the OZcoin widget seems to be keeping up with my estimated hashrate, but isn't reporting any shares in the worker section. Also, the Estimated Payout section is having issues. None of this is really unexpected.
December 18, 2012, 12:40:45 PM
Graet,
One of my miners shows offline but I'm not so sure. My rejects went from 6 to 500+ and climbing consistently now like every share from that one miner is rejecting.
-Wave
One of my miners shows offline but I'm not so sure. My rejects went from 6 to 500+ and climbing consistently now like every share from that one miner is rejecting.
-Wave
December 18, 2012, 12:01:48 PM
Moving to a = 0.5 while i look at more figures
and 4% fee on advice
Interesting choice, the variance will still be infinite but just barely so.and 4% fee on advice
If you reduce a further maybe 3% fee can be enough.
Any way to make it selectable by the user? .5 .6 .7 or .8 with appropriate fees? Could get interesting...
ooc was interested in that too, it *may* be possible yes
I'd like to make this interesting, but don't want to go broke doing it
I don't think any of us wants you to go broke.
So is this new payout method for only those who are using higher difficulty targets?
December 18, 2012, 09:40:37 AM
Understood, if you would like to test seperatly I will "donate" my hasing power at up to 10% fee PPS for 1 day at each level...
-Jason
-Jason
Thanks man
I think for the "trial period" I will put fees to 3% - a little sweetener for miners while they try it
Please expect them to go to 4% after trial.
I don't think any of us wants you to go broke.
So is this new payout method for only those who are using higher difficulty targets?
We could have used a different formula and forced everyone to diff1 for this payout method but thought allowing Variable difficulty it would be better for miners
December 18, 2012, 09:39:33 AM
Any way to make it selectable by the user? .5 .6 .7 or .8 with appropriate fees? Could get interesting...
ooc was interested in that too, it *may* be possible yes
I'd like to make this interesting, but don't want to go broke doing it
I don't think any of us wants you to go broke.
So is this new payout method for only those who are using higher difficulty targets?
December 18, 2012, 09:32:30 AM
Understood, if you would like to test seperatly I will "donate" my hasing power at up to 10% fee PPS for 1 day at each level...
-Jason
-Jason
December 18, 2012, 09:29:06 AM
Any way to make it selectable by the user? .5 .6 .7 or .8 with appropriate fees? Could get interesting...
ooc was interested in that too, it *may* be possible yes
I'd like to make this interesting, but don't want to go broke doing it
December 18, 2012, 09:26:41 AM
Any way to make it selectable by the user? .5 .6 .7 or .8 with appropriate fees? Could get interesting...
Hahaha now thats gambling
December 18, 2012, 09:22:11 AM
Any way to make it selectable by the user? .5 .6 .7 or .8 with appropriate fees? Could get interesting...
December 18, 2012, 08:51:54 AM
Moving to a = 0.5 while i look at more figures
and 4% fee on advice
and 4% fee on advice
Not sure how I feel about that. The huge POTENTIAL upside is gone. But I will run for a day for sure.
December 18, 2012, 08:50:08 AM
I guess I'm down under if the average share is below 5. Is that correct? 
Does not appear that way:
I'm at 5021 shares, avg/ Dif 10.938 and payout is 0.03173515, or per share of 0.00000632.
-Wave
0.8 gave too much variance, see a=0.5 example in 1st post
December 18, 2012, 08:48:29 AM
Moving to a = 0.5 while i look at more figures
and 4% fee on advice
and 4% fee on advice
Not sure how I feel about that. The huge POTENTIAL upside is gone. But I will run for a day for sure.
December 18, 2012, 08:40:50 AM
I guess I'm down under if the average share is below 5. Is that correct?
Does not appear that way:
I'm at 5021 shares, avg/ Dif 10.938 and payout is 0.03173515, or per share of 0.00000632.
-Wave
December 18, 2012, 08:38:42 AM
Moving to a = 0.5 while i look at more figures
and 4% fee on advice
and 4% fee on advice
Jump to: