[1500 TH] p2pool: Decentralized, DoS-resistant, Hop-Proof pool

[Red Emerald]

Pool is dying now.. Lost 1/4 of the hashrate. Perhaps it'll come back if there's a good day. Although that would take a 4-7 block streak to bring luck just up to 100%PPS levels for 90 days.

I think it's strange that bad luck streaks at non-PPS pools freak miners out less than bad luck at p2Pool - a pool where a pool op cannot maliciously cause bad luck (lookin' at you, Bitclockers 2011 - early 2012).

I was under the understanding that hashrate has nothing to do with luck.
Successfully finding coins fluctuates and could be said to vary based on luck but hashrate is constant for a given hardware barring bugs.

He is saying that because of the luck, miners are leaving the pool, not that the luck is lowering the hash rate.

I think its also likely that GPU miners are going offline in general and ASIC miners are not pointing at p2pool because of the initial reports of problems with stratum.

[taltamir]

Pool is dying now.. Lost 1/4 of the hashrate. Perhaps it'll come back if there's a good day. Although that would take a 4-7 block streak to bring luck just up to 100%PPS levels for 90 days.

I think it's strange that bad luck streaks at non-PPS pools freak miners out less than bad luck at p2Pool - a pool where a pool op cannot maliciously cause bad luck (lookin' at you, Bitclockers 2011 - early 2012).

I was under the understanding that hashrate has nothing to do with luck.
Successfully finding coins fluctuates and could be said to vary based on luck but hashrate is constant for a given hardware barring bugs.

He is saying that because of the luck, miners are leaving the pool, not that the luck is lowering the hash rate.

I think its also likely that GPU miners are going offline in general and ASIC miners are not pointing at p2pool because of the initial reports of problems with stratum.

Thanks, I misunderstood.
I mixed up this thread and another I have open about a recently reported bug with tanking hashrates that happens on certain miner configurations (for all pools)

[organofcorti]

Just finished a multivariate Augemented Dickey Fuller test on the p2Pool data shares per round data since Novemeber last year, to find if a trend has actually occurred.

p value for trend occurring: 0.96
p value for drift occurring: 0.99
p value for no trend or drift in data: 0.019

So (and check this out, kano ) there has been no trend in the data since Novemeber last year.

Why does it look like there has been some kind of cyclical trend? This is probably due to the method used to illustrate "luck". Any interpretation is likely to vary from actuality in some way, and I recommend you don't use luck charts in general for any important decision about mining.

I'll try to post boxplots and smoothing spline results later.

[rav3n_pl]

Looks like new chapter on NPW about P2Pool is comming

[organofcorti]

As a follow up to my last post, here's a better way of judging luck:



the boxplots show the median, quartiles and outliers. The median (the line through the middle of the box) should be ln(2). As you can see, there is significant variation - but not a great deal more variation recently than in previous times.

Looks like new chapter on NPW about P2Pool is comming

I'm flat out like a lizard drinking atm so if I can make some time for it, sure. But I get a feeling that everytime theres a new p2Pool miner, or new data, someone's going to question whether there's some new problem with the luck. Responding to the same sort of issues every few months is a bit disheartening. It's a Sisyphean task.

[kano]

Just finished a multivariate Augemented Dickey Fuller test on the p2Pool data shares per round data since Novemeber last year, to find if a trend has actually occurred.

p value for trend occurring: 0.96
p value for drift occurring: 0.99
p value for no trend or drift in data: 0.019

So (and check this out, kano ) there has been no trend in the data since Novemeber last year.

Why does it look like there has been some kind of cyclical trend? This is probably due to the method used to illustrate "luck". Any interpretation is likely to vary from actuality in some way, and I recommend you don't use luck charts in general for any important decision about mining.

I'll try to post boxplots and smoothing spline results later.

Ah well - it just looked like it - but I guess, as we all know, maths often proves something isn't how it looks

[organofcorti]

Ah well - it just looked like it - but I guess, as we all know, maths often proves something isn't how it looks

I know. Everytime I learn something new in probability and statistics I start with an intuitition of what it's about and how it works. I'm almost always wrong.

[btharper]

As a follow up to my last post, here's a better way of judging luck:



the boxplots show the median, quartiles and outliers. The median (the line through the middle of the box) should be ln(2). As you can see, there is significant variation - but not a great deal more variation recently than in previous times.

Looks like new chapter on NPW about P2Pool is comming

I'm flat out like a lizard drinking atm so if I can make some time for it, sure. But I get a feeling that everytime theres a new p2Pool miner, or new data, someone's going to question whether there's some new problem with the luck. Responding to the same sort of issues every few months is a bit disheartening. It's a Sisyphean task.

Any way to get a more in-depth statistical analysis automated enough to host it somewhere like p2pool.info?

[maqifrnswa]

I know it's bad luck and all, but there has been 5 blocks found in the past week (should have been 11-14).

[organofcorti]

Any way to get a more in-depth statistical analysis automated enough to host it somewhere like p2pool.info?

It's possible to automate it. Don't know if p2Pool.info could host it though.

[zvs]

i've officially made a run for the hillz

[rav3n_pl]

We need more power. Pool have to be over 600GH to have chance to make over 3 blocks per day and pay like it should.
NOW is best time to join p2pool ;]

[lenny_]

We need more power. Pool have to be over 600GH to have chance to make over 3 blocks per day and pay like it should.
NOW is best time to join p2pool ;]

Yeah, but who gonna pay for extra 300Gh/s of GPU hashpower? As we know, FPGA are useless on p2pool, and GPU miners are leaving slowely, moving to newer machines. P2pool has to change its strategy or it will vanish.

[gyverlb]

We need more power. Pool have to be over 600GH to have chance to make over 3 blocks per day and pay like it should.
NOW is best time to join p2pool ;]

Yeah, but who gonna pay for extra 300Gh/s of GPU hashpower? As we know, FPGA are useless on p2pool

Bullshit: the last block I found on p2pool was with a Spartan6, they work perfectly on p2pool. Only BFL FPGAs are useless on p2pool

and GPU miners are leaving slowely, moving to newer machines. P2pool has to change its strategy or it will vanish.

If/When I get my Avalon(s) I'll put them on p2pool...

[zvs]

We need more power. Pool have to be over 600GH to have chance to make over 3 blocks per day and pay like it should.
NOW is best time to join p2pool ;]

figures that like 3 hrs after i switch pools i get a block

guess it's not so bad tho cause of the 1/2 transaction fee bonus

[K1773R]

We need more power. Pool have to be over 600GH to have chance to make over 3 blocks per day and pay like it should.
NOW is best time to join p2pool ;]

Yeah, but who gonna pay for extra 300Gh/s of GPU hashpower? As we know, FPGA are useless on p2pool

Bullshit: the last block I found on p2pool was with a Spartan6, they work perfectly on p2pool. Only BFL FPGAs are useless on p2pool

not quite true, u can even use BFL FPGA's with p2pool if you use cgminer in combination with --bfl-range

[rav3n_pl]

Only trouble witch p2pool have devices that cant handle reset work messages and/or cant return nonces found in moment that they found them.
BFLs and some FPGAs returning results from work only when they do full nonce, with in case of p2pool is 50% chance to return work after longpool and all of them are DOA instantly.
GPUs and all "good" devices not have that issues - they return work instantly and asking for new one when hit by longpool/work reset.
Crucial is to make nonce as soon as possible, so changing threads (work units that are calculated in same time) to 1 on any device can actually make better efficiency (less rejects and more full nonces) even if hash rate is bit lower.

[spiccioli]

not quite true, u can even use BFL FPGA's with p2pool if you use cgminer in combination with --bfl-range

--bfl-range does not work with singles, but only with minirigs, if I remember correctly.

spiccioli

[Subo1977]

i think a few Day's with this "good" L*, the people goes away.

[zvs]

typical

my switch died though anyway, so i was going to be screwed regardless =p

(12 hrs of like 1 computer mining)