Pages:
Author

Topic: Deepbit hopping (Read 6132 times)

sr. member
Activity: 252
Merit: 251
February 20, 2012, 11:05:09 PM
#44
Pool Ops that offer PPS to their members and hop with their hash power should be exposed.  

Hopping proportional pools and exploiting their resources is a natural step in the evolution process of bitcoin pools.

Large proportional pools have been on the decline for ages & will eventually die out completely once the core miners realize they are working for someone else;
They are paying out of their own pocket so that somebody else may profit.

I think Inaba on page 2, the master of Eclipse pool, has put it eloquently in the last sentence:
Quote from: Inaba
A new block starts and they will jump to 700 - 900 GH/s, then like clockwork at ~40% their hashrate plummets to 100 - 150 GH/s.  
Those poor miners making up the 150 GH/s are getting raped like a two bit whore.

Of course, those 101+% pools will disappear too, but only once the proportional model disappears first. Right now, non-hopping, perpetual proportional miners are the suckers of the system, the prey. Through either lack of knowledge or sheer ignorance, they distort the balance of pools so as to create an incentive for predators.
True equality in mining rewards will exist, but only when these miners finally understand they are damaging their own earnings.

(This model relies on the rational assumption miners' main motive is profit;
It does not take in account irrational motives such as supporting a pool for ideological reasons, or wanting to give free money to pool hoppers out of generosity etc., so it is still possible there could be proportional miners 3-5 years down the road)
hero member
Activity: 518
Merit: 500
February 16, 2012, 04:05:09 PM
#43
For interest I added another five days data to try to increase the size of the dataset (a few over 600 blocks) because Deepbit has had a few 2-3 hour solves recently (including a 184 minute monster).

Distribution is still very noisy with an average solve time of 26 minutes (Sd of 27 minutes) and average hash rate 51,440 shares/minute (s.d. 3246).  Testing different regression fits gave less than 5% R^2.
hero member
Activity: 518
Merit: 500
February 08, 2012, 04:25:36 PM
#42
Pool hopping is detectable by the pool op. If a user's per-share reward value is consistently higher than average, they are gaming the payout system. In aggregate data like the total per-share reward vs round time, observation by outsiders will be masked by the total pool size and the modest counter-measures, but on a per-user basis, there will be users who earn the average expected per-share reward, and hoppers that stick out like a sore thumb.

Agreed. But since I made the post, Deepbit has linked to their own thread were is shown that hoppers are not only detected, but also labeled as such. Just no action is being taken against them - for now.

I hadnt seen that when I made my analysis, I just tried proving deepbit can be and is being hopped. That much is beyond doubt now.

I do agree because of deepbits size, the current impact on non-stop miners by hoppers is trivial and statistically noise, that was never my concern.
legendary
Activity: 1512
Merit: 1028
February 08, 2012, 04:17:57 PM
#41
Pool hopping is detectable by the pool op. If a user's per-share reward value over time is consistently higher than average, they are gaming the payout system. In aggregate data like the total per-share reward vs round time, observation by outsiders will be masked by the total pool size and the modest counter-measures, but on a per-user basis, there will be users who earn the average expected per-share reward, and hoppers that stick out like a sore thumb with higher earnings.

Hoppers will also submit a consistent number of shares per round even when they are long (or shares per block if just hopping on new network block information), and by quitting long rounds, are identifiable.

The time scale of the graph in the first post should be logarithmic, so there is an even spread of round quantiles over time.
hero member
Activity: 518
Merit: 500
February 08, 2012, 03:39:33 PM
#40
For curiosity I did a graph of the last 321 blocks.  Time per block to the nearest second versus shares per minute.  I too can get a slight decline in hashes returned for the longer blocks.  Average shares/minute = 51,300, Standard Deviation = 2,127.

Better yet, I added a trend-line and calculated the R^2 values.  The result, with R^2 less than 10% tells me lots of noise and any relationship, if there is one, is weak.  The longest block in the sample had a hash rate higher than average - a random result Smiley
hero member
Activity: 518
Merit: 500
February 08, 2012, 12:25:41 PM
#39
Would somebody who is on PPS be effected more than Proportional? Because i'm on Proportional, the avg. payout was higher within a few days time at my current Mhashs.

PPS would not be affected.
Of course, that has to be put in perspective, considering deepbit charges, what 10% fee for PPS? That makes hopping a total non issue by comparison, but you still get paid 10% less than on any (hop proof) zero fee pool though.
member
Activity: 63
Merit: 10
February 08, 2012, 12:19:33 PM
#38
Would somebody who is on PPS be effected more than Proportional? Because i'm on Proportional, the avg. payout was higher within a few days time at my current Mhashs.

I think there is a lot of sarcasm around here that i'm taking for fact and getting confused lol.
hero member
Activity: 518
Merit: 500
February 08, 2012, 12:08:05 PM
#37
no no, no one but p4man thinks deepbit is getting hopped.

God you're such an idiot.
"No one but me" and Deepbit themselves Roll Eyes

Welcome to my ignore list again. I wonder why I ever let you out.
hero member
Activity: 518
Merit: 500
February 08, 2012, 12:03:24 PM
#36
I never understood the purpose to that, my payout is more or less consistent within 1-2%, very rarely (maybe once every other day) i will only get 80/160,000 shares and it'll be higher or lower than normal, but as far as i can see Pool hopping hasn't done anything to my payout if it's happening at all. (Deepbit)

If the hopping were to happen constantly, you wouldnt see it from your payout, it would always be "normal" but x% lower than 100% PPS for the current difficulty. Im not saying I think the impact from hopping is significant though, at least not compared to the fees you are paying anyway Wink.
member
Activity: 63
Merit: 10
February 08, 2012, 11:39:15 AM
#35
I never understood the purpose to that, my payout is more or less consistent within 1-2%, very rarely (maybe once every other day) i will only get 80/160,000 shares and it'll be higher or lower than normal, but as far as i can see Pool hopping hasn't done anything to my payout if it's happening at all. (Deepbit)
legendary
Activity: 1750
Merit: 1007
February 08, 2012, 11:29:51 AM
#34
Pool Ops that offer PPS to their members and hop with their hash power should be exposed. 

Exposed for being savvy businessmen. The miners are getting exactly what they deserve. What's the problem?

1)  They're actively hurting users of other pools.  Not every pool can handle swings of hundreds of GH/s, especially the smaller pools that still use hoppable methods.  It increases stale rates and the frequency of miners idling.  The pool ops can't afford to upgrade to stronger servers just to service a proxy-pool that only shows up for a small time frame.

2)  They're not being transparent, and it can be far more dangerous than the FUD that gets spread about Deepbit.  They should at least have on their site available for all to see what pools they hop, and what pool(s) they are currently hashing on.
hero member
Activity: 518
Merit: 500
February 08, 2012, 10:21:22 AM
#33
Yeah, I highly doubted deepbit was getting heavily hopped.

Non sequitur, depending how you define "heavily".My point still stands.

AFAICT, they are being hopped by something like 5-10% of the hashing power. That I cant prove an effect on the payouts of non stop miners is entirely to be expected since Id need a much bigger sample for that than just the last few days available on their site; there are many reasons you can get slightly above expected payouts (excluding their high fees of course) on such a short span, the most important one being pool luck.  Other factors are stales, since I have no data for them. Since they occur more frequently at the beginning of new blocks, it could chance the result significantly.
hero member
Activity: 518
Merit: 500
February 08, 2012, 09:07:52 AM
#32
I just analyzed Bitclockers stats and posted the results here:
https://bitcointalksearch.org/topic/m.737310

No need to for any regression analysis there me thinks lol
Short story: poor (/idiot) constant miners there are paying 10-12% to the hoppers, +2% fees to the pool.

I also tried calculating that for deepbit using second intervals, but either I screwed up somewhere or its statistical noise with pool luck and stales being a bigger factor, as I ended up getting about 0.2% better than expected earnings for constant miners.
donator
Activity: 532
Merit: 501
We have cookies
February 07, 2012, 08:45:21 PM
#31
So yes, GPUMAX is responsible for sending purchases to deepbit on occasion but it's not a popular pool for those that purchase shares.  I think one of the things that you might want to look at is when we're doing updates or have suspended purchases our miners fail-over to their offline pools.  So this could cause a spike in hashes that could make it look like they're hopping the pool when in reality it's just them failing over. I have no problem adding Deepbit to the not supported list for purchases if it's causing that much concern but I can't control where people fail-over to when we're not processing purchases.
Yes, I noticed some users are getting enormous bursts of up to 160+ GH/s sometimes after your service started and this is completely fine for me unless they are hopping, so you can keep Deepbit in your "supported" list. Hoppers would direct their purchase to some proxy anyway.
The hopping incidents were completely different because its very clear to me if their starts and stops are correlated with many of our blocks.
sr. member
Activity: 378
Merit: 250
"Yes I am a pirate, 200 years too late."
February 07, 2012, 08:29:58 PM
#30
So yes, GPUMAX is responsible for sending purchases to deepbit on occasion but it's not a popular pool for those that purchase shares.  I think one of the things that you might want to look at is when we're doing updates or have suspended purchases our miners fail-over to their offline pools.  So this could cause a spike in hashes that could make it look like they're hopping the pool when in reality it's just them failing over.  Technically they are hopping but not the traditional way.  I have no problem adding Deepbit to the not supported list for purchases if it's causing that much concern but I can't control where people fail-over to when we're not processing purchases.
hero member
Activity: 742
Merit: 503
February 07, 2012, 08:27:17 PM
#29
Pool Ops that offer PPS to their members and hop with their hash power should be exposed. 
donator
Activity: 532
Merit: 501
We have cookies
February 07, 2012, 06:01:13 PM
#28
Oh, so you know what's the difference Smiley
hero member
Activity: 518
Merit: 500
February 07, 2012, 05:35:32 PM
#27
Do you really not understanding why we are asking this

No. But if you like lots of blank space, here you go":


I even resisted the urge to add to -100.000
I guess what you really wanted me to show was this:



While what you really should see is this:



I can stretch and zoom it any other way you want, it keeps showing a little under 10% hopping.
hero member
Activity: 518
Merit: 500
February 07, 2012, 05:11:52 PM
#26
Bitclockers would be a great one.  Their stats are fun to watch.  A new block starts and they will jump to 700 - 900 GH/s, then like clockwork at ~40% their hashrate plummets to 100 - 150 GH/s.  Those poor miners making up the 150 GH/s are getting raped like a two bit whore.

Eeeck. I think this chart on their own website already shows what you mean:



Im having a vague idea when they found blocks looking at that. Ill see if I can plot it against their blocks tomorrow. I got a feeling its indeed not going to be pretty  Shocked
donator
Activity: 532
Merit: 501
We have cookies
February 07, 2012, 05:09:10 PM
#25
Quote
The point about graphing the Y axis from zero would better show how much (or little) decline there is.  For instance, is deepbit losing 5% or more from hoppers, or less.  It is hard to tell given the way it is presented.
Speed never drops to zero. As i said over and over, its just blank space at the bottom. Or top, why not add a lot at the top?
The chart is fairly clear I think, from the sampled data there seems to be about 10% hopping.
Do you really not understanding why we are asking this, or you just want the graph to be more impressive ? :)

I was talking about monitoring all the hoppers, not the total/average hashrate.
Okay, but even so, the hoppers might just have "pooled together" now.
Well, you can believe in this if you want to.
I just told what my observations show me, this opinion is based on really good stats collection.
Pages:
Jump to: