Pool hopping... ethical or not?

[k9quaint]

So do tell...what is the purpose of all definitions?

To point a person towards the concept that a word names.

Fair enough.  So it seems like you've kind of proved my point.  If a definition serves merely as the barest of guideposts and especially when addressing a specific audience it's goal is not to result in  a mutually understood term .  Then you are providing what I (and perhaps others) would call a poor definition.

So yeah.  You're bad at it.*

QED baby.

*Unless of course you meant that a mutually understood term is the goal and just said it badly...which also demonstrates my point. 

What is the definition of pedantic? 

Kinda depends, that word without looking it up probably has at least three pretty distinct senses.   None appear to apply here.   What made you think it did?

I was referring to definition number two: overly concerned with minute details or formalisms
I thought it might apply to a discussion in which two people are arguing over the definition of the word definition. 

[JoelKatz]

So here is the question. I had a potential customer to my mining contracts ask about a pool hopper in which I said no. But, if I am renting a VPS, and he wants to run a pool hopper, should I allow it?

If he promises not to hop polls that have an explicit prohibition on pool hopping, I'd say your responsibility is sufficiently indirect that you're not a bad person if you allow him to do it. You are perfectly justified in saying "no" if he doesn't promise to stick to only pools that explicitly allow hopping though.

[organofcorti]

The other thread is long and full of meandering and not on-topic posts. If you want an intro to what pool hopping is and why it works, you could try this blog post although it does require some active reading.

Also, try this game which might give you an idea of what's going on if you have the time to play with it.

Otherwise, the short story is you aren't at risk at a hopper proof pool. The best known are double-geo pools (like eclipse mc),PPLNS pools (eg mineco.in), SMPPS pools and variants (arsbitcoin, bitpit, eligius) and pure PPS pools (deepbit PPS).

[teukon]

Agreed.

Just to clarify, if you use a pool with a proportional reward system and you do not hop you are losing out (possibly by as much as 20%), even if said pool pretends to stop pool hopping by hiding statistics.  I would recommend a simple PPS pool with low variance for a beginner as it's easy to understand and carries pracitcally no risk and instead usually comes with a heavy fee (deepbit PPS for example), however if you are prepared for higher variance (high variance in absolutely no way affects your expected return; you just have to have the mental strength to be able to mine for a few days and receive nothing sometimes) then you can get a much better return elsewhere.

One thing worth mentioning is a new kind of mining called 'merged mining'.  This has been available for a few days now and few pools are using it.  Using it gives you namecoins for free as you mine bitcoins (at the moment this provides an income bonus of a little over 30% but expect this to fall in the coming weeks).  There are a few pools which use merged mining such as slush's pool so have a look for 'merged mining' on the forum if you're interested (namecoins can easily be traded for bitcoins).  Again, if you don't intend to hop, it is highly recommended that you choose a hopping-proof pool.

eligius.st - merged mining with SMPPS.  hop-resistant (technically hoppable but with limited effect) and with fairly low variance.

simplecoin.us - merged mining with PPLNS.  hop-proof and has a big promotion but with higher variance (shamelessly recommended because it's the pool I use and I want others to join so I get lower variance ).

Pro tip:  Sometime soon the Namecoin difficulty is going to fall sharply and at this time merged mining will provide approximately double the income of a normal Bitcoin pool.  This period will last for 2016 Namecoin blocks.

[organofcorti]

eligius.st - merged mining with SMPPS.  hop-proof and with fairly low variance.

teukon - just thought I'd mention that SMPPS pools can be hopped to reduce payback time, at least in simulation anyway. It won't increase your total payout, but will reduce the amount you are owed at any point in time.

As far as I can tell there are no strategies to optimise payments for double-geo and PPLNS.

[teukon]

eligius.st - merged mining with SMPPS.  hop-proof and with fairly low variance.

teukon - just thought I'd mention that SMPPS pools can be hopped to reduce payback time, at least in simulation anyway. It won't increase your total payout, but will reduce the amount you are owed at any point in time.

As far as I can tell there are no strategies to optimise payments for double-geo and PPLNS.

Man!  I should really trust my instincts.  Thanks for this!

I actually drafted that last post and said that I wasn't 100% sure that SMPPS was hop-proof but changed my mind because it looked like I was promoting my favourite pool too much and who was I to tell people that there might be a hopping problem with a reward system designed to be hop-proof!

Yes, it is definitely suspicious and I'm glad to have a little confirmation on this from someone that really knows what they are talking about.  I'm simply too busy to look at all the reward systems carefully and evaluate them mathematically (I reject unnecessarily complicated and inelegant systems out of hand).

The double-geo that I remember coming across a while ago on the other hand feels very much like a mathematically sound reward system.  I'm afraid I haven't looked into this in detail either but I'd be very surprised to learn that the double-geo reward systems have a hopping problem, at least with certain parameters fixed.

PPLNS is absolutely solid for a fixed N.  It is quite possible that a pool is varying N with time (as difficulty changes) in a naive way which would allow some mild hopping opportunities but I expect most PPLNS pool operators would make any changes to N gracefully (possibly accepting a bit of risk to themselves).  The same problems could be faced by starting and ending a PPLNS pool (a naive implementation will be a trapezoid scheme but a graceful one will avoid this).

Honestly though, my instincts tell me that most of the reward systems out there, even though hoppable, are perfectly fine to use (very few hop them correctly and make very little for their efforts anyway).  The proportional reward system is to be avoided though.  It is very hoppable and a 24/7 miner will lose out in a big way.

[organofcorti]

SMPPS PPLNS PPS double-geo

thanks for the replies, i think i'll go back to mining school first and come back to this a bit later.

Just start at EclipseMC first. Then go to mining school. You'll be safe at EclipseMC, and the payout system has low variance. Plus the owner is a nice guy, responds to requests quickly and has been thoroughly honest.

[johnj]

SMPPS PPLNS PPS double-geo

thanks for the replies, i think i'll go back to mining school first and come back to this a bit later.

PPS - Pay per Share. If the difficulty is 1,000 and the block reward is 50btc, each share is worth .05 BTC.  The pool op takes the hit on variance. They take the gain from the short rounds and hope it's enough to pay out the loss on longer rounds.

PPLNS - Pay per last N shares.  Pool hoppers profit by getting in on the beginning of a new round on a proportional pool hoping to get shares in on a short round.  By paying by the last N shares, hoppers lose their edge by having their early-round shares becoming worthless after N shares have passed since their last submitted share.

SMPSS - Simple Maximum PPS. Like PPS, but keeps a log of who submitted what share when, and pays out only a maximum of the equal PPS shares. Example: On a normal PPS, (at a difficulty of 1000 and reward of 50), you'd get .05 BTC/share regardless of if that round is solved in 5 shares or 5000 shares, but a series of unlucky blocks can wipe out server funds, and then you have a dead pool. On SMPPS though if the server funds ever become less than what needs to be paid out, miners are only paid their share of what's available - and when shorter rounds are solved (thus the pool fund gains a surplus), miners are compensated for their previous under-paid work. This method alleviates the pool op of a lot of risk, while also ensuring that in the event unlucky rounds deplete pool funds that miners will be compensated.

And yeah, double-geo is new to me haha

[Meni Rosenfeld]

Summary of mining pool reward systems
Analysis of Bitcoin pooled mining reward systems (work in progress)

eligius.st - merged mining with SMPPS.  hop-proof and with fairly low variance.

teukon - just thought I'd mention that SMPPS pools can be hopped to reduce payback time, at least in simulation anyway. It won't increase your total payout, but will reduce the amount you are owed at any point in time.

You yourself made the error of including SMPPS in a list of "hopper proof" methods.

As far as I can tell there are no strategies to optimise payments for double-geo and PPLNS.

Lie-in-wait. But no, assuming the attacker has no posterior knowledge of future blocks and that the atomic action is finding a share and submitting it to a pool of choice, double-geom and unit-PPLNS provably cannot be optimized.

and who was I to tell people that there might be a hopping problem with a reward system designed to be hop-proof!

SMPPS was "designed to be hop-proof" by someone who hasn't got a clue what he's talking about. I've explained why it's broken until I went blue in the face ever since it was spawned.

I'm simply too busy to look at all the reward systems carefully and evaluate them mathematically (I reject unnecessarily complicated and inelegant systems out of hand).

Way ahead of you, see links above.

at least with certain parameters fixed.

You can change the parameters, but you'll need to either modify the scores or using an invariant scoring in the first place.

PPLNS is absolutely solid for a fixed N.

No it's not. To be absolutely solid you need to use the "unit-PPLNS" variant (to which I refer in my PPLNS post simply as "the correct method").

The same problems could be faced by starting and ending a PPLNS pool (a naive implementation will be a trapezoid scheme but a graceful one will avoid this).

The correct way to end a PPLNS pool is to immediately pay out the expected values of outstanding scores.

Honestly though, my instincts tell me that most of the reward systems out there, even though hoppable, are perfectly fine to use (very few hop them correctly and make very little for their efforts anyway).

I don't think that's true. slush and SMPPS should be avoided. ESMPPS might sort of work out ok. There are very good hopping-proof methods so there's no justification to settle for anything less.

SMPSS - Simple Maximum PPS.

Shared Maximum PPS.

[teukon]

Thank you very much for this treasure trove of information!  As I said, I've not had time to look at these methods carefully but have been operating mainly on instinct.

I'm very interested to read about unit-PPLNS and why PPLNS for a fixed N is flawed.  As I'm writing this I'm realising that when I fixed N I somehow fixed difficulty in my mind too.  It seems that given the lag involved in PPLNS (when N is greater than 1) that miners could take advantage of changes in difficulty to hop the pool (I'm fairly sure this could be easily patched up).  If the flaw with non-unit PPLNS with fixed N is nothing to do with difficulty changes then that is fascinating and I hope to learn something new.

and who was I to tell people that there might be a hopping problem with a reward system designed to be hop-proof!

SMPPS was "designed to be hop-proof" by someone who hasn't got a clue what he's talking about. I've explained why it's broken until I went blue in the face ever since it was spawned.

Such a shame.  I wonder how they managed to convince themselves that their reward system was good.  I must admit that when I first read this reward system it seemed bad (just as with BitMinter and it's shift system).

at least with certain parameters fixed.

You can change the parameters, but you'll need to either modify the scores or using an invariant scoring in the first place.

Agreed.  I just meant to suggest that changing the parameters without due care can change the expected values of certain shares in a way might be exploited by hoppers.  I'm almost certain that some PPLNS pools are sufficiently careless about this.

Honestly though, my instincts tell me that most of the reward systems out there, even though hoppable, are perfectly fine to use (very few hop them correctly and make very little for their efforts anyway).

I don't think that's true. slush and SMPPS should be avoided. ESMPPS might sort of work out ok. There are very good hopping-proof methods so there's no justification to settle for anything less.

Hmm...  Again I'm operating on instinct so I'm happy to stand corrected by someone who's done the work.  When I read SMPPS it felt as though the severity of pool hopping would be significantly reduced (compared to proportional) but I could easily be wrong here.  "perfectly fine to use" is perhaps a little strong but I only mean to suggest that if you restrict yourself to only "correct" reward systems then you must reject almost every pool in existence.  When selecting a pool with low variance it is more helpful to know which of the pools have less of a hopping problem (factoring in: the percentage hoppers can ideally expect; preventative measures bought in by the pool such as hiding information and banning hoppers; and the difficulty in hopping it).  Personally I'm interested in using correct reward systems and don't mind high variance.

[teukon]

How have I gone this long here without discovering this information?

I've skimmed through some of the material you have and learnt a lot.  I'm disappointed that the difficulty changes are exactly the problem with PPLNS for fixed N as I was hoping for some cool new probability fact.  To me, the essential idea behind PPLNS is solid but one does have to be careful when changing parameters (here, both N and difficulty are parameters and I completely overlooked difficulty as a parameter in my original post ).  I'll read your method for coping with adjustments of these parameters but I'd guess there's only one way of doing it.  I have so little time these days for this kind of fun (1.25 am here) but reading your work is certainly not a waste of my time.

[Meni Rosenfeld]

and who was I to tell people that there might be a hopping problem with a reward system designed to be hop-proof!

SMPPS was "designed to be hop-proof" by someone who hasn't got a clue what he's talking about. I've explained why it's broken until I went blue in the face ever since it was spawned.

Such a shame.  I wonder how they managed to convince themselves that their reward system was good.  I must admit that when I first read this reward system it seemed bad

I think it has something to do with falling for the myth that score-based methods penalize intermittent miners, then starting a holy crusade stating that any method is good as long as it's not score-based. And, the vulnerabilities of SMPPS are different and subtler than of proportional.

(just as with BitMinter and it's shift system).

Actually a shift system can work. I've discussed here how to do it correctly.

Honestly though, my instincts tell me that most of the reward systems out there, even though hoppable, are perfectly fine to use (very few hop them correctly and make very little for their efforts anyway).

I don't think that's true. slush and SMPPS should be avoided. ESMPPS might sort of work out ok. There are very good hopping-proof methods so there's no justification to settle for anything less.

Hmm...  Again I'm operating on instinct so I'm happy to stand corrected by someone who's done the work.  When I read SMPPS it felt as though the severity of pool hopping would be significantly reduced (compared to proportional) but I could easily be wrong here. 

Oh, it's reduced all right, in a sense. The hopper can't get more than 0-fee PPS out of it, and the honest miners' losses depend on the long-term dynamics and usually come in the form of increased maturity time rather than reduced expectation.

"perfectly fine to use" is perhaps a little strong but I only mean to suggest that if you restrict yourself to only "correct" reward systems then you must reject almost every pool in existence.  When selecting a pool with low variance it is more helpful to know which of the pools have less of a hopping problem (factoring in: the percentage hoppers can ideally expect; preventative measures bought in by the pool such as hiding information and banning hoppers; and the difficulty in hopping it).  Personally I'm interested in using correct reward systems and don't mind high variance.

I'd say fixed-N PPLNS is good enough in practice. That opens up quite a few options, such as MMC. And you have some great, even if small, double-geom pools such as EclipseMC and yourbtc.

I'm disappointed that the difficulty changes are exactly the problem with PPLNS for fixed N as I was hoping for some cool new probability fact.

Sorry, making up new probability laws is beyond my powers.

[teukon]

I think it has something to do with falling for the myth that score-based methods penalize intermittent miners, then starting a holy crusade stating that any method is good as long as it's not score-based. And, the vulnerabilities of SMPPS are different and subtler than of proportional.

Yeah, what was up with that?  Why would weighing shares to given them all the same expected value hurt intermittent miners?  Where do people get these ideas from?

Actually a shift system can work. I've discussed here how to do it correctly.

I don't doubt that, it just felt at first glance that it was an attempt to reduce server workload.  The high complexity was very off putting and it felt like it would require some careful patching and scaling to be unhoppable.

I'd say fixed-N PPLNS is good enough in practice. That opens up quite a few options, such as MMC. And you have some great, even if small, double-geom pools such as EclipseMC and yourbtc.

I'll look into these.  Right now I would not switch from even variable N PPLNS to double-geom if it meant giving up merged mining (obviously as it's about 2.4 times a profitable right now) but if there is a merged mining double-geom pool out there (sensible parameters) with a solid server and a reputable and attentive operator then I'd send it half of my hashing power and have it as a backup for the other half.

[teukon]

Also, Meni, we are far off topic here.  This thread is supposed to be a flame war about the ethics of pool hopping, not a place for intelligent discussion on the mathematics of alternative reward systems.  We are very much in danger of being branded trolls!

[Matthew N. Wright]

[d.james]

LOL  
those kids are probably considered "unethical" to their neighbors.

[mu50stang]

I started a new proportional mining pool at http://www.tntmining.com/.  If anyone is interested in helping us out getting started.  I am offering a 3 btc bonus to who ever finds a block between now and Christmas.  12/25/11