Pages:
Author

Topic: Optimal pool abuse strategy. Proofs and countermeasures - page 2. (Read 30459 times)

legendary
Activity: 2058
Merit: 1452
I REPEAT, READ THE PAPER AGAIN! especially the part with the funny equations. don't just go "wtf", and skip it.
Yeah, what about them. They translate the long text into a mathematical formula. Both are theoretical models that seem to lack a certain factor that happens in practice...

This is why I ask, how exactly can one cheat in practice.
same question i asked.
https://forum.bitcoin.org/index.php?topic=9928.0
moral of the story? lrn2 search
legendary
Activity: 1442
Merit: 1005
I REPEAT, READ THE PAPER AGAIN! especially the part with the funny equations. don't just go "wtf", and skip it.
Yeah, what about them. They translate the long text into a mathematical formula. Both are theoretical models that seem to lack a certain factor that happens in practice...

This is why I ask, how exactly can one cheat in practice.
legendary
Activity: 2058
Merit: 1452
I just red the paper AGAIN!
lol


How do you get a higher reward? If the block required 200 shares and there are 2 people in the pool then you get:
- cheater does 43 shares then leaves, the non-cheater has to do 157 shares, the block is found, each gets proportional payment, cheater has a chance of 21.5% of finding the block, the non-cheater has a chance of 78.5%. The cheater gets 21.5% of the reward, the non-cheater gets 78.5%.
- non-cheater A can do 43 difficulty 1 hashes in 10 minutes, non-cheater B can do 157. The block is found, each gets a proportional payment, just like above, the total contribution is shares/time, in both scenarios, and the payoff is the same
I REPEAT, READ THE PAPER AGAIN! especially the part with the funny equations. don't just go "wtf", and skip it.
legendary
Activity: 1442
Merit: 1005
I just red the paper AGAIN!

I still don't see how in practice you could abuse a pool. If you are contributing 43% of an average round period, you get 43% of that round's shares and a 43% chance of finding the block.

How do you get a higher reward? If the block required 200 shares and there are 2 people in the pool then you get:
- cheater does 43 shares then leaves, the non-cheater has to do 157 shares, the block is found, each gets proportional payment, cheater has a chance of 21.5% of finding the block, the non-cheater has a chance of 78.5%. The cheater gets 21.5% of the reward, the non-cheater gets 78.5%.
- non-cheater A can do 43 difficulty 1 hashes in 10 minutes, non-cheater B can do 157. The block is found, each gets a proportional payment, just like above, the total contribution is shares/time, in both scenarios, and the payoff is the same
donator
Activity: 2058
Merit: 1054
43.5% of expected time to complete a block with the current hash rate.
Actually that's just an approximation, which doesn't take into account changes in hashrate etc.
legendary
Activity: 2618
Merit: 1007
Just read the paper!

43.5% of expected time to complete a block with the current hash rate.
donator
Activity: 2058
Merit: 1054
No, he will always mine for the pool with the least shares in the current round.
Ok. That's like having two miners and increasing load on the one with less submissions so they equal up... I don't see the bad thing in this. If he pushes to a single pool all his shares, or pushes 50% to a pool and another 50% to another pool, his reward should be the same in all cases.
The less shares in the current round in a proportional pool, the greater the expected reward per share submitted. Consider also the examples I gave in this comment. The bad thing is that hoppers get too much reward for their contribution to the pool (and to the Bitcoin network at large), while honest miners get too little.

If all proportional pools are >43.5% he will mine for a score-based pool.
Proportional pools are > 43.5% what? I don't understand what is the percentage of.
This means that for every proportional pool, the number of shares in its current round is more than 43.5% of the current difficulty.
legendary
Activity: 1442
Merit: 1005
No, he will always mine for the pool with the least shares in the current round.
Ok. That's like having two miners and increasing load on the one with less submissions so they equal up... I don't see the bad thing in this. If he pushes to a single pool all his shares, or pushes 50% to a pool and another 50% to another pool, his reward should be the same in all cases.

If all proportional pools are >43.5% he will mine for a score-based pool.
Proportional pools are > 43.5% what? I don't understand what is the percentage of.
donator
Activity: 2058
Merit: 1054
Given 2 blocks, and an ability to generate 100 shares per block, and two pools, a pool hopper will try to jump the pool after each share, right?
No, he will always mine for the pool with the least shares in the current round. If all proportional pools are >43.5% he will mine for a score-based pool.
legendary
Activity: 1442
Merit: 1005
HELP, I don't see the proof.

Given 2 blocks, and an ability to generate 100 shares per block, and two pools, a pool hopper will try to jump the pool after each share, right? He will effectively generate 200 shares, in the two pools, each block he will generate 100 shares. Just like he would use a single pool. He will spend the whole time between two shares, in one of the pools, at the reward of a share.

Number of times waiting for shares to come up and compute hashes = Number of shares received in pools

Why don't I see the problem?
legendary
Activity: 2618
Merit: 1007
PPS has the downside that you pay for stale shares of others - even if the pool operator keeps as little as possible, this still means your income will be lower than in a pool with a different model. A proportional pool with 0 hoppers might be ideal "fairness wise", a score based pool raises variance and shifts some risks but this canbe tuned.

PPS however will very likely be worse, since it is not ver predictable for smaller pools how much they would really earn during 2016 Blocks to calculate correct values. This means they either have to have higher fees on PPS (deepbit for example has (or had, as they are down atm)) 10%(!) extra fee on PPS payouts.
full member
Activity: 238
Merit: 100
This to me just sounds messy. Why not just keep it simple and stay with PPS?? Then no weird stuff..  is there anyway to cheat with the Pay Per Share System? Not that I can think of anything. Skipping a share? Then you get a big pool of stales.. no way to get around that one..

Pay per share (PPS) is very dangerous for pool operators, especially the small ones. In PPS, the pool operators have to pay for bad lack streaks with their own pockets. It can make them go bankrupt because 10-20% bad luck streaks are not uncommon. Moreover, there can be malicious users who do not submit winning shares. With difficulty over 400,000, the loss for the user is negligible, for pool operator it is huge.
member
Activity: 83
Merit: 10
This to me just sounds messy. Why not just keep it simple and stay with PPS?? Then no weird stuff..  is there anyway to cheat with the Pay Per Share System? Not that I can think of anything. Skipping a share? Then you get a big pool of stales.. no way to get around that one..

Just keep it simple.. Just wasting time and energy. IMHO
hero member
Activity: 793
Merit: 1026
luv2drnkbr, I agree with you. Raulo's formulation of "not having contributed if you leave the pool" is incorrect. I have never raised my dissent because the analysis of pool hopping does not depend on it as a premise. It is still a way to benefit on the expense of others in proportional pools, which is unsustainable if done by everyone.

The correct way to view this is to ignore the past, look at the present and see how it affects the future. If I submit a share now, what is the expected benefit to the pool? What is my expected payout? Are they the same? In my scoring method, the contribution and reward are always equal (up to fees). In a proportional pool, the reward is higher than the contribution in young rounds, and lower in old rounds.

On a proportional pool, suppose the current round already has shares twice the difficulty. If you submit a share, what is your expected contribution? You have a probability p=1/difficulty to find a valid block with reward B, so it's pB. What is your expected payout for this share? No matter what happens, you will get less than (pB/2) reward, so the expectation is less than pB/2. So you're better off mining solo which has pB expectation per share.

What if the round is very fresh, say, no shares at all? Your contribution to the pool is still pB. But now you have a probability of p to get the whole reward B to yourself, this contingency alone adds pB to the expectation; probability p(1-p) of getting half the block - since 1-p is quite small, this already puts you at almost 1.5pB; and with probability p(1-p)^2 you get a third, and so on. So your expected payout is roughly log(1/p)pB, which at current difficulty is 13pB. So shares submitted early on will give you a very large expected payout, much more than your fair share for your contribution to the pool.

I did not understand that proportional pools existed.  Thank you for that explanation Holy-Fire.
donator
Activity: 2058
Merit: 1054
luv2drnkbr, I agree with you. Raulo's formulation of "not having contributed if you leave the pool" is incorrect. I have never raised my dissent because the analysis of pool hopping does not depend on it as a premise. It is still a way to benefit on the expense of others in proportional pools, which is unsustainable if done by everyone.

The correct way to view this is to ignore the past, look at the present and see how it affects the future. If I submit a share now, what is the expected benefit to the pool? What is my expected payout? Are they the same? In my scoring method, the contribution and reward are always equal (up to fees). In a proportional pool, the reward is higher than the contribution in young rounds, and lower in old rounds.

On a proportional pool, suppose the current round already has shares twice the difficulty. If you submit a share, what is your expected contribution? You have a probability p=1/difficulty to find a valid block with reward B, so it's pB. What is your expected payout for this share? No matter what happens, you will get less than (pB/2) reward, so the expectation is less than pB/2. So you're better off mining solo which has pB expectation per share.

What if the round is very fresh, say, no shares at all? Your contribution to the pool is still pB. But now you have a probability of p to get the whole reward B to yourself, this contingency alone adds pB to the expectation; probability p(1-p) of getting half the block - since 1-p is quite small, this already puts you at almost 1.5pB; and with probability p(1-p)^2 you get a third, and so on. So your expected payout is roughly log(1/p)pB, which at current difficulty is 13pB. So shares submitted early on will give you a very large expected payout, much more than your fair share for your contribution to the pool.
hero member
Activity: 793
Merit: 1026
I'm not math-savvy enough to get the whole thing, but I am a gambler and I understand expectation, and the basic premise is false:

Quote
If one stops mining for a pool that has not found a share, one still gets a payment [...] even though now you contribute nothing to the pool chance of success. This enables cheating.

It doesn't matter if you contributed to the pool and a hash wasn't found.  It still took your previous work as well as the work of everybody else to find the hash.  You DID contribute to finding that solution.  And you are getting paid for the amount of work that you did.  If you then go and continue working on your own, that doesn't matter.

If I am playing in a re-buy poker tournament on a team's bankroll (a team I'm a part of), and if my win would get chopped with the team, but then I bust out, and the team doesn't want to re-buy me back into it, I can still go into my pocket and buy myself in on my own money.

If somebody on the team wins the tournament, I still get part of it because I AM part of that team, and if I cash on my own money with my re-buy, I get all of it because the team didn't pay for that.

That's not unfair to anybody.  I was working for the team, and if the team wins money, I should get paid for my role in the team effort.  If I then go off and put the effort in by myself and win, then I should get all of it.  There's no conflict there.

The basic premise of your paper is false.  The EV is the same for the miner--it's merely a function of his total effort put in, pool or no pool.

What you are describing is simply a "gambling system" that doesn't change the house edge.  You've stumbled upon the Martingale system for blackjack and now think you have an edge.

You don't.  Nothing has changed.  The premise is false.

I will demonstrate what I'm saying must be true:  If everybody in the pool mines for a bit then goes solo, the pool has done some work but now will never find a solution.  However, one of the solo miners will (assuming no outside people in this example).  The chance of any individual miner finding a winner is exactly the same as it was when he was part of a pool, it's just that when he was in a pool, he would have had to give most of his winnings away, but in return he got part of the winnings if somebody else won.  By mining solo, he has merely given himself a bigger potential payout but at a cost of it being less likely.  But his average monetary gain over the course of his lifetime will be the same, whether or not it's paid in small increments or in random 50 btc spurts.  Thus, if the two are equal, it cannot be true that switching to one or the other at some magical point in time will somehow yield a higher profit.

100% of $5 is mathematically equivalent to 10% of $50.  The only difference is variance.

The premise is false!
donator
Activity: 2058
Merit: 1054
It looks like the paper is unavailable from http://bitcoin.atspace.com/poolcheating.pdf . Where can I get it now?

The web hosting I use, started to become pretty unreliable recently. I need to find something different. But if you PM me your email, I can email the paper to you.
Thanks! This was obviated by slush's gracious hosting.

Thanks! Saved to disk this time.
full member
Activity: 238
Merit: 100
It looks like the paper is unavailable from http://bitcoin.atspace.com/poolcheating.pdf . Where can I get it now?

The web hosting I use, started to become pretty unreliable recently. I need to find something different. But if you PM me your email, I can email the paper to you.
donator
Activity: 2058
Merit: 1054
It looks like the paper is unavailable from http://bitcoin.atspace.com/poolcheating.pdf . Where can I get it now?
Pages:
Jump to: