 |
November 10, 2013, 07:00:41 PM
Last edit: November 10, 2013, 09:15:08 PM by cunicula |
|
NOTE: This is about the implications of game theory, not adjustments to how the network operates.
Hypothetical (bad) Strategy for dealing with selfish mining:
1) Say that a new pool can form at the drop of a hat. The new pool offers big rewards. All the miners are greedy and they jump to this pool. The pool does selfish mining to be able to afford the big rewards.
2) Gavin asks everyone to switch to a good pool and everyone listens.
3) The new pool closes down.
Why is this bad? Because as soon as (3) occurs we can go back to (1). As long as people are tempted by higher rewards the selfish pools can keep appearing and shutting down. The people who mined at the bad pool will get rewarded and be encouraged to repeat their behavior.
Hypothetical (good) Strategy for dealing with selfish mining:
1) Say that a new pool can form at the drop of a hat. The new pool offers big rewards. All the miners are greedy and they jump to this pool. The pool does selfish mining to be able to afford the big rewards.
2) Gavin picks a pool at random and asks everyone to switch to this pool. Gavin tells this pool to do selfish mining for 1 week. All other pools will be forced to close. The selfish pool can take a higher than normal % because it is a monopoly. (miners don't like this but they don't have a choice; the other selfish pool would have exploited them anyway)
3) All the miners are sad because Gavin had a pool exploit them for one week. (But we still like him because he is a strong leader)
Why is this actually better? Because it imposes a significant penalty on miners' bad behavior. 25% of miners were bad selfish guys. It is infeasible to identify them one by one. However, they can be punished through a collective penalty on all miners.
Now miners realize that if 25% of people are bad, everyone gets punished. Everyone is made worse off, including the bad guys. The bad miners have a strong incentive to never use untrustworthy pools again. They will only mine with reputable players. If not they will risk collective punishment.
The latter strategy is called tit-for-tat. If Gavin to credibly announce a plan like this, then miners would never willingly participate in selfish mining. So even though the outcome seems awful, we would never see it come to pass. This is how nuclear arsenals maintain peace between superpowers. It can work for the more humble purpose of selfish mining as well. (we also see it used in households; your boys are fighting and you don't know who's to blame -> send them both to their rooms)
(You will say that the bad miner will not consider his effect on the outcome because his marginal effect is too small, but this is wrong. The key insight is that under this arrangement, the bad miner wants the attacking pool to fail. He would like to contribute his small marginal effect to making this happen. If the attacking pool is successful, then the bad miner is strictly worse off. There is no private gain from attacking. If the bad miners see a mounting attack, they will try and prevent it just like good miners will. If the attacking pool fails, then the bad miners will end up with more income. That is all that matters. The bad miners, just like the good miners, have an incentive to make sure the attack never happens. Their pocketbook depends on it.)
The existence of strategies like tit-for-tat is why that stupid selfish mining paper is wrong. The authors did not consider the possibility that miners could condition their behavior choices based on past events. (I'm not sure how they can call what they are doing game theory without considering this, but LOL that is what they did.)
Anyways, if we commit to using violence to defend ourselves, then we can keep the peace.
|
