"In reality, humans display a systematic bias towards cooperative behavior in this and similar games, much more so than predicted by simple models of "rational" self-interested action."
That doesn't apply in non-cooperative games (which is what masternodes create).
Then why did you bring it up in the first place if it doesn't apply?
You misunderstand. The Prisoner's Dilemma DOES apply in both cooperative and non-cooperative games, the bias towards cooperative behaviour DOES NOT and cannot apply in non-cooperative games.
In other words: the Bitcoin protocol is a cooperative game. If every node tried to cheat there'd be chaos. Everyone on the network runs by the same rules, for fear of being rejected by the network if they do not.
However, Bitcoin mining is a non-cooperative game. Mining pools can and do DDoS each other in order to get more miners mining on their pool. We know this happens, and in the altcoin world small pools shut down regularly because they can't afford the upkeep when they are regularly DDoS'd. The two things that prevent this system from imploding are: 1. the community freaks out whenever a pool is above 30%, and so a pool that is a bad actor is forced to setup or buy out other pools to mask their behaviour; 2. pool fees are so tiny (some of them have no fee) that it's not that financially beneficial.
For another example of mining being a non-cooperative game prone to cheating, see
Errata Security's post that applies another aspect of game theory to show how miners can cheat, or see
Ittay Eyal and Emin Sirer's paper on Selfish Mining.
Also, to demonstrate I'm not proposing a merely theoretical attack (and this applies to Darkcoin's masternodes too, despite Brilliantrocket thinking that DDoS attacks are "March's problem"), I'll quote from that ErrataSec blog post:
The operators of one pool might try to DDoS the URLs of other pools. This would prevent those pools from sending out work to their members, lowering their hash rate, and making it more likely that non-DDoSed pools would win the lottery. People do this, and this has lead to arms races among pools. They are under frequent DDoS attacks, and are frequently having to upgrade their server infrastructure to cope with the attacks. Overall, it’s been a nuisance and hasn’t dramatically enabled anybody to significantly win at the mining game.
In Darkcoin - and again, this is a general criticism of the broken architecture and not FUD or a slight against anyone - masternodes are in a non-cooperative game. They have no obligation to cooperate with anyone, as long as they are publicly perceived to be following the mathematical constraints in the protocol. The huge reward they are paid on a daily basis (
Brilliantrocket states it as 576 coins per day for all masternodes) makes attacking the competitors feasible and desirable.
Practically speaking, then, the question is: is there sufficient reason for a masternode to cheat in this manner? Well, let us say that Kim owns 3 masternodes, and is thus stuck wanting to reclaim 30 000 Darkcoin. Network is operating efficiently, Darkcoin is popular, and there are 2 000 masternodes (this was mentioned to me as the proposed maximum, I stand to be corrected). Let us assume (as it appears to be right now) that the block reward is based on minsubsidy and not on any sort of declining emission, and thus for the purposes of our example 576 Darkcoins a day is the total masternode reward for the period. Thus, Kim receives 0.15% of that reward, a total of 0.864 Darkcoins a day. For him to recoup his 3000 Darkcoin it will take him 3472 days, or about 9.5 years. I'm aware that he still has the original 3000, but that's a sunk cost - untouchable - and he thus needs to recoup it in order to operate as a profitable business. That it's an asset that doesn't in and of itself depreciate is not a reflection on profitability.
Now I know that the knee-jerk reaction is "but those 3000 Darkcoin only cost him $9000 at the time, he's recouped that already!" Of course, this is irrelevant, as Kim would have been better off just holding his 3000 Darkcoins instead of throwing them at a masternode and hoping to recoup it. Thus, Kim is motivated - eminently so - to destroy the other masternodes. If he attacks a single masternode long enough for them to give up and dump their coins, the price will merely take a small hit and recover, but he will increase his daily reward.
The reason this system (as defined in the example) will approach Nash equilibrium is because all of the masternode operators know that they have to keep the number of masternodes under 100 to have any profitability (because they will then be able to roll-over their 1k investment every 6 months), and thus there is motivation to be a bad actor. More importantly, if you get on the forums and claim that you are also being DDoS'd (maybe turn your masternode off for a few hours every week) you can play the victim, and nobody will know that you're a bad actor. A handful of these bad actors is enough to destroy the system from within.
Edit: was off by a factor of 10, as I thought it required 10k DRK not 1k DRK