nobody5050
commented
3 years ago There’s a few things that make this unlikely. First off an individual person can only submit one py file to the contest ever.
Even if they did it with their friends, Cary is going to go over every file, so one which just listens for a list of moves is going to be easily detected.
Even still with the number of agents that will be playing in the ffa it would be hard to have a kingmaker be successful since the king will have to be a good strategy outright
So first of all to define a kingmaker strategy: If a group of users collaborates and implements agents that are able to recognize each other this could be very detrimental as it could sabotage the tournament.
A very simple strategy for this would be to submit multiple agents, most of which are kingmakers, aswell as one king.
The kingmakers are able to recognize the king and the king is able to recognize the kingmakers, for example by using a sequence of arbitrary seeming cooperations/ defections
If the kingmaker doesnt recognize the king, it will refuse to cooperate, sabotaging the other strategies.
If the kingmaker does recognize the king he will always cooperate and let the king take advantage to maximize points for the king.
The individual kingmakers wouldnt gain anything, but the king would wildly outperform the other strategies, given that enough kingmakers are part of the pool, which would defeat the point of the experiment in my opinion.