0
Hey folks, just started an iterated prisoner's dilemma tournament on campus. We're coding bots to compete, and I want to hear about the cool strategies you all might use or have heard of. I'm tinkering with a 'Grudger' approach - cooperating until the other defects. After that, it's defect all the way. Will this just make my bot a pushover, or is it viable? Alternative strategy suggestions welcome!
Submitted 6 months, 1 week ago by IterativePrisoner
0
0
Consider something with a longer memory than Grudger. Analyze patterns in your opponents' play, predict their next move, and then backstab 'em when they least expect it. Like a 'Faux-Friend' bot - plays nice, lulls them into a false sense of security, then BAM! Defect when the stakes are high.
0
0
Hey! Advanced strats could involve machine learning techniques where your bot learns the most effective strategy against each opponent over time. The computational expense might be a factor, but if you can pull it off, you'd potentially outsmart those bots using static strategies.
0
Your Grudger approach, that's like holding a grudge right? I dig it. Once they cross you, they're dead to your bot. But here's a thought, maybe instead of defecting forever, hold the grudge for a set number of turns. Give 'em a taste of their own medicine, then throw in a random act of cooperation to mix things up. Keep them on their toes.
0
0
Ever heard of the 'Pavlov' strategy? It flips based on win or loss, not the opponent's last move. Cooperates or defects and then simply repeats last choice if it won, else it changes. Add some randomness to the mix, and it'll be less predictable. Could mess with those bots relying on pattern recognition. Make 'em sweat, ya know?
0
imo, you gotta try a tit-for-tat strategy! It's simple: cooperate on the first move, then mirror whatever the other bot did last time. If they defect, you defect right back next round. If they play nice, you play nice. It's fair and it tends to do pretty well in tournaments from what I've heard.