It depends on what sort of informativeness we want to achieve. What is the “ideal” distribution of strategies in the pool, not biased by popular vote? If there was a simple metric for strategy complexity, an analogue of Kolmogorov prior might work. But the actual pool was very far from that. Having more DefectBots and TitForTats would actually move us in that direction.
Having more DefectBots and TitForTats would actually move us in that direction.
Only by coincidence. It’s not like authors of simple strategies are systematically more likely than authors of complex strategies to try to cheat the metarules.
It depends on what sort of informativeness we want to achieve. What is the “ideal” distribution of strategies in the pool, not biased by popular vote? If there was a simple metric for strategy complexity, an analogue of Kolmogorov prior might work. But the actual pool was very far from that. Having more DefectBots and TitForTats would actually move us in that direction.
Only by coincidence. It’s not like authors of simple strategies are systematically more likely than authors of complex strategies to try to cheat the metarules.