The default cost values are those used for the Formula 1 combat bot league. This is why they are labeled F1 Default. I have no idea where the specific values came from or why they were chosen as they were. Before my time.
While we are on the subject, I personally do not favor costs that are directly associated with DNA length, DNA instruction content or specific, lowlevel DNA command execution, at least not for those commands not directly related to a morphological action like all of the current "DNA Command Costs". Personally, I never use these. If our goal is to evolve complex behaviour and reward morphological effeciency, then I view DNA command costs as counter productive towards that end in that the selective pressures they create work against the formation of complex cognition. I much prefer costs on morphology and on morphological actions - taxing the phenotype, not the genotype and not the neurology embodied in the DNA makes more sense to me.
But surely there should be an execution cost you may say. Thinking longer or more or using more inputs or more calculations to decide on which way to move or face or when to shoot for example, surely that should be taxed more, even if the end morphological result is simply to move or shoot the same as a bot that spent no instructions thinking. Again, I would disagree. I want to encourage complex behavior, not select against it.
I am not opposed to say, an abstracted execution cost which taxes cognitive ineffeciency, say a cost on number of genes fired each cylce for example. After all, there is a biological carrying cost to having more neurons or bigger brains, but that cost should be small relative to morphological actions such as moving or shooting. I want organisms to be able to evolve a lot of DNA that only rarely gets used, say for mate selection at reproduction time, and not be taxed for simply carrying it around. The advantages of being smart should far outwiegh the disadvantages.