Oops. I meant that for all possible biases torwards heads, not always or never. Except that only has 3/4 success.

I'm just saying comparing it to a human is unfair. Humans use feeble subsets of the universal prior that are easy to imitate.

After observing the coin tosses for a while, a typical intelligent person, just applying common sense, would notice that 90% of the tosses come up heads, and infer that perhaps all the coins are biased towards heads. They would become more certain of this with time, and adjust their answers accordingly.

A bayesian could do better if you allowedtot to remember things.

coin lands on heads: P(headsbiased|heads) = .9.5/.5 =.9 P(tailsbiased|heads) = .1.5/.5=.1 another heads: P(headsbiased|heads)=.9.9/(.1.1+.9*.9)=.987 P(tailsbiased|heads)=.0121