Me! Ooh me! (Insistently raises hand hoping the teacher will notice).
A slightly better analogy would be when the light have a higher probability of being green than red, say that after becoming green they have a somewhat smaller chance of changing to red each second, than they do of changing to green if they are currently red. If all start red, more and more of them become green until finally Prob(being green) x Prob(this copy changes to red next second | it is now green) = Prob(being red) x Prob(this copy changes to green next second | it is now red) where “|” means “given”. And then the numbers fluctuate around an equilibrium which has more greens than reds.
At the expected equilibrium frequencies of Red and Green, the rates of green mutations per red copy will not necessarily equal the rates of red mutations per green copy, but the total numbers of to-green mutations will be expected to equal the total numbers of to-red mutations. More properly, if we have natural selection slightly favoring green, the events should be not mutations but substitutions, the probability of which is the product of the probability of a mutation times the probability that it substitutes.