Nearly Precise Probability Circuit

Wow, I can't believe I finished a build. It's my first time.
Each output of the circuit in the picture has about 1/20 chance of being on. Only one output will be on at a time, for some unknown reason. Each output has about a 1/20 chance of being on because there are ten torches, and when a torch activates it pulses off and on.
There are two reasons each output doesn't have a precisely 1/20 chance is that there can be no torches on. This problem becomes increasingly small with more torches.
I'm going to use this in the RedSlime, to replace connection permanence. Rather than keeping track of the permanence of connections based on how often the two neurons fire in unison, every time the two neurons fire in unison there will be a chance of connecting, and every time the source neuron fires and the other doesn't, there will a chance of disconnecting.
It is important that I know the odds, because I will need to chose a permanence value and then simulate that by selecting the odds of connecting and odds of disconnecting.