Now, in the expectation of obtaining the simplest possible self-reproducing organizm, let the reactions take place a billion times per second in each cube. And let's further assume that the reactions have been going on for 4.5 billion years, the estimated age of the earth.

Scientists Fred Hoyle and Chandra Wickramasinghe has estimated that the chance of obtaining the simplest self-reproducing system by random combination of molecules is at best somewhere in the neighborhood of 1 in 10 to the fortieth power attempts. But if out of extreme genorosity we reduce the required number of proteins from 2,000 to only 100, then the probability is still 1 in 10^2,000 [1 in 10 to the 2,000 power].

Now, if you add up all the possible attempted billion-per-second combinations in our hypothetical primordial soup, you wind up with only 10 to the 74th power throws of the chemical dice. that means the odds of getting the required self-producing system out of our soup would be 1 in 10¹⁹²⁶ [10 to the 1,926th power]. We wouldn't expect that to happen in the entire course of the earth's history!

Of course, a diehard gambler might say it's highly unlikely but it just could happen by chance. But this is a completely meaningless use of the word chance. In order for a statement about an event with a nonzero probability of happening to be meaningful, we would have to observe enough repetitions of the event to establish a statistical pattern. Only this would allow us to say, "This event has probability p of happening."

For example, we say that when we toss a coin there is one chance in two that it will turn up heads. This probability is established by examining the behavior of the coin over several hundred trials. Now, if you have an event with a probability of one in a million, it would take hundreds of millions of trials to establish this. And if the event has an estimated probability of 1 in 10^2,000, you would need many times that number of trials. The basic point is this: What is meant by a probability of 1 in 10^2,000 is that a certain statistical pattern corresponding to this figure will be observed over the required vast number of trials. If there is no possibility of performing these trials (as is certainly the case here), then there is no meaning to saying an event happens with that very small probability.