Unwin is a risk management consultant who carries a torch for Bayesian inference, as against rival statistical methods. He illustrates Bayes' Theorem by taking on, not a murder, but the biggest test case of all, the existence of God. The plan is to start with complete uncertainty, which he chooses to quantify by assigning the existence and non-existence of God a 50 per cent starting likelihood each. Then he lists six facts that might bear on the matter, puts a numerical weighting on each, feeds the six numbers into the engine of Bayes' Theorem and sees what number pops out. The trouble is that (to repeat) the six weightings are not measured quantities but simply Stephen Unwin's own personal judgements, turned into numbers for the sake of the exercise. The six facts are:

For what it is worth (nothing, in my opinion), at the end of a ding-dong Bayesian race in which God surges ahead in the betting, then drops way back, then claws his way up to the 50 per cent mark from which he started, he finally ends up enjoying, in Unwin's estimation, a 67 per cent likelihood of existing. Unwin then decides that his Bayesian verdict of 67 per cent isn't high enough, so he takes the bizarre step of boosting it to 95 per cent by an emergency injection of 'faith'. It sounds like a joke, but that really is how he proceeds. I wish I could say how he justifies it, but there really is nothing to say. I have met this kind of absurdity elsewhere, when I have challenged religious but otherwise intelligent scientists to justify their belief, given their admission that there is no evidence: 'I admit that there's no evidence. There's a reason why it's called faith' (this last sentence uttered with almost truculent conviction, and no hint of apology or defensiveness).

Surprisingly, Unwin's list of six statements does not include the argument from design, nor any of Aquinas' five 'proofs', nor any of the various ontological arguments. He has no truck with them: they don't contribute even a minor fillip to his numerical estimate of God's likelihood. He discusses them and, as a good statistician, dismisses them as empty. I think this is to his credit, although his reason for discounting the design argument is different from mine. But the arguments that he does admit through his Bayesian door are, it seems to me, just as weak. That is only to say that the subjective likelihood weightings I would give to them are different from his, and who cares about subjective judgements anyway? He thinks the fact that we have a sense of right and wrong counts strongly in God's favour, whereas I don't see that it should really shift him, in either direction, from his initial prior expectation. Chapters 6 and 7 will show that there is no good case to be made for our possession of a sense of right and wrong having any clear connection with the existence of a supernatural deity. As in the case of our ability to appreciate a Beethoven quartet, our sense of goodness (though not necessarily our inducement to follow it) would be the way it is with a God and without a God.

On the other hand, Unwin thinks the existence of evil, especially natural catastrophes such as earthquakes and tsunamis, counts strongly against the likelihood that God exists. Here, Unwin's judgement is opposite to mine but goes along with many uncomfortable theologians. 'Theodicy' (the vindication of divine providence in the face of the existence of evil) keeps theologians awake at night. The authoritative Oxford Companion to Philosophy gives the problem of evil as 'the most powerful objection to traditional theism'. But it is an argument only against the existence of a good God. Goodness is no part of the definition of the God Hypothesis, merely a desirable add-on.

Admittedly, people of a theological bent are often chronically incapable of distinguishing what is true from what they'd like to be true. But, for a more sophisticated believer in some kind of supernatural intelligence, it is childishly easy to overcome the problem of evil. Simply postulate a nasty god — such as the one who stalks every page of the Old Testament. Or, if you don't like that, invent a separate evil god, call him Satan, and blame his cosmic battle against the good god for the evil in the world. Or — a more sophisticated solution — postulate a god with grander things to do than fuss about human distress. Or a god who is not indifferent to suffering but regards it as the price that has to be paid for free will in an orderly, lawful cosmos. Theologians can be found buying into all these rationalizations.

For these reasons, if I were redoing Unwin's Bayesian exercise, neither the problem of evil nor moral considerations in general would shift me far, one way or the other, from the null hypothesis (Unwin's 50 per cent). But I don't want to argue the point because, in any case, I can't get excited about personal opinions, whether Unwin's or mine.

There is a much more powerful argument, which does not depend upon subjective judgement, and it is the argument from improbability. It really does transport us dramatically away from 50 per cent agnosticism, far towards the extreme of theism in the view of many theists, far towards the extreme of atheism in my view. I have alluded to it several times already. The whole argument turns on the familiar question 'Who made God?', which most thinking people discover for themselves. A designer God cannot be used to explain organized complexity because any God capable of designing anything would have to be complex enough to demand the same kind of explanation in his own right. God presents an infinite regress from which he cannot help us to escape. This argument, as I shall show in the next chapter, demonstrates that God, though not technically disprovable, is very very improbable indeed.

The argument from improbability is the big one. In the traditional guise of the argument from design, it is easily today's most popular argument offered in favour of the existence of God and it is seen, by an amazingly large number of theists, as completely and utterly convincing. It is indeed a very strong and, I suspect, unanswerable argument — but in precisely the opposite direction from the theist's intention. The argument from improbability, properly deployed, comes close to proving that God does not exist. My name for the statistical demonstration that God almost certainly does not exist is the Ultimate Boeing 747 gambit.

The name comes from Fred Hoyle's amusing image of the Boeing 747 and the scrapyard. I am not sure whether Hoyle ever wrote it down himself, but it was attributed to him by his close colleague Chandra Wickramasinghe and is presumably authentic.⁵⁸ Hoyle said that the probability of life originating on Earth is no greater than the chance that a hurricane, sweeping through a scrapyard, would have the luck to assemble a Boeing 747. Others have borrowed the metaphor to refer to the later evolution of complex living bodies, where it has a spurious plausibility. The odds against assembling a fully functioning horse, beetle or ostrich by randomly shuffling its parts are up there in 747 territory. This, in a nutshell, is the creationist's favourite argument — an argument that could be made only by somebody who doesn't understand the first thing about natural selection: somebody who thinks natural selection is a theory of chance whereas — in the relevant sense of chance — it is the opposite.