packbat | How Likely is a Random Universe to Produce Your Twin?

Dan Shive (best known as the creator of El Goonish Shive) recently wrote a brief argument why alternate universes would probably not contain alternate "you"s. His argument looks correct, as far as it goes, but it is qualitative - lacking numerical estimates - and I don't see why it has to be. The data exists. Surely ballpark back-of-the-envelope numbers could be produced.

...but not trivially. Dan Shive's challenge can - and I think should - be broken down as follows.

First: how strictly identical must two individuals be to be called twins, for our purposes? There are two parallel criteria to be examined, each of which can be held to a stronger or weaker standard:

Genetic matching

When it comes to similarity on a genotypal scale, two defined benchmarks stand out for measurement purposes: perfect duplication and identical twin duplication. The former would have all three-billion-plus base pairs match; the latter would require that the match be no less near than is typical in monozygotic (colloquially: identical) twins. This latter standard would suggest measuring the difference at some specified stage of development, perhaps birth.

Personality matching

In many ways, the genetic criteron is dissatisfying, primarily because one does not think of oneself as one's genes. Identical twins are different people, for example. This criterion may be held to three levels of severity:

Indistinguishable - trading places would result in no apparent change with respect to behavior. This is almost certainly impossible, and not very good SF, either.
Birds of a feather - each would predictably have similar reactions to the same items. For example, if you showed the same movie to each, one may be nearly certain that either both would like it or both would dislike it.
Dissimilar - the most common kind in SF: the kind associated with evil twins, heroic alternates, etcetera. No restriction on personality.

Second: How divergent are the universes that created these individuals? From least to most:

Identical up to moment of conception: how much chance is involved in a particular human fertilization in (a) selection of sperm and (b) mutation of genome during development of embryo?
Identical up through moment of mate selection: the previous, plus: how much chance is involved in (c) the genetics of the egg and (d) the commitment to a particular relationship (or cheating thereon)? The latter leads on to:
Identical up through birth of parent generation: (e) how much chance is involved in mate selection? (f) how probable is the chance selection of identical child genes from different parents (cf human genetic variation)? (g) how large a gap can be spanned by mutation, and how probable are different degrees of mutation?
Diverging before birth of one or more parents: (h) how quickly do distinctive genomes spread through population, and how likely are they to meet again in different combinations in later generations?

Now, I lack the knowledge of biology to, first, nail down these questions to their most correct forms, and second, assign probability estimates to relevant steps in the chain. But the most superficial examination of the situation seems to suggest at least one thing: any alternate universe measurably diverging a significant period before the birth of an individual is vanishingly likely to contain a copy of that individual. Which, of course, is what Dan Shive has pointed out.

And, as an obvious consequence of this, even if such a universe contained a duplicate of yourself, it would still be vanishingly unlikely for it to contain duplicates of anyone not your direct descendant. (Which would make for a heck of a paternity test, I have to tell you!)