<>then, another thought: those are computer programs. that means, that they'll do the same >things with the same input (random functions excluded). I think that even a small difference >in computer strength would be enough to provide the sort of results you show. for example: >say that chess1.0 has an elo of 1800, chess2.0 an elo of 1850. among humans, you'd expect >results like 12-8; with computers results such as 20-0 aren't surprising.

Sorry but all these is simple nonsense!
If Progr-1 is only a little stronger than Progr-2 then a 20-0 is just impossible. And actually the results are just comparable with these of humans.
To be more specific:

Let's suppose that we have a database of computer Chess programs and a database of human Chess players. And after many games between the members of the 1st database against each other and the members of the 2nd database against each other, we establish an ELO classification for each database. And then we choose 2 programs(P1,P2) from the 1st base with ELO's A and B. And then we choose 2 humans(H1,H2) from the 2nd with ELO's A and B too.
Then according with ELO system definition, if a match between P1 and P2 has x% probability to end +w1 -w2 =w3, then a match between H1 and H2 has a x% probability also, to end +w1 -w2 =w3. The probability is the same or to reverse it: If the expected score between H1 and H2 is something then we have EXACTLY the same for P1 and P2 also. There is no reason that a small difference in computer's strength would provide different result than that of humans.