there is a typo in the moves-whites 21st was red1 not rf1-so the bb5 option doesnt really exist

secondly international ratings didnt exist at all in 1966 as far as i know

i took two minutes because i had to play two matches that evening-my opponent arrived as late as possible for this game so i only had 2 minutes to make all my moves -run to the railway station and then catch a train to the next match -otherwise i wd have arrived too late for that one

of course under modern rules my opponent here wd have lost on forfeit being just a second late instead of 58 minutes late as he was

the rules of this competiton allowed for adjudication after completion of 24 moves-hence whites apparent resignation when there is still a little fight left-white in fact lost on adjudication.

But especially thanks for adding the colour to this game, through the story you tell, which enables us to appreciate this game beyond just the moves.

9.a4 is the most usual move. White knows he'll probably have to play it at some point, and this is a good moment. The bishop will have to retreat eventually, but it's better to wait until Black spends some time to persuade it to leave. Also, delaying the bishop retreat allows White to assess how the position is shaping up before deciding which square will be best for the bishop.

Then White makes some tactical boo-boos and loses quickly. Very quickly.

An interesting point. Of course two pieces aren't *always* worth more than a ♖; under the right circumstances even a ♕ may not be worth more than a ♖.

What I wonder is, how much variation is there? To put the question in precise terms, if we looked at a sample of 100 master games and used a standard engine -- say, Rybka or Fritz at 20 plies -- to calculate the change in the value of a position before and after the swap of ♗+♘ for a ♖ whenever this is possible, what would the mean change in the evaluation be? More to the point, what would the dispersion about the mean be, i.e., the root mean square of the variation?

Presumably, since this is thought to be a bad trade most of the time, the mean change would be negative. It *should* be equal in magnitude to the difference in values of the weights the programmers assign to the pieces, i.e., 1.0 or close to it. If the rms of the variation is much less than that, then the rule of thumb is highly reliable. But if the opposite turned out to be true, it would imply that the "bad" trade is often good.

However, I'd think more useful than a statistical study might be a discussion in a middle game strategy/tactics book by a GM -- indicating positions (such as this game) which favor the side exchanging the rook for the two minor pieces, or favor the other side, or is otherwise equal. It might also make for an interesting game collection.