gsgs wrote on 07/05/19 at 04:39:16:
at 2000 Elo you should go for "traps" . Play some unusual sideline
which you prepared and hope that the opponent won't find
the best moves.
Hmm, there should be programs which search for such lines
automatically ...
Quite a "provocative" post.
A. On a higher level of abstraction, we all hope that our opponent will make a mistake, otherwise there is no way to win. If I have two candidates of the same quality, with one containing a trap and the other not, I know which one I would choose. But playing "for" traps has a different connotation, meaning to choose an
inferior move solely because it contains a trap. In that case I agree with IsaVulpes, it's bad for your chess.
B. Nothing wrong with sidelines
per se, but the key word is
prepared. If you have some idea what to do when your opponent answers in the theoretically approved way, then go for it. If the only thing you know is the trap, and against any other play you will fold like a cheap tent, it's time to go back to school.
C. If a player studied only traps, how strong could they become? A key problem is that traps typically work only once per opponent. (I did know a player who fell for the
exact same trap
four times, but that's a different story.) Based on a tiny sample of players at the club, my guess is typically about 1700. There was one trappy player who peaked at 1900, but that was when he temporarily started playing theory. Later he decided theory was too much work, and fell back to 1800.
D. A program that searched for traps would be a good way around the once per victim problem. But could a program that found a trap tell you how likely it is that your human opponent might fall into it? It would have to be a sophisticated program, maybe with some learning feature based on human data. The books by Mueller and Knaak,
222 Opening Traps (2 vols, 2007-2008), are based on human games, and the authors indicate how many games they found in the database. If you needed a trap today, without waiting for a futuristic program, that would be a good place to start.