Author: Robert Hyatt
Date: 11:39:57 10/06/03
Go up one level in this thread
On October 05, 2003 at 13:43:33, Ricardo Gibert wrote: >There was a discussion about the value of playing oddball lines by amateur >programs to avoid the effective book lines of the stronger programs i.e. >http://www.talkchess.com/forums/1/message.html?319397. I thought about this a >little and a trick occurred to me that should work on most programs e.g. >Ruffian. > >The easiest way to explain it is to jump into some examples: > 1. e3 e5 2. e4 > 1. c3 e5 2. c4 > 1. d3 d5 2. d4 > 1. d3 e5 2. d4 > >What is this? Isn't White just dumping a tempo? Yes, but the idea is for white >to get the computer opponent out of book while retaining the advantage of >hundreds of years of opening theory for your own program! > >It can be carried out with 2 possible motives in mind: >1- Reach a playable middlegame with a huge time advantage on the clock. >2- Play a sharp gambit defense in reverse. This is the idea of 1.d3 e5 2. d4 >i.e. 2...exd4 3. Nf3 > >I would assume somebody has thought of this before and that some engines are >able to handle this easily. In fact, an engine that can't I would say has a bug >IMO. Which engines are able to cope with this trick effectively? I don't at the moment. I did several years ago, however. The idea is to simply "flip" the board by changing the color of all pieces, and then swapping everything on rank 1 with rank 8, rank 2 with rank 7, etc. Now, in the above you end up with e4 e5 (or whatever) with white to move rather than black. You find the move then "flip it" back to black to move. I quit doing it when I added book learning, as it caused some massive grief, because even though it had found a move as book by the above trick, I didn't do the trick in the book learning code itself as it was complicated. But it could be done. It wasn't that unusual for programs to handle this years ago. I started doing it in Cray Blitz when humans did that to it (before it was cray blitz) in the late 1970's. It was a known "trick" that worked well. We picked up transpositions back then because we used the same hash signature approach I use today, where many programs did not. IE Sargon actually had a "tree" for their book and changing the order meant both orders had to be entered into the book or it wouldn't find the transposition. Dan/Kathe eventually went to the same hash signature approach everyone uses today, and I'd guess they found the same problem with e3 e5 e4. However, I saw a variation of this in the 1996 WCCC event with the "Ruy Lopez, Crafty variation." e4 e5 Nf3 Nc6 Bb5 a6 Bc4. That took everyone out of book, and yet they played the obvious developmental moves such as Nf6 and so forth. I didn't plan it, as my "random 0" mode does a search on all book moves and takes the one with the best score, and Bc4 came out better than Ba4 or Bxc6. It turned out to be _much_ better as we went 20+ moves in book while everyone else was out instantly. We ended up playing 3 programs (at least) with a 2:1 time odds advantage (40 moves in 2 hours vs 40 moves in 1 hour) or whatever the time control was. It was funny. What was _really_ funny was I had posted such a game where Crafty beat chess genius, in my office, while tuning up for the game, and I had pointed out how stupid the move looked to me. But I forgot about it, and it started happening in Jakarta. I didn't know anything about it until one morning on ICC, Bruce Moreland was asking everyone "What should I play after e4 e5 Nf3 Nc6 Bb5 a6 Bc4?" That was the first I had heard about it. No internet access from the Jakarta playing hall so news came out of there a day or too behind. Books can be funny things...
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.