Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: The Mirage of backsolving

Author: chandler yergin

Date: 19:06:15 05/17/05

Go up one level in this thread


On May 17, 2005 at 21:45:48, Komputer Korner wrote:

>It is incredible that once a dishonest or false idea takes hold in the
>marketplace, how very difficult it is to overturn it. Once it becomes a sacred
>cow, most people just accept it on faith. Thus unfortunately is the case with
>backsolving. Let us suppose that you have either bought, inherited or downloaded
>an electronic openings book without any annotations. How are you going to know
>what lines will be in your repertoire/ After soul searching and examining lots
>of the opening lines, and other advice, you decide to play lines that fit your
>style. However within these lines there are many options at each move node. Let
>us say that you have one important position with 4 move choices at the 4th move.
>How are you going to get these multi move choices annotated so that you will
>know which lines to memorize for (over the board) or which lines to play for
>internet or correspondence chess. If you can't afford GM or IM analysis of your
>repertoire and don't trust your own analysis, you are left with either having an
>engine analyze the opening nodes or you backsolve. If you backsolve you are
>forced to put in a lot of extra move lines because backsolving depends on an
>ending result or an evaluation far up the tree(deep into the game). So if you
>are a committed backsolver, you accept this and you search for a lot of GM games
>in your specialty opening repertoire to add to your opening book. You will have
>to add an evaluation to the end of each one of these lines and then you
>backsolve. However,  there is an average of 3 positions in every GM game where
>the assessment of the game is overturned (in other words, the assessment goes
>from advantage from one side to the other or an equal position goes into an
>advantage to one side or Vice Versa). This means that you are never sure that
>the result really represents the value of the opening node  line that you want
>to have annotated and that backsolving indeed gives you as it backs up the
>result evaluation to the opening node position.  So you cheat a bit; you don't
>put the complete GM games. You put in the 1st 30 moves or so and then have an
>engine evaluate the final positions after 30 moves or wherever the engines give
>a definite advantage to one side and then backsolve again. There is a separate
>problem with equal and unclear positions but more on that later. However, you
>wasted engine midnight analysis sessions on the 30th move analysis which you
>probably will never get to play as your opponent will have veered off on another
>line far sooner. The point is that backsolving is doing its annotation work (on
>positions that you have your engine analyze) far up the tree (beyond the 30th
>move) on positions that you will never encounter. Even if you back that up and
>only analyze positions after the 20th move and then backsolve you will still
>never encounter those positions in a real game. Opening study really means
>opening study. It doesn't mean ending study. It doesn't mean mid game study. Of
>course you should do some of those as well but only after you have a solid
>repertoire. You are far better off putting your engines to work on that 4th move
>(in other words study openings the old fashioned way).   Now for the unclear
>lines that the backsolvers 30th move analysis showed that I promised to talk
>about earlier. These lines represent a special problem. If you backsolve them,
>then you haven't really accomplished anything because the opening node that is
>the ultimate target of your backsolving has your 4 opening  choices without any
>annotations before you started backsolving. Having no annotations is really the
>fact that you don't know what is going on therefore  they are unclear.   So let
>us say that 2 of the lines are backsolved to be unclear and one line was large
>advantage for one side and the other line was backsolved to be equal. So the
>only one we haven't talked about is the equal line.  It has the same problem as
>the unclear line. Since most games have 3 position overturns as explained above
>you are not really sure whether the line was always equal from the 4th move on
>or else it became an advantage and then was turned back into equality because of
>mistakes. if every game was played perfectly, then backsolving would be worth
>it. However that is not the case. So to sum up, put your effort into evaluating
>opening positions and opening strategies and traps. There is no magic bullet
>that will solve chess and backsolving doesn't solve anything.

Total gibberish!

I don't believe you are a Chess Player!



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.