Author: Adam Oellermann
Date: 06:09:08 08/09/01
Go up one level in this thread
On August 09, 2001 at 08:59:01, Uri Blass wrote: >On August 09, 2001 at 08:54:27, Adam Oellermann wrote: > >>On August 09, 2001 at 07:30:57, Uri Blass wrote: >> >>>On August 09, 2001 at 07:13:21, Jeroen van Dorp wrote: >>> >>>>On August 09, 2001 at 06:19:01, Uri Blass wrote: >>>> >>>>>I do not have the impression that weak players use the chess computer method. >>>>>They have enough time at tournament time control to do 1 ply search and 1 ply >>>>>search after the move that they plan to play but they do not do it and a common >>>>>mistake of amatuers is not to play a move simply because they did not analyze it >>>>>or to lose material or to do a positional mistake because they did not consider >>>>>the opponent reply even for one second. >>>>> >>>> >>>> >>>>You should really read the article, if you didn't already. The way Graham tells >>>>it is a one line synopsis of an article, which is a synopsis of a research >>>>project. That's logical. >>>> >>>>The comparison lies not so much in raw calculating, but in the necessity of >>>>*assessing* each chess position from ground zero. >>>> >>>>The extensive use of their memory sets strong chess players up to constantly >>>>recall both the current situation as well as the development of the position, >>>>leading to pattern recognition, and better results. With this they can -based on >>>>the position characteristics- develop a plan and the candidate moves much more >>>>easily. >>>> >>>>As the weak chess player lacks this level of pattern recoginition, he has to >>>>assess the position a bit like a basic chess program does: start crunching all >>>>move sequences. >>>>For a computer program this still can result in very strong play - because of >>>>node speed - but as the human brain hasn't got the ability to not only tune up >>>>to that speed, but store and retrieve the analysis tree in a useful manner, the >>>>*weak* player will oversee even the most simple countermoves of the opponent. >>>> >>>>Basically the research seems to indicate that *memory* is very important, and >>>>the lack of it leads to more basic and flawed methods of calculation, *not* that >>>>the strong player thinks like a human and the weak like a computer. >>>> >>>>Nothing new under the sun, as these results confirm earlier research. >>>>And we all *know* that, as we play the same. >>>>Well know fact is that drilling with (simple) tactical combinations makes you >>>>aware of the possibilities of recognizing them in real play. >>>> >>>>I am *not* a strong chess player, but with my training and after-game analysis >>>>I'm always looking almost *automatically* for these positional characteristics. >>>> >>>>Just a few days ago I have been looking at numerous games with the sicilian in >>>>the ECO B50's range, as I have good results with those opening as black, but >>>>very often am stuck with a weak pawn on d6. >>>>I kick my opening books aside and specifically tried to identify the >>>>*characteristics* of games in which black won the game, although the d6 pawn (as >>>>a plus pawn for white) was nicked off the board. >>>> >>>>Not looking for a 'move sequence' how to solve it, but characteristics of piece >>>>placement and strong and weak points. A basic chess program won't do that, as >>>>well as a weak chess player. >>> >>>I agree that recognizing patterns is importnat but calculating is also important >>>and I believe that training in 1 ply searches can help most of the players >>>including the GM's to play better. >>> >>>1 ply searches are not going to help in most of the cases to get a new idea but >>>one case when it helps to get a new idea can help to get get a better result. >>> >>>Uri >> >>If we're going to transplant computer-chess terms into human chess playing, we >>should probably say at least "1 ply search with quiescence". What does 1 ply >>help? It just saves you accidentally leaving a piece en prise. > >No >It also can help you to find a good positional move. > >There are cases when humans do not consider a good positional move and looking >at the move for only 1/2 second can be enough for them to understand that it is >probably a good positional move that they need to analyze. > >Uri I agree completely with this sentiment, but I think calling the result a consequence of a 1-ply search is misleading. A human looking at the position for an extra 0.5 seconds after that 1 ply will *not* simply apply a static eval, but a quirky combination of eval, memory and highly-pruned search which is impossible to quantify. Although there is no real analogue for this in computer chess, quiescence extension comes a lot closer than just flat 1 ply, to my mind. - Adam
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.