Author: Jonas Cohonas
Date: 02:53:02 05/19/05
Go up one level in this thread
On May 19, 2005 at 05:46:11, chandler yergin wrote: >This maybe a good time to reflect on what the Analysis Module & > >Engine is really doing; and what the Opening Trees really are. > >The Opening Tree consists of positions, not games. > >For every position in the Tree, the Program can display what moves > >have been played, how often, by whom, and what level of success > >they have had. The statistical % is used by the Program > >when out of book. It can be misleading, and I think Dann Corbit > > covered this in his Post, when he indicated that a move might > > reflect an average score of 70%, but this % comes from older games. > > or, the line may have been abandoned when a refutation was found. > > Unless the Database is constantly updated, by new games, > >the latest in Theory will be missing. > >The advantage in studying Openings using the Tree, is that there are > >no transposition problems. > >The moves or move order used to reach the position is irrelevant. > >There are some that seem to be confused about what the Analysis > >module and Engine is doing; so I address this to them. > >If any Programmers can add to this, I welcome your comments. > >At any time in any position, the Analysis module will examine > >and evaluate every possible move 1/2 Ply at a time. > >Example: Many times on CCC a position or FEN is given, and the Question > >asked.. "How long does it take for your engine to 'find' such and such > >a move?" > >This is misleading.. as the Engine has already found it almost instantly. > >You just aren't seeing it, because you are "Scrolling" only the PV! > >If the Postion has 20 'possible' moves; the Module has 'evaluated' > >20 PV's Simultaneously, and ranked them in order of 'best' move first. > >It will do this continually 1/2 Ply at a time. > >Each time is one iteration. > >The more 'possible moves' there are, the more time it takes for each iteration > >as the depth of search goes deeper, the Tree gets larger. > >All the Module or engine is doing is "Evaluating", based on static > >positional factors which are programmed in. > > Material, King safety, Space, strong and weak squares, > >the 2 Bishop advantage, Pawn structure, Open files, half open files, > >and many other factors the Programmers feel are important. > >THere is also a learning algorithm where the Program can learn from > >it's mistakes, and is self correcting in that regard. > >I can only speak of my experience with Chessbase; some of you have Bookup > >or other Database Programs. > >In Chessbase 9 for example, when I want to analyze a position, > >I Open Chessbase 9, Open a Board, insert the Fen, and Start an Analysis > >Module. The nice thing about this is, that I can Clip and Paste the entire > >Analysis into a word Processor or CCC POst with Diagram. > >It shows exactly the Iterations, time, eval, etc.. > >Many here use only the Engine EVals which they "Cut & Paste" > >and it does not show the detail of what the Engine is doing. > >The Opening Tree is really very shallow, and the Program > >gets 'out of book' early. > >In these cases, it is usually in the middle to late Opening > >phase where there are many 'possible' positions. > >This is where the time factor is critical. > >When playing against the engine in these situations, many times, > >the "best" move is far below the horizon and the Program cannot see it. > >This is where the Top PLayers shine... and get their advantage. > >Under Tournament time controls, the Time allocation software > >determines when the Program must move. If the time is 40 moves > >in 2 hours, the average time for each move for the Program > >is 3 minutes. As you can see, this is often not enough, > >and the Program may make what looks like strange moves. > >It may see that in 40 Ply, it will have a serious disadvantage, > >and will sac pieces just to postpone the inevitable. > >Endgame Table Bases give the Program 'exact' play. > >It also makes them make strange moves. > >I have seen the Program give up a Queen, simply to get into > >the EGTB's. This is certainly not human behavior, but is > >typical of Computers. > >Another interesting point I might make is the Opening Tree > >that comes with the Program. > >Earlier I mentioned that the Tree uses Statistical information > >from the number of games played, and the success rate of the moves. > >This can be very misleading.. for example; The Urusov Opening. > >The statistics are very much in favor of Black! > >Well.. only because it is seldom played; and most of the White > >PLayers, including Steintz & Lasker had blundered. > >Lasker lost in 7 moves I think, Steintz did not do much better. > >If you want to perform an interesting experiment, Open the > >Database Program.. Open the Tree, start your analysis module, > >browse through the Tree, and watch the Evaluations.. > >What may look as 'winning' for one move, can turn into a disaster > >on the next! > >I'd be interested in any and all comments, and your own personal > >experiences with Databases, Engines, etc. if you'd care to share. > >Some may have some good ideas on how best to use the Computer > >to study Openings. > >Correspondence players for sure. > >I'm sure Komputer Korner would appreciate it too.. > >Thanks for listening, > >Chan This way of writing is just as annoying as the one long paragraph "style", i read a couple of lines, got annoyed and wrote this message.
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.