Author: Ratko V Tomic
Date: 20:05:50 10/08/99
Over the last couple days I downloaded and installed two versions of updates for Chessbase programs (2xHiarcs 7.32, 1xFritz 5.32 and 1xJunior 5). Well, I am ready now to uninstall the whole "improved" thing and reinstall the old and "unimproved" stuff. Here are the "improvements" I noticed in about day and half after the installation: 1. The new search "feature" (bug?) while still in the book dumps the search info into the window where it used to show the opening move options. So, you can't look any more at the standard opening moves as you play. (Of course, you can still "Browse tree" but that's not playing a chess game any more but studying the opening.) The new "feature" kicks in at around 3rd or 4th move and from there on they clobber the display of the book moves with the analysis. 2. The program's moves were "improved" by an animated flicker of the piece the program moves. This improvement can't be turned off. Maybe they had five year olds beta testing (if they do that any more), since only they could possibly like that kind of amusement (to break the boredoom of having to play chess, I guess). 3. The analysis window gets erased between moves with a different color than the background, for a fraction of a second, then it gets erased again with the normal background color. That makes the whole window flash as if a lightening is firing behind it. Again, my five year old may get amused by that kind of "special" effects. Anyone older risks epilepsy while playing fast games. 4. Tried some bad lines in Hiarcs (which it loses right after it gets out of the book), they're still in the "improved" book. So much for "Weed the tree" procedure. 5. The "learning" feature is as dumb as ever. If you turn it on, it quickly ends up playing the same opening line over and over, while if it had lost some game, it drops the whole big chunk of the opening where that line occurred, even though it lost for some other reason much later in the game. If you turn the "learning" off (and also set file attributes on the opening books to read-only, since it doesn't seem to respect the mere configuration setting to "no learning", I guess it helps them cheat on mechanized tests which turn the learning off; I am thankful, though, they don't require [as yet] the hard disk to be disconnected during the opening to disable the "learning" for real) -- well then every time it comes out of the book you have to wait however many minutes it takes it to discover from scratch the same move it discovered the last time you played that line. Is just remembering the moves played (at given level) without skewing the variety of opening play too complex to program? It seems by bundling together the remembering of the line played along with skewing the opening choices, they've tuned themselves to perform for the SSDF's mechanized testing (so they can appear stronger by winning the identical game ten times in a row, I guess), as for customer's convenience and time, who cares, let them buy faster machine if they don't enjoy waiting. Why can't they just remember the result of calculation from the last time (so they don't waste my time, thinking the same thing over and over with the same result over and over) without reducing the opening variety. In case they lose, they need to back off one or two moves (at most) from the point their score went negative, but not drop the whole Sicilian after they lose couple games in Sicilian due to the program's misjudgment 10-15 moves after the book ended. Now, for fairness, all programs, if they have learning, it's geared to give them an edge in a mindless kind of automated machine-machine play, while at the same time forcing their customers to either waste time on repeated calculations or to play the same opening line over and over. So this is not just a CB problem. I wish some of the folks who write these programs would try it out occasionally as a user, and see if they enjoy either of the two choices above. If they wish to get an edge through such "inspired" means, it would be enough to have an option for "mindless machine-machine" play and do this kind of so-called "learning" in that mode only. Customers are not computers who don't mind playing the same line over and over and who don't mind waiting for the other computer to calculate the same thing over and over.
This page took 0.01 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.