In the not-too-distant past, writing an amateur chess computer program was strictly the province of college students gunning for an A in a programming course, or professional programmers messing around with chess on the side as a means of learning new programming techniques. Internet web sites are rife with old DOS programs, still downloadable, which afford an interesting glimpse into the history of chess programming in the 1980's and early 1990's. Most of these programs shared some common tendencies -- they were middlin' to fairly strong tactically but played terrible positional chess.
As the years have passed, programmers have learned new techniques for getting their chess programs to play better in closed positions. And these techniques aren't the sole property of professional chess programmers anymore -- many amateur programs play strong chess, both tactically and positionally (though, to be honest, chess computers still have a long way to go in the positional area).
Seven of these impressive amateur programs have been collected on a new CD from ChessBase: Young Talents. Playing against these programs is a real education on how far chess programming techniques have come in the last few years, even on the amateur level.
Young Talents uses the Fritz6/Junior6 interface (prior ownership of these two programs is not required) and contains seven different chess programs for you to play against and analyze with:
Each of these engines has its own unique "personality" and style of play. A few technical details are available at the ChessBase GmbH website, but I'd like to provide you with my own impressions of each of these engines.
AnMon5.07 is a strong program, no doubt about it, one of the strongest of the Young Talents engines. It plays in a very tactical style. Its style of play put me very much in mind of Fritz2's: it attacks relentlessly with threat after threat until you finally succumb. Of the seven engines, it's the most "old-fashioned", playing in the most "computer-like" style. This isn't a slam against AnMon; I won't be disparaging about it until sometime in the far forever future when I somehow manage to beat it (fat chance!). A gauge of the relative strengths of the Young Talents engines is to compare their ranges of ratings in the "Handicap and fun" mode parameters window. In it, AnMon shows a rating range of 1300-2100 (after you've "crippled" it; keep in mind that this is "Handicap and fun" mode we're discussing), making it one of the strongest programs of the group.
Goliath Light is the most "tweakable" of the seven engines on the CD. It's the only one that has engine parameters that you can play with and reset. These are similar to Fritz6's and include contempt value, selectivity, learning, combination search, a tablebase toggle, and three style settings: normal, careful, and risky. The biggest feature of Goliath Light is its speed -- this baby screams! It's about as fast as Fritz5.32 in kilonodes per second -- so it's obviously a strong tactician. In the "Handicap and fun" mode rating range test, Goliath Light scores the highest: 1305 to 2105.
Gromit3.1 displays a slightly lower range on the "Handicap and fun" mode slider than do most of the other engines (1250 to 2050), but its style of play makes it a lot of fun as a competitor. The programmers have opted to build positional knowledge into the engine. This explains why it's slower than many chess engines, but the positional ideas it displays make up for its lack of tactical speed. I was really impressed by many of the moves Gromit made in the games I played against it. Gromit frequently passed up the anticipated "computer-like" tactical ideas to play more positional moves. It's one of the most "human-like" engines on the Young Talents CD and I expect it to become one of my regular sparring partners.
Ikarus0.18 is the "baby" of the group -- it's one of the newer engines on the CD (just look at the version number). It's unusal among chess engines in that it's programmed in Delphi rather than the "usual" programming languages C or C+. This also explains the weird results you get when you look at the slider in "Handicap and fun" mode -- it displays a negative rating for its low-end value. I'm not sure how it rates compared to the other engines, but it's always kicked my butt, even in "Friend" mode. It's a very strong engine and plays a highly tactical game.
Patzer3.11's name is a bit of a misnomer -- there's nothing weak about this engine. It's not the best blitz player around, but it really kicks at longer time controls (G/15 and up, in my experience). Its "Handicap and fun" range is 1300 to 2100, putting it on a par with the other engines on the CD. It's a good positional engine; it played a very human-like game in its contests against me. It's fast becoming one of my favorite opponents (along with Goliath Light, Phalanx, and SOS).
PhalanxXXII was the most interesting of the seven engines (in my opinion). I was looking forward to a wild sacrificial game against it (based on the engine's description at the ChessBase GmbH site), but Phalanx fooled me. I played a couple of games against it immediately after replaying a couple of Steinitz' games from Reti's book Masters of the Chessboard. Steinitz was the master of the pawn push. He often drove his pawns relentlessly forward to deny space to his opponent. To my surprise (I might even say "shock"), Phalanx does the same thing. Control of space is evidently given high priority in its evaluation algorithm. As I played against Phalanx, it seemed intent on throttling me ("python-like", to use the hoary old stereotype). I give this program high marks for positional play -- the pawn pushes it made were all strong ones, and it never gave my Knights a chance to exploit any holes it may have created. Despite the relatively low rating range in "Handicap and fun" mode (1150-1950), Phalanx was without question the most unique and interesting of the seven engines and I'm looking forward to my next game against it.
SOS was another surprise. Computer chess programs are traditionally very weak in the endgame. The endgame requires a great deal of knowledge to play properly and brute force calculation just doesn't cut it -- there are a lot of endgames you need to know how to play instead of trying to calculate them out over the board. SOS exhibited some really impressive endgame technique when I played it. I tried the old anti-computer trick of swapping pieces off the board in the hopes of whacking it in the endgame. But my hopes were mercilessly dashed by SOS. It handled the endgame better than most chess programs I've played -- and it did so without the use of tablebases. It definitely lived up to its billing on the ChessBase GmbH site. It displays a "Handicap and fun" rating range of 1300 to 2100, so it's on a level with the other top engines on this CD.
All of these engines are very strong chessplayers, but if you're looking for something really unique as far as style of play is concerned, I highly recommend Gromit, Phalanx, and SOS from this CD -- they're definitely different from any other chess programs I've played (and I have well over 100 in my collection, so I do have a bit of experience in this area). It's really surprising to see how far chess programming has come in the last decade and the fact that these seven engines are essentially part-time "hobbies" for their creators is nothing short of amazing.
I've been having a lot of fun with the Young Talents engines and I'm sure you will too. These engines (especially in combination with the multiple engines that come with Fritz6) are like having a chess club on your computer, all at your beck and call 24 hours a day. As stated previously, the Young Talents CD comes complete with the Fritz6/Junior6 interface, so no other software is required (although these engines will work within any of our existing 32-bit interfaces).
Until next week, have fun!
You can e-mail me with your comments, suggestions, and analysis for Electronic T-Notes. If you love gambits and sacrificial play, stop by my
Chess Kamikaze Home Page
and the Yahoo Chess Kamikazes Club.