Home Calendar E-mail Challenge Tournament Reports Chess Puzzles



Chessmaster 9000
vs.
Fritz 6



By Roger McIntyre

A little over a year ago I bought Chessmaster 8000 and was somewhat impressed with all the tutorials and the look and feel of the game but I was very curious to know how strong the engine was so I pitted it up against Fritz 6 in a 12 game match (read my report here). Fritz 6 totally outplayed Chessmaster 8000 in this match by winning 5 games to Chessmaster's single victory with the rest of the games drawn for a 8 to 4 final score. Even after this trouncing I was impressed with the Chessmaster program due to it's teaching ability and the playing personalities. Chessmaster's personalities are much more fun to play against than the levels Fritz offers.


Chessmaster 9000 Enhancements

There are a number of changes introduced in the Chessmaster 9000 program. Even more tutorials have been added including "The Psychology of Competition Chess" course taught by IM Josh Waitzkin and an Endgame quiz by GM Larry Evans. A Blunder Alert feature has been added that can notify you when you make a significant mistake. Even more chessboards and pieces (true 3D) have been added to the many it already had. 10 new Grand Master personalities have been added but one, Garry Kasparov, was curiously removed. A few of the lower rated personalities have also been remove. Most of these were rated well below 1000. I thought Chessmaster 8000 had too many personalities in this range anyway so they won't be missed. Maybe more ratings greater than 1500 should be added. An Endgame Database which has all 3, 4 and some 5 piece endgames has been included. If you choose the "Full installation" option when you install the program Chessmaster and all the GM personalities will use this database to play perfect endgames. The Chessmaster 9000 engine "The King" written by Johan de Koning has also been improved.

Another notable difference between Chessmaster 9000 and Chessmaster 8000 is the strength of play of the different personalities. The ratings in Chessmaster 9000 are probably 200 to 300 rating points stronger than in Chessmaster 8000. So whatever your rating was playing Chessmaster 8000, expect it to be 200 to 300 rating points lower when you play Chessmaster 9000. The Chessmaster team claims that they established the personality ratings by actual games played against real people and other programs and that they believe they are accurate to within 50 points of their "real world" USCF equivalent.


Chessmaster 9000 vs U.S. Champ

On 29-30 September 2002 Chessmaster 9000 played a 4 game match against the reigning U.S. Open Champion Larry Christiansen. In each of the games Chessmaster used a different personality. Chessmaster impressively won this match 2 - 1.

WhiteBlackOpeningResult
CM9000 (Alekhine)Christiansen, Larry MD41: Semi-Tarrasch Defence0-1
Christiansen, Larry MCM9000 (Fischer)A30: English Symmetrical0-1
CM9000 (Botvinnik)Christiansen, Larry MA28: English Four Knights1-0
Christiansen, Larry MCM9000 (Chessmaster)A07: Barcza System1/2


Chessmaster 9000 vs Fritz 6

Since Chessmaster's engine has been improved and did so well against Larry Christiansen I was eager to see how it fared against Fritz again. I conducted this match just as I did the previous Chessmaster 8000 vs Fritz 6 match. Each program would use its opening book. Once out of their book they would have exactly two minutes processing time for each move. No endgame databases would be used. Only one program would be running on my 1.00 GHz AMD Athlon processor at a time. This is important because I found that if both programs were running on the same computer at the same time Chessmaster tended to hog the CPU and perform much stronger than Fritz. The match would consist of 16 games where each program would play 8 games with the white pieces and 8 games with black.

The first thing I noticed when the match got under way was that Fritz's opening book was far more extensive than Chessmaster's. Fritz's book took it deeper into 9 of the games to Chessmaster's 2. Once Chessmaster began evaluating the positions the evaluation numbers (number of positions looked at and depth) looked similar to its last version. Fritz claimed to be looking at about 5 times more positions than Chessmaster and it's depth of exhausted levels was always higher than both Chessmaster's exhausted levels and critical lines, usually by 3 or more levels! But there seems to be something bogus with this because of the 5 announced mates in this match Chessmaster announced 3 of them first! Also, when large evaluation changes occurred in the games sometimes Chessmaster whould show the changes first and sometimes Fritz would. It looked to me like the two programs were roughly evaluating to the same depth.

Here are some notable positions that occurred in some of the games...

If Chessmaster used its Endgame Database it would have improved its results in two of the games. In game 1 Chessmaster missed a drawing opportunity when it move 57... Rh1? This allowed Fritz to cut his king off from the defense of the pawn's queening square with 58. Re6! and eventually Chessmaster would have to sacrifice his rook for the pawn and lose the game. If Chessmaster had just 2:24 instead of the allotted two minutes it would have found the correct continuation 57... Ke7!, or if it was using its Endgame Database it would have found this drawing move instantly.

The second game that Chessmaster would have improved upon if it used its Endgame Database is game 7. In the actual game Chessmaster played 101... Rg6? and eventually drew the game by the 50-move rule. In this position black has a forced win with 101... Re5! This move cuts the white king off from the defense of his knight and the knight will soon fall. Even when I gave Chessmaster more than 15 minutes it did not deviate from the incorrect 101... Rg6? Fritz, however, recognized 101... Re5! as the correct move after only three seconds.

Game 5 was a very interesting game. Fritz was a pawn up and chasing Chessmaster's king all over the board. During the game Fritz continually evaluated an advantage of over two points but Chessmaster insisted it was close to an even game. In the diagrammed position on the left Fritz evaluated its advantage to be 2.53. Chessmaster's evaluation was 0.09. Who was right? On move 89 the game was declared a draw based on the 50-move rule. Chessmaster must have saw something in the position that Fritz didn't.

Game 2 was a game that Chessmaster evaluated to be close all the way until the end. Fritz claimed a slight advantage for its black pieces. In this situation Chessmaster's evaluation was correct but its play was not. Chessmaster played 58. Qc8+ but probably should have played 58. Qb6 to trade off the queens. Without the queens this would lead to an easy draw for both sides. Chessmaster would have a couple other opportunities to trade off queens during the game but it refused to do so and Fritz showed it how the queen and knight dominates the queen and bishop.



Chessmaster's evaluation of the rook/pawn ending shown above that occurred in game 9 was much better than Fritz's evaluation. Fritz missed the win when it moved 60. Rxh6? Chessmaster knew that the winning move was 60. Rg6! Even when I gave Fritz 15 minutes to see the error of its ways it persisted on 60. Rxh6? After 60. Rg6! white should eventually get connected passed pawns that will march down the board for the win. As the game went however, the top right diagrammed position was reached. Fritz still thought it had a huge advantage, evaluating the position as 2.69. Chessmaster's evaluation was 0.41. This is a typical position where humans tend to evaluate better than computers. A decent human player can look at this position and know that white can never queen the pawn. If the king goes down it help queen it, black's rook will simply check it until it abandons the idea or triggers the three-fold repetition rule. It's impressive that Chessmaster saw this. Eventually the game was drawn by the 50-move rule. Two moves before the 50-move rule would occur Chessmaster made a move that allowed Fritz the opportunity to queen the pawn - one move too late. I'm not sure but I thought I saw the Chessmaster crack a smile.
In most of the endgames Chessmaster's evaluatation was more accurate than Fritz's. Another example of this was in game 6. The game came down to a knight and pawn versus a bishop endgame. Since the white king was in front of the pawn defending its queening square, it was pretty obvious that black would never be able to promoted his pawn. Fritz's evaluation in this position was -1.50 whereas Chessmaster's evaluation was -0.13. Eventually the game was drawn by the 50-move rule.

Maybe the biggest blunder committed by Fritz in this match occurred in game 15. In this position Fritz played 79. Rd6? in which Chessmaster responded with 79... Rg2! and quickly gained a huge advantage and won the game. I let Fritz evaluate this position several times and it didn't come up with the same move every time. There is an element of randomness in its choices. Sometimes it did come up with the correct move 79. Kd2 which should draw. This match was touch move so I had to accept its first selection which lost it the game.

Below are the results and games from this match.

Click on the game number to see the game.
GameWhiteBlackResultCMFritz
1Fritz 6CM 90001 - 001
2CM 9000Fritz 60 - 102
3Fritz 6CM 90001 - 003
4CM 9000Fritz 60 - 104
5Fritz 6CM 9000 - 4
6CM 9000Fritz 6 - 15
7Fritz 6CM 9000 - 15
8CM 9000Fritz 61 - 025
9Fritz 6CM 9000 - 36
10CM 9000Fritz 6 - 36
11Fritz 6CM 9000 - 47
12CM 9000Fritz 60 - 148
13Fritz 6CM 9000 - 48
14CM 9000Fritz 6 - 59
15Fritz 6CM 90000 - 169
16CM 9000Fritz 61 - 079


Conclusion

Fritz won this match 9 to 7. Fritz won 5 of the games, Chessmaster won 3 and there were 8 draws. Unlike the match against Chessmaster 8000, Chessmaster 9000 put up a much tougher fight. Clearly Chessmaster's engine is much improved. Generally speaking it seemed Fritz performed better in extremely complicated positions with lots of pieces on the board and Chessmaster did better when the games entered the endgame. The results were too close for me to say definitively that Fritz is the stronger program, especially given the fact that Fritz had an opening book advantage and if Chessmaster had been using its Endgame Database the result would be tied 8 - 8. I am very curious as to what rating this program will earn with The Swedish Chess Computer Association. It most certainly will move up the ranking list.


Last modified: 12 January 2003
Roger McIntyre