---

Here is an article we have received from Mr. Leonard Dickerson, Life Master and past champion of Tennessee (he came in second last year).

We have not changed a single word from this article.

I, a USCF Life Master and past champion of Tennessee, requested that chesstiger.com allow me to evaluate their program against my copy of Chess Genius 1.3 running on a Palm Vx. They bravely agreed to the match and suggested that I conduct a match with greater than 20 games so that the results would be statistically relevant. Well, the results are in and the objective evidence of the attached 26-game match proves that claims of Chess Tiger 14.5’s superiority are not exaggerated!

With 17 of 26 (14 wins and 6 draws), it has slightly over a 65% winning margin. Based upon the USCF win expectancy probabilities, this correlates to a 110 point difference in rating strength. Hence if one regards Chess Genius as performing at expert strength (i.e., 2000 level) when played at 60 min/Sudden Death, then Chess Tiger is definitely playing at the Advanced Expert Level (i.e., 2100 strength).

Both programs performed better as Black! When playing Black, Chess Genius was 5.5 of 13 as Black, but only a paltry 27% (3.5 of 13) as White. Chess Tiger was 9.5 of 13 as Black for a dominating 73% winning percentage but scored an average—but respectful—57% (7.5 of 13) as White. So if you want to win, let the machine have White!

Genius played the Black side of Queen pawn openings better than Tiger, but did not do so well in open King pawn games. Don't be misled, though. Genius managed to get the advantage out of most of the openings (as much as an actual pawn up—not evaluated advantage—early with no disadvantages) but could still not hold on to draw! Tiger essentially spots Genius a pawn and then goes on to beat him! (See games 4, 8, 18, and 22) But some credit should still be accorded Genius because Tiger's king pawn opening book seems to run 2 or 3 moves deeper than Genius. Overcoming an opening disadvantage to take the initiative was to no avail, though, because Tiger simply outplayed Genius in the endgame—even from a pawn down! So, Genius plays slightly better in the openings but much worse in the endgame.

I personally feel that Tiger is a much more dangerous opponent for humans. This is based upon his strength when playing e4 and ability to press the initiative. When playing d4 or closed games, Tiger is not quite so intimidating. Consider that when he opened d4, he only won 2 of the 7 games! When he played e4 6 times, he won 4 of the games. The two e4-games that he did not win were due to their closed nature: a Ruy Lopez and a transposed Caro-Kann.

Even though Genius had missed some wins before, I did not become convinced that Tiger was clearly superior until game 21 when Tiger won the Black-side of an inferior heavy-piece endgame where he seemed doomed to passivity. Then game 22 confirmed my evaluation because White wins a pawn on the 18^th move with no disadvantages and then goes on to lose! If more proof is needed, then witness game 24 when Tiger foolishly sacrifices a pawn for only light pressure but Genius immediately caves in under the illusion of an attack and fails to even draw. Game 8 is another depressing game of Genius being a pawn up in a safe position but not being able to retain even the draw. Tiger is simply the better program.

Might the results had been different had the games been played at 2 hours/SD? Maybe. But consider that of the two games played at this rate (games 9 and 21), Tiger scored 1.5 points. Also, who has a four-hour battery in their Palm Pilot to venture these longer games or the inclination? If you have the time for a long game, then it is likely that you will be able to find a stronger game on a dedicated computer or on a program running from a personal computer. So for a quick opponent, 1 h/SD is practical—which makes this match-up reasonable.
 

Leonard Dickerson

Game 9 and Game 21 were played at 2 hours/SD. Games 4, 5, 9, and 25 were valid draws claimed by Chess Genius that Chess Tiger did not speedily acknowledge as such. Hence I replaced its inconclusive "*" with the draw result.