"It plays the most reckless chess you have ever seen, but for the most part it gets away with it. It truly plays chess like Mikhail
Tal, its namesake. . . [and] gives you the impression that you are playing against a real person."
Komputer Korner

"It is the most *interesting* program I have ever played, even if it has the most crazy excursions..."
Alastair Scott

Chess System Tal - A Promise Finally Accomplished?

by Fernando Villegas

To create a real breakthrough, a program capable of a qualitatively different, human-like approach to chess, is not something
new in the agenda of Mr. Chris Whittington. Years ago he already had that ambition, with two important differences: first, he
did not talked in that time about "idealism" versus "materialism", but of implementing a selective search approach as opposed to
a brute force one. And second, he did not achieve it. Not that his "Complete Chess System" and its somewhat lesser brother,
"Chess Champion", were total failures. Complete Chess System was the first commercial chess software in both mass market
and professional niches that included a respectable collection of games, even with some database capabilities, and a very pretty
interface. In fact it included a lot of features that were new, innovative and gave a kind of distinction to an otherwise relatively
cheap product. And to beat the chess engine was not a piece of cake. On a Pentium 200, Complete Chess System scores
2160 ELO on the LCTII test. That is equivalent to the around 2000 ELO of the computers that were available in the year of
Complete Chess System's birth. Not bad at all for most of us, but of course at the same time there were far stronger programs
and/or far better marketed programs. And what is more, there were the so called "professional" programs that until now have
gotten the lion's share of any discussion or review. No matter how good a "mass market" program is, it has been and it will be
scarcely discussed by people like us, fans and critics. Look at the cases of ChessMaster and Power Chess, both very strong
programs by Johann de Koning and Kittinger but nevertheless they are rarely on centre stage. So, because it was not one of
the strongest and because clearly was just a "mass market" program, we could say without much risk of error that the first try
of Chris for getting a different yet effective chess program was at least a relative failure, if maybe not so on the economic side,
because Complete Chess System did sell a good number of units.

Past History, Actual Success

But now all that is old history. It seems that Chris decided to dedicate a far greater amount of time to his chess project as far as
his other concerns in his Oxford software company fell into place and began to run smoothly. And Chess System Tal is the
result of around five years of hard and dedicated work in association with Thorsten Czub, a guy who operated, as Chris has
told us, as a mix of friend, spiritual support, and also as an intelligent partner in the process of developing heuristics capable of
giving the engine a capacity to look at each position as a master could. The result: a chess program that seem to finally satisfy
the aspirations of its father.

However, there were a lot of dangers in the way of Chess System Tal, apart from the obvious one born from the intricacies of
his goal, of the sheer idea to imitate the intuitive, attacking, daring style of Tal or at least of a good chess player trying to be like
Tal. Then, Chris is not a secretive person. During the long incubation period of Chess System Tal he did not hold back his
criticisms about his colleagues average way of doing things. He manoeuvred himself many times in the position to be asked to
put his program where his mouth was. There was always the danger that his announced marvel (and how many people not only
waited for, but desired such a thing?) could result in a great, appalling fiasco. And with all this, when he carried Chess System
Tal to Paris, became the paradigm of danger and suicide: how would a human-like, very slow, speculative program do against
sheer calculators?

But then good things began to happen. First, Chess System Tal gave a very good show in Paris. It won a lot of games against
strong opposition and took 10th place among 34 competitors. You must evaluate that as a great success in as much as Chess
System Tal was not designed to play computers at all, but human beings, and that means it is prone to take risks that are useful
against our breed, but which tend to fail against the cold soul of other computers.
Even more, Chess System Tal made a good showing in the worst possible stage, a blitz tournament against programs searching
20 to 30 times more positions per second.
And even more, the Paris version was poisoned by an "almighty bug" as Chris himself described it, one that severely impaired a
lot of evaluation functions so much that he says it is incredible that Chess System Tal was capable of playing chess at all.
Second, the first customers of Chess System Tal, which was being delivered commercially precisely in the week of Paris
tournament, began to express their satisfaction.

What you see

What you first see of Chess System Tal is an envelope which says: "...while all chess programs are now strong, Chess System
Tal is more dynamic, more romantic, daring and graceful than any other..." That's what you see and that's what you get, but
before we take a look at its playing qualities, we will first tell you something about the frills. First, we can say that Chess System
Tal keeps more than a few of the good display features of its grandfather, Complete Chess System. Chris has a lot of
specialized programmers in his company, Oxford Softworks, people who really know how to make a good GUI. The result
has an air of mass market product, but what's the problem? In terms of display options, to purchase a mass market product is
an advantage: it means that it was not made by the guy who serves the coffee, that you will not get a poor CGA-style screen.
And clearly it does not mean that the engine is rubbish. Times have changed: strength became commonplace and so the frills are
now important. Maybe as time goes on you won't care too much about these things, but in the very first moments it is nice to
get something that from the very beginning tries to hypnotize you and gives you more for your money.

CST gives much more information about its internal thinking than any other program. In fact, this is due to the way as this
program was designed, in as much as many of the things he shows are new features, new evaluation functions until now either
not implemented in other programs, or that do not show. You can see not only the usual main line, score, number of nodes per
second and depth of search, but also what the program thinks about board activity, the percent of search used by tactical
extensions, hash table behaviour, degree of development, attack to king, futility search in percentage to total search and so on.
Some options let you take a look at how the program evaluates each move of a line in relation to different evaluation
components, such as mobility, development, material, etc. To any technically biased fan, that is to say almost all of us, this is a
great source of entertainment.

Clearly the Chess System Tal creators have made an effort to get an attractive interface in terms of board options, but as
usually happens many of the options are pretty to see, but impossible to use for playing a game. Fortunately, Whittington and
company surely thought the same and you always have the option to use a simple monochromatic and well designed 2-D
board. In fact, the novelty in the display lies in the "inner eye" , a kind of face that incarnates the program's ego and that
grimaces during the game expressing how he feels about his position. Depending on which layout you choose, the face seems a
mask (the Phantom of the Opera would have loved it) or the face of a moron, but in any case, if you do not want to see the
score for avoiding the temptation of also seeing the lines Chess System Tal is looking at, you always can suppose, just staring at
the smile of the inner eye, how badly your game is going and how near the defeat is.

And now the strength

And now let us see the main feature of any program and the foundation of Chess System Tal's claim of being a creature of
different kind: the chess engine. In short, I think Chris and associates have a real breakthrough here. We are talking of a
program which barely calculates, 3500 to 3700 positions per second on a 200 MHz MMX Pentium, and nevertheless gives
you a first class game and one that really has style. I cannot say the rating of Chess System Tal because, as I have said here
many times, to implement positional tests is not the kind of task I like to do and not the kind of measure I think adequate to
evaluate a program that in the end is made for playing people, not other computers and not just solve problems. My system has
been and will be just play against the monster and from that, thorough a qualitative but well founded judgement, to get my
conclusions. I say "well founded" judgement because I have and play against almost all top programs, middle class programs
and even sometimes, for the sake winning, of course, against shareware programs, old commercial programs, etc. So I have a
direct, human feeling of what a program is in comparison with his competitors.

Well, Chess System Tal is really a great opponent. It is because not only as any chess program calculates fast enough and to
out search any human being, but because of his unrelenting attacking attitude. When you are playing seriously, when you are not
going to take-back anything, this is appalling. Even if you suspect that some attack has a hidden weakness, you usually do not
have the time or the calculation power or the masterly experience to know where it is and how face it and get some advantage
of it. In other words, with Chess System Tal you face almost the same thing which happens in a real game against a very strong
human opponent. That reality includes the likelihood that the opponent can make mistakes and the agony to find them and
exploit them before the opposition can cut your throat.

In as much as in a chess game there is more expressive power than in 1000 words, let's take a look at one of the games Chess
System Tal won against me at 40 moves in one hour. Not that I played the best, far from it, but anyway the method of Chess
System Tal is instructive of his style and power.

[Event "40/1"]
[Site "?"]
[Date "1997.??.??"]
[Round "?"]
[White "Chess System Tal"]
[Black "Fernando Villegas"]
[Result "1-0"]

1. e4 e5 2. Nf3 Nc6 3. Bb5 a6 4. Ba4 Nf6 5. 0-0 Be7 6. Re1 b5 7. Bb3 d6 8. c3 0-0 9. h3 Na5 10. Bc2 c5 11. d4 Qc7 12.
d5 Bd7 13. Bg5 Rfe8 14. Nbd2 Nb7 15. a4 Kh8

{Here I began what it seemed to me a novelty. Well, bad moves are always a "novelty" for real and sound theory and at the
same time are a normal occurrence in patzer play. The idea: to exchange my king bishop or remove white's queen bishop.}

16. Kf1 Ng8 17. Bxe7 Nxe7

{OK, I got it, but now...}

18. axb5 axb5 19. Rxa8 Rxa8 20. Nh4 g6 21. Qf3 Kg7 22. Qe3 Qc8 23. f4 f6 24. fxe5 dxe5

{Otherwise the white king pawn can move and the white king bishop will aim sooner or later at g6.}

25. c4 b4

{Probably a mistake, but the effects will be seen much later.}

26. Bb3 Qc7 27. Qg3 Nd6 28. Kg1 Nf7 29. Rf1 Nd6 30. Kh1 Ne8

{Another mistake. Exchanges that follow from it are the root of hard times to come.}

31. Nf5+ Nxf5 32. exf5 g5 33. h4 h6 34. hxg5 hxg5 35. Ne4 Qd8 36. Qg4 Qe7 37. Qh5 Qf7 38. Qxf7+ Kxf7 39. Nxc5

{After his successive previous threats, CST just collects a pawn, but it is more than enough.}

39... Ke7 40. Nxd7 Kxd7 41. c5 Ra5 42. Rc1 Nc7 43. d6 Na8 44. Be6+ Kd8 45. Kg1 Ra7 46. Rc4 Ra1+

{And after that check, I felt myself entitled to resign. Nothing can stop the b and c pawns.} 1-0


CST - Paris version - is Overwhelming in Attack

by Fernando Villegas, November 10, 1997

After completion of the test I did with the killing line of Robert Sherman, I begun my own series of tournament games at 40/60
only to discover that CST is really what it is supposed to be, at my cost... I have all the top programs, with the only exception -
by now - of Schedder - and of course, not being myself more than a 2100 player, I lose 8 times of 10 after agonizing fights and
the rest are difficult draws, with a miraculous winning here and there, but you can believe me that never before I had been
mauled in such a way as CST did with me. Is not just tactical shots, but unrelenting attack and at least the Paris version running
now in my program does not launch just "wild" swings, but very well thought ones. He handles things in such a manner you
always have the feeling that "If" you only have one more time to do that last final movement, all would be fine, but... I lost three
games in a row, but they were so interesting even for me, the loser, that I scarcely finished one I was beginning the next. I was
all the time expecting the "wild" and unsound movement to punish the beast, but the opportunity never arose. In one of them, in
a ending after more than 45 moves and where he was a pawn above me, I was sure I could handle that in order to get a
technical draw, but the bastard saw it all and say goodbye... If any of you is interested in the game, I can send it even at the
cost of my face and reputation. It is just that I want to share with you my feeling that this is really something...