The first talks began around Kasparov's games against Deep Blue in 1996, and after long talks with Jaume Farràs in the Barcelona town of Moià, explaining to him the technique of computers to play chess, since I am a computer scientist. Debates began to turn to how to make chess more difficult for the machine.
Well, nowadays they seem like naive origins in the face of new techniques like those of AlphaZero, but at that time the machine had not yet definitively beaten Kasparov, and the discussion made more sense.
As a computer scientist, discussions then focused on machine
power were not enough, and using
heuristics often requires transferring knowledge from experts to algorithms, a point that brings difficulties and shortcomings. Nowadays, the machine acquires this knowledge by learning, and games such as chess have been very suitable problems to consolidate advances in this field; we should be aware that human skills also have to be developed as machines incorporate advances, and games are a very suitable field to develop certain skills.
The ideas for C'escacs began when a little research led me to the
discovery of Gliński's hexagonal chess, and the first tries were made with a cardboard hexagonal board that I built. Thus began our first steps into the world of hexagonal chess. The hexagonal geometry increased the search space significantly, and could be further increased by making a larger board... and if some pieces incorporated more moves it would also help; We already knew Capablanca's chess, and the Chancellor's and Archbishop's pieces.
Finally, in 2007, the talks and ideas materialised on a cardboard board, and we started the first games with a primitive version of C'escacs, which initially used a piece called Prince, with the same moves as the king, but without royalty. For this first version there was a program in a Java Applet to test the moves, save and visualise the games from a web page in Geocities (a service that Yahoo! closed down), where there were also published the rules.
We already had the first version of the game and the website when we contacted Dave McCooey, who very kindly listened to our ideas. His comments were of great help in finalising the initial design of C'escacs, particularly the pawn moves, which had not yet been defined. As a result of these email conversations the pawn was defined with the option to advance double and triple steps in the starting hexes; much later the knight ride was also incorporated, a difficult decision, but clearly necessary for similar reasons to those of the pawn movement.
McCooey also criticised us on two other points, which we did not change immediately, as we had to mature them:
- The Prince's piece, which we used in the early versions, warned us that it was a very powerful piece. Moreover, on the 169 hexes board it was very slow. His diagnosis was absolutely right, and it was the weak point of the game. They still had no alternative, but the idea was to strengthen the power of the pawns, which could not build solid chains. Seeing that it was a piece of little use, we put aside the original idea of adding some kind of royalty, as in Tamerlane's chess, and began to think of alternatives. We also did some testing by adding the knight move and naming the new piece Lion, but the knight also suffered from a speed problem; adding more power to the piece was not the aim, and the pawn problem was not solved.
- He also criticised castling, saying that it contributed nothing. In fact this criticism later led to the incorporation of the King's Leap, the origin of castling, both in orthodox chess and in C'escacs, as the idea arose from the historical existence of this move. With the addition of the King's Leap we make a nod to the history of chess, including the reference to the old King's Leap of European chess, and the existence of both moves makes the mainly evasive function more evident. There was no point in forcing a castling, indeed, but the Rook's move can, on occasion, represent an additional defence. Together with the King's Leap we incorporate the double castling, which allows a player to advance moves in a defensive position, but also allows a player with a tactical advantage to take an aggressive position, connecting with one move both Rooks on a column, although it is neither its function nor the objective of this move.
The Java Applet was successful and I was repeatedly asked to incorporate an engine that would allow games to be played against the machine; unfortunately the Applet was only intended to force valid moves and the programming had to be rewritten, and we had no time to devote to the project.
Dave McCooey carried out the endgame analysis on the 169-hexes board, confirming that two knights cannot mate a lone king, and also the endgame analysis of the Prince, a piece that has now been discarded in C'escacs.
McCooey's other contribution was with regard to notation, drawing our attention to the naming of the columns, suggesting that we use for the central columns the same names as Gliński's columns, extending the columns left and right, but leaving the letter F for the central column. I found this point very interesting, but Gliński's notation is problematic, and I observed that by numbering the lines and advancing the numbering through the columns two by two the notation was much more coherent. This notation arose spontaneously, and on further research we found references in this coordinate system, obviously already invented, although it is rarely used, except in technological fields, given its suitability.
In 2008 I acquired an Omega Chess (difficult to get in Spain). This chess variant is interesting, and incorporates a piece called Champion. We played a few games of C'escacs using a hexagonal adaptation of this piece, instead of the Prince, but it was cumbersome because of the diagonal jumps and, possibly because we were not well versed in Omega Chess, it was of little use to us. Making some adaptations to obtain simpler moves than the Champion is almost a requirement to use it in hexagonal chess games. We then called it Almogavar, but it was not the piece that C'escacs needed either.
The ultimate alternative came with Elephants, just the piece we wanted, which we adapted from Eastern (Burmese) chess by researching the history of chess, and finding that Sittuyin is still played today. The name Elephant has drawbacks in some languages, notably Arabic and Russian, as the Bishop is called Elephant, for historical reasons. In these languages the name Halberdier seems to us to be the most appropriate and consistent.
The game was finally defined in 2018, although it was not published on the website until 2022, once all the changes had been thoroughly tested and confirmed; another change incorporated was the position of the Dragon, now definitely called Wyvern, although the name Dragon is still accepted. This change of position means that the piece has no weaknesses in the openings, delays its participation in the game in a positive way, and moves the Bishops forward, giving them more ease of movement at the start of the game.
As for the
early aspirations of making chess
difficult for the machine, perhaps it does not make much sense today, but it could be a
new way to play chess, if the machine
help is a factor, and people direct the machine's strategy. This would already be a practically different game from chess, where the assisted players would have different skills than a chess player of today, and probably the game of C'escacs is not yet complex enough for such a challenge... But I have a new project of an extension of C'escacs to a virtual game, to be played only on computer. Hardly on real boards, as it would require twenty-six C'escacs boards (that's 4_394 hexes). Lovers of strategy games would do well to remember that chess is king.