One hundred years ago, the first chess computer

[sje]

One hundred years ago in 1912, the first chess computer was designed and built by Leonardo Torres y Quevedo of Spain. It played the white side of KRK and could always force a win although with an occasional departure from optimal play. In a way, it was also the first chess tablebase.



Links:

http://en.wikipedia.org/wiki/El_Ajedrecista

http://en.wikipedia.org/wiki/Leonardo_Torres_y_Quevedo

http://blog.modernmechanix.com/2006/03/ ... -opponent/

[Ajedrecista]

Hello:

I have searched in a Spanish newspaper archive (ABC) and there is an article (written in Spanish, of course) dated on 25th March, 1953, with an improved version of El Ajedrecista. Links:

http://hemeroteca.abc.es/nav/Navigate.e ... 5/015.html

http://hemeroteca.abc.es/nav/Navigate.e ... 5/017.html

The first photograph (attached with a bunch of drawings) shows the complexity inside the machine, while the second is an exterior view.

The article begins with a brief review of older machines, continues with the inner workings of El Ajedrecista and finishes explaining the output of the moves while recognizing the huge merit of Torres Quevedo. Please enjoy!

Regards from Spain.

Ajedrecista.

[sje]

The machine's algorithm is explained in the text at the bottom of the second document. ("el rey" = "the king", "la torres" = "the rook", "un paso" = "one square", etc.) There are six possible move actions; the exact move made for an action is dependent on the relative positions of the men.

The machine could not handle KQK due to incomplete stalemate detection for that class. Still, an amazing piece of machinery.

[sje]

I'm sorely tempted to code its move selection algorithm and run it through each of the 187,328 KRK WTM positions checking it against the KRK WTM tablebase.

The Wikipedia article states that there were some starting positions where the machine could exceed the fifty move limit, but I believe that claim is false. Reasoning: the machine makes progress of a sort on each move. The longest WTM mate is 16 moves, so even if two out of three machine moves waste a move, there's still two moves left under the fifty move limit.

CookieCat says:

Code: Select all

[Event "Unknown event"]
[Site "Unknown site"]
[Date "2012.01.05"]
[Round "-"]
[White "Unknown player"]
[Black "Unknown player"]
[Result "1-0"]
[FEN "7K/8/8/8/8/8/2k5/1R6 w - - 0 1"]

1 Ra1 Kb2 2 Ra8 Kc2 3 Ra3 Kb1 4 Kg7 Kb2 5 Rg3 Kc1 6 Kf6 Kc2 7 Ke5 Kb2 8 Kd4 Kc2 9 Ra3 Kb2 10 Rc3
Ka2 11 Kc4 Kb2 12 Kb4 Ka2 13 Rb3 Ka1 14 Kc3 Ka2 15 Kc2 Ka1 16 Ra3# 1-0

[bob]

Hello:

I have searched in a Spanish newspaper archive (ABC) and there is an article (written in Spanish, of course) dated on 25th March, 1953, with an improved version of El Ajedrecista. Links:

http://hemeroteca.abc.es/nav/Navigate.e ... 5/015.html

http://hemeroteca.abc.es/nav/Navigate.e ... 5/017.html

The first photograph (attached with a bunch of drawings) shows the complexity inside the machine, while the second is an exterior view.

The article begins with a brief review of older machines, continues with the inner workings of El Ajedrecista and finishes explaining the output of the moves while recognizing the huge merit of Torres Quevedo. Please enjoy!

Regards from Spain.

Ajedrecista.

I have a signed photo of Claude Shannon with a KRK machine he developed in the 1950's... I was not aware there was one done earlier... Shannon's is based on relays and such, not mechanics.

[Gerd Isenberg]

and some cpw links on that topic

https://chessprogramming.wikispaces.com/El+Ajedrecista
https://chessprogramming.wikispaces.com ... +y+Quevedo

According to Henri Vigneron it could take up to 63 moves until mate.

Henri Vigneron (1914). Les Automates in La Nature, translation: Robots from Computer Chess Compendium by David Levy (Editor) pp: 273 - 278; also printed in David Levy and Monroe Newborn, All About Chess and Computers, Springer 1982, pp. 14-23

[sje]

I have a signed photo of Claude Shannon with a KRK machine he developed in the 1950's... I was not aware there was one done earlier... Shannon's is based on relays and such, not mechanics.

I saw a photograph of Shannon's electromechanical chess endgame computer, printed in Adventures in Chess by Edward Lasker. The machine was about the size of a large footstool. The playing board was about 30 cm on a side and it looked like each square had a switch and a lamp. The report by Lasker is that Shannon's machine could handle endgames with up to six men.

I would dearly like to get the schematic. Was Shannon working for Bell Labs at the time? Maybe the machine is still intact and is sitting in a storeroom somewhere in Murray Hill.

I have a link to a surplus electronics store that sells the relays that were used in telephone switching offices prior to the Transistor Age. These relays can be had by the thousand at only US$0.25 each, so it might not be too expensive to re-create the machine and perhaps add some improvements.

[sje]

Here's a computer which has its logic implemented with 415 DPDT relays: http://web.cecs.pdx.edu/~harry/Relay/

[Gerd Isenberg]

According to Henri Vigneron it could take up to 63 moves until mate.

Wrong! Sorry for the miss-information. I re-read the paper in the Computer Chess Compendium and found nothing about 63 moves, but Levy mentioned in the chapter foreword less than 50 moves. I don't know where I read the 63 somewhere, its still from the early cpw times I did not used references and quotes carefully.

edit

Hans-Peter Ketterling mentions more than 60 moves
http://www.schachklub-tempelhof.de/?q=quevedo

Aus der ungünstigsten Stellung erfordert die Mattführung gegen bestes Gegenspiel etwas über 60 Züge, was nach den FIDE-Regeln remis durch Überschreiten der Fünfzig-Züge-Grenze wäre.

guess we need to implement the algorithm

[bob]

I have a signed photo of Claude Shannon with a KRK machine he developed in the 1950's... I was not aware there was one done earlier... Shannon's is based on relays and such, not mechanics.

I saw a photograph of Shannon's electromechanical chess endgame computer, printed in Adventures in Chess by Edward Lasker. The machine was about the size of a large footstool. The playing board was about 30 cm on a side and it looked like each square had a switch and a lamp. The report by Lasker is that Shannon's machine could handle endgames with up to six men.

I would dearly like to get the schematic. Was Shannon working for Bell Labs at the time? Maybe the machine is still intact and is sitting in a storeroom somewhere in Murray Hill.

I have a link to a surplus electronics store that sells the relays that were used in telephone switching offices prior to the Transistor Age. These relays can be had by the thousand at only US$0.25 each, so it might not be too expensive to re-create the machine and perhaps add some improvements.

I believe that "6 man ending" statement was in error. I talked to Claude Shannon when he gave me this photo and he was emphatic that it was a KR vs K and that it would always mate with the rook, although not always "optimally".

That thing you are describing seems to be about the right size. In the photo I have, he is kneeling down beside the thing with someone else in the photo. When I get to the office tomorrow I will see if my memory is correct about the specifics of the photo and what the thing could do.

More to follow...