The Cambodian Chess Riddle

Cambodian Chess (John, 2/4 November 2012, extended on 12 November)

In the first edition (1994) of The Encyclopedia of Chess Variants, David Pritchard described a chess game allegedly played in Cambodia. His authority was a copy, now in the Pritchard archive in the Musée Suisse du Jeu, of a letter written by the late John Gollon to Philip Cohen, reporting information he had received in 1969 from a U.S. serviceman serving as an interrogator in Saigon, who in turn had received it from a Cambodian-born guerrilla officer he was questioning. The name 'P. A. Hill' had been added as a manuscript annotation to the words 'U.S. serviceman'. However, this description was challenged, and in The Classified Encyclopedia of Chess Variants (2007) I took it upon myself to omit the game from the main text and to mention it only in an editorial note, with the comment that the authority for its existence appeared to reduce to a single informant whose statements were at variance with all other known testimony.

There have since been several developments.

Immediately upon publication of the Classified Encyclopedia, I received a letter from Peter Blommers saying No, there was indeed supporting testimony; the Japanese collector Okano Shin possessed a set, and he himself had a photograph of one. He subsequently sent me photocopies of pages from several books in Japanese, by Umebayashi Isao and Okano Shin or by Okano Shin alone, with translations into English of the relevant items. There were minor differences between the games described by Hill and by Umebayashi Isao and Okano Shin as translated (the names of some of the pieces were quite different, there was a slight difference in the elephant's capturing power, and the board in the photograph had diagonal lines across it in the manner of Myanmar chess), but these seemed to me to be no more than the differences which might be expected in the recollections of separate informants; if anything, they added confidence that the testimonies of Hill and of Umebayashi Isao and Okano Shin were independent (in particular, the presence of the diagonal lines across the board did not seem likely to be a detail that somebody had invented). I therefore reported this new testimony in Variant Chess 55 (front page and page 4) and repeated it in Variant Chess 64 (pages 177-178), making clear that I accepted it as sound. Further apparent confirmation subsequently came to light in the shape of a report that Umebayashi Isao had once visited Cambodia and had bought a book on Cambodian Chess there, and I reported this also in Variant Chess 64 (page 234).

However, I was recently told by Yasuji Shimizu, initially through Peter Michaelsen and then directly, that all this appeared to have been founded on misunderstanding. The set photographed was in fact owned by Umebayashi Isao, and was not a survival but a modern reconstruction based on the information in the first edition of the Encyclopedia. Umebayashi Isao knew of no other information about the game, and he obtained the names of the game and of its pieces from a Cambodian dictionary. The descriptions of the game in the various Japanese books were again based on what appeared in the first edition of the Encyclopedia, and it is my conjecture that the difference in the elephant's capturing powers slipped in as a mistranscription or mistranslation somewhere along the way. As for the diagonal lines on the board in the photograph, Yasuji Shimizu told me that the catalogue of an exhibition held in 2002 included a photograph of a board and men for Chator (Malay Chess, 'Main Chator' in the Encyclopedia and the Classified Encyclopedia) owned by Umebayashi Isao since 1999, Chator being another game whose board features diagonal lines, and the identical graining of the wood showed the two boards to have been the same. This catalogue also included some Makruk boards, without diagonal lines, but all these had stepped rims around the outside. He therefore conjectured that when Umebayashi Isao and Okano Shin needed a board for the photograph in their book, which was published in 2000, they found the Makruk boards unsuitable because the rims came too near to the squares for men to be placed on the square corners, and so they used the Chator board.

I reported all this in a posting on 2 November, updated on 4 November, acknowledging that what I had written in issues 55 and 64 of Variant Chess appeared to have been wholly misguided, and that we were back to where we were when I was working on the Classified Encyclopedia: the game currently played in Cambodia was Makruk (Thai Chess) with one or two minor variations, the authority for the game described to P. A. Hill in 1969 appeared once more to reduce to a single informant whose statements were at variance with all other known testimony, and what had appeared to be independent confirmation of the existence of this game had in fact all been taken directly or indirectly from what appeared in the first edition of the Encyclopedia. Be it noted that Umebayashi Isao and Okano Shin were in no way to blame for the misunderstanding; a caption under the relevant photograph said 'Re-creation', and had I been able to read Japanese I would have realised this.

It then occurred to me to ask a question which might have been asked earlier: do the 'Hill' rules as given by Gollon produce a playable game? These rules can be found in the Encyclopedia and the Classified Encyclopedia and also in issue 55 of Variant Chess (downloadable via 'BESN and VC' alongside), but for present purposes the following summary is sufficient: the board is 9x9, the men are King, Rook, and Knight with their ordinary chess moves, Elephant and Official, whose moves are subsets of the king's move, and Fish, which moves and captures one step forward until it reaches the sixth rank, when it gains additional powers, and the initial array is RNEOKOENR (shades of Xiangqi!) with 9xF on the fourth rank. There is a diagram showing the initial setup on the front page of Variant Chess 55.

I quickly found that these rules do not produce a playable game, in that Black has a very simple strategy which gives White the choice of accepting a draw by repetition or sacrificing material. All he has to do is to mirror White's moves; for example, if White starts 1 Nc3, Black replies 1...Nc7, and so on. In ordinary chess, White can defeat this by giving check or by capturing the mirror-image man (for example, 1 e4 e5 2 Qh5 Qh4 3 QxQ). Here, no such maneouvre appears to be possible until it is too late. Initially, none of White's pieces can cross his fish line (apart from the knight, which will be immediately captured if it does), so to make progress White must sooner or later advance a fish, and Black's facing fish will simply take it. This will leave White a fish down without apparent compensation, which cannot be good, and Black's men (apart from his extra fish) will still mirror White's so he can continue with the mirroring strategy if he wants to. True, once White has advanced a fish he will be able to bring a piece to the square it has vacated, but this square is now commanded by Black's extra fish, so how will this help him? Suppose 1 Na3 Na7 2 Ke2 Ke8 3 Fc5 Fxc5 4 Nc4, intending to meet 4...Nc6 with 5 Nxd6+ giving check and preventing Black from continuing the mirror play; yes, but why should Black play the mirror move 4...Nc6? He can play 4...Fxc4 instead, and his fish ahead has become a knight ahead.

Had the game ever been played at all seriously, this simple non-losing defensive strategy would have come to light. I conclude that not only does the authority for the game reduce to a single informant whose statements are at variance with all other known testimony, but that something must have gone wrong somewhere along the line and in fact no such game existed. Whether the error lay with Hill's Cambodian-born informant, who had perhaps seen makruk played in childhood but had never played it himself and had misremembered the details, or whether Hill himself did not properly understand what he was being told, or whether Hill and Gollon got their lines crossed, we can only guess. I don't know how the U.S. military obtained its Vietnamese-speaking interrogators, but I suspect that it took the best linguists among its draftees, sent them on a crash course in Vietnamese, and set them to work, and that their knowledge of the language was often little more than was needed for the purposes of military interrogation. But unless Hill or his informant reads this and comes forward, we shall never know.

My thanks to Jean-Louis Cazaux for trying out this strategy against the Zillions implementation of the game, and verifying that the computer (which won't have thrown a fish away unless it could see compensation within its look-ahead horizon) did indeed give up and concede the draw.