TalkChess.com

Finally my mini-Shogi engine Shokidoki managed to win the UEC Cup, which was played this Sunday in Tokyo. It was a close finish: after a 9-round round robin Shokidoki tied with GA-Sho!!!!! for 1st/2nd place with 8 wins/1 loss, while Ferdinand Mosca's program Lima tied with the former long-term champion 1/128 Rigan for 3rd/4th place with 7 wins/2 losses. So playoffs were necessary. In these play-offs Shokidoki beat GA-Sho!!!!! (for the second time, as this was also GA's only loss in the round robin). While 1/128 Rigan beat Lima (although in the round robin it had lost to it and GA, while it had beaten Shokidoki). So the final result was

1) Shokidoki
2) GA-Sho
3) 1/128 Rigan
4) Lima

Tony Hecker's program TJshogi5x5 finished 5th, losing against the top four but beating all others.

Congratulations!
Does it mean Shokidoki is current #1? I remember you won a gold medal recently in world championships?
Lima also did very well, only 1 point behind.

I'm surprised that there are no draws, is this specific to mini shogi or was it that gaps between engines were too large? Top engines seem close in strength.

Congratulations.

Congratulations H.G.!

Congrats HG.

Is this the official site?

http://minerva.cs.uec.ac.jp/~uec55shogi ... tournament

mar wrote:Congratulations!
Does it mean Shokidoki is current #1? I remember you won a gold medal recently in world championships?

Shokidoki won gold medal at the ICGA Computer Olympiad this year, but there were no Japanese there. In 2013 it won gold in Yokohama, before 1/128 Rigan (which won the UEC Cup in 2010-2014, and gold in the Olympiad of 2010).

Last year 1/128 Rigan lost to Sjaak II, but was still 1 point before Shokidoki and Lima, because it beat both of them, and Shokidoki also lost to Sjaak II and beat Lima. Lima then won the play-offs and ended 3rd. The loss against Rigan that year was my fault, not Shokidoki's. I had made a typo in the opening book, so it came out of the opening after 3 moves with a score of -7.

I'm surprised that there are no draws, is this specific to mini shogi or was it that gaps between engines were too large? Top engines seem close in strength.

The rules of mini-Shogi do not allow draws. After 4-fold repetition white loses, stalemate is a win, and there is no move limit.

hgm wrote:Finally my mini-Shogi engine Shokidoki managed to win the UEC Cup, which was played this Sunday in Tokyo. It was a close finish: after a 9-round round robin Shokidoki tied with GA-Sho!!!!! for 1st/2nd place with 8 wins/1 loss, while Ferdinand Mosca's program Lima tied with the former long-term champion 1/128 Rigan for 3rd/4th place with 7 wins/2 losses. So playoffs were necessary. In these play-offs Shokidoki beat GA-Sho!!!!! (for the second time, as this was also GA's only loss in the round robin). While 1/128 Rigan beat Lima (although in the round robin it had lost to it and GA, while it had beaten Shokidoki). So the final result was

1) Shokidoki
2) GA-Sho
3) 1/128 Rigan
4) Lima

Tony Hecker's program TJshogi5x5 finished 5th, losing against the top four but beating all others.

Code: Select all

Cross table:

                  S G R L T S K T T B
1.   Shokidoki    # 1 0 1 1 1 1 1 1 1  8
2.   GA-Sho!!!!!! 0 # 1 1 1 1 1 1 1 1  8
3.   1/128 Rigan  1 0 # 0 1 1 1 1 1 1  7
4.   Lima         0 0 1 # 1 1 1 1 1 1  7
5.   TJshogi5x5   0 0 0 0 # 1 1 1 1 1  5
6/7. Sleeping     0 0 0 0 0 # 0 1 1 1  3
6/7. Komachan     0 0 0 0 0 1 # 1 0 1  3
8.   Triple Crown 0 0 0 0 0 0 0 # 1 1  2
9/10.Tokin-Chan   0 0 0 0 0 0 1 0 # 0  1
9/10.Broaden55    0 0 0 0 0 0 0 0 1 #  1

Congrats I knew Shokidoki is very strong and this time finally took the top spot. Missing here are Sjaak and Nebiyu to complete the winboard gang .

Gerd Isenberg wrote:Congrats HG.

Is this the official site?

http://minerva.cs.uec.ac.jp/~uec55shogi ... tournament

Yes it is and just click the link for 9th UEC Cup.

Congratulations! In your writeup, GA-Sho!!!!!!'s punctuated name makes you sound very enthusiastic about your results.

How tied is Shokidoki's strength to the smaller board size? Do you plan to assail the standard 9x9 board in the future?

The GA-Sho engine is new, and the exclamation points seem to be part of its name. ( http://minerva.cs.uec.ac.jp/~uec55shogi ... 2%E7%B2%F1 ) It was always 1/28 Rigan that was the big opponent these past years.

Shokidoki already does 9x9 Shogi, but (as expected) it sucks at that. It was in fact originally written as a 9x9 Shogi for the occasion of the 2010 ICGA Olympiad in Kanazawa. Only at the Olympiad I learned that mini-Shogi existed, and was actually on the program. I adapted the engine to it overnight in my hotel, and it finished second. In the 9x9 Shogi it only won two games. (But that was my goal: not lose everything.)

Regular Shogi is a highly knowledge-driven game, and state of the art in computer Shogi is to use knowledge-extraction algorithms from huge databases of human professional games. Shokidoki does none of that. It just has a simplistic evaluation written by someone who doesn't play the game at all (namely me). In mini-Shogi the other players face the same problem, though, as the game was invented as recent as 1974, and huge game databases on it do not exist. It is much more tactical than 9x9 Shogi, because of the absence of a closed rank of Pawns.

Since 2010 I have been optimizing the engine for mini-Shogi, but in general this benefits its 9x9 Shogi capabilities as well. The previous version was only slightly weaker there then the well-known Shogi engine SPEAR (scores >40% against it in 20-min games, >50% in 1-min games). Which is not nearly world top, but written by a strong Shogi player, and has an evaluation of ~200MB.

And Shokidoki also plays 6x6 (Judkins) Shogi and 7x7 Tori Shogi (the latter not being a sub-set of regular Shogi, but having completely different pieces).

Since then I