The FFO endgame test suite

On this page you find a set of endgame test positions compiled by Marc Tastet of the French Federation of Othello . Most of the top programs have tried these positions and the results are published in FForum, the newsletter of the FFO.

My own opinion on the test suite is as follows: The positions are too easy in the sense that good programs find the correct moves in almost all positions in a few seconds using their midgame search and evaluation. On the other hand, a standardized set of test positions makes it possible to size up the competition. I have used the test suite to evaluate modifications of Zebra's endgame code.

If you are a new programmer and have only recently started working on an Othello program: These positions are hard, it takes 2 hours and 6 minutes for Zebra to solve all positions for exact score on a Thunderbird/1333. To the best of my knowledge, this is the best performance reported for the test.

All the positions:

#40, black to move#41, black to move#42, black to move#43, white to move
Turner vs MonnomEclipse vs LogistelloPenloup vs ShamanBrightwell vs Suekuni
Bruxelles 1997Internet (7-12) 1996Cambridge 1998WC 1997 (final)
#44, white to move#45, black to move#46, black to move#47, white to move
Shaman vs TastetTamenori vs ShamanCaspard vs JuhemBrightwell vs Tastet
WC 1995WC 1995 (final)FC 1994WC 1997
#48, white to move#49, black to move#50, black to move#51, white to move
Brightwell vs TastetLazard vs CaliNicolet vs FeinsteinTastet vs Caspard
Paris Open 1997 FC 1991WC 1996WC 1997
#52, white to move#53, black to move#54, black to move#55, white to move
Edmead vs LazardNicolet vs MurakamiLogistello vs BrutusLogistello vs Hannibal
Cambridge 1996WC 1996 (final)IOS Tournaments 1997Internet (1-6) 1997
#56, white to move#57, black to move#58, black to move#59, black to move
Logistello vs HannibalTaniguchi vs RalleLogistello vs MurakamiTastet vs Parsons
Internet (1-6) 1996WC 1984 (final)Match 1997MSO 1997


Zebra's results

For each position, the following information is requested:

Note: For some positions, there are several optimal moves:

Zebra has verified that these are the only positions with several optimal moves and that there are no other optimal moves.

Zebra configuration (thanks to Dan Lidström, he performed the test):

Complete logs are available here, a summary is given below:

Perfect play determination
#DateScoreTime lockTime exactNodes exactNPSLine
4001/02/03+380.8s2.4s19.2M8026ka2 b1 c1
4101/02/03+00.1s3.2s20.8M6600kh4 a3 a2
4201/02/03+60.8s3.6s28.6M7912kg2 h1 c2
4301/02/03-120.1s3.9s23.7M6163kc7 h4 h5
4401/02/03-140.1s3.7s22.3M5950kd2 g5 b8
4501/02/03+62.3s58.7s472M7993kb2 c1 g5
4601/02/03-81.2s9.9s66.6M6685kb3 c1 b1
4701/02/03+40.1s4.0s23.6M5920kg2 b8 b7
4801/02/03+280.2s53.3s361M6794kf6 g5 g3
4901/02/03+160.3s48.5s380M7850ke1 h4 g6
5001/02/03+100.6s294s2.09G7000kd8 e8 g8
5101/02/03+60.3s51.5s324M6267ke2 h2 g7
5201/02/03+029.8s54.9s338M6241ka3 f2 e1
5301/02/03-23.8s430s2.99G6893kd8 b6 a5
5401/02/03-2123s657s4.62G7000kc7 g6 g5
5501/02/03+00.9s2411s14.9G5998kg6 h7 c1
5601/02/03+20.5s267s1.31G4856kh5 h4 h3
5701/02/03-10167s406s2.73G6755ka6 b5 b4
5801/02/03+40.6s465s2.53G5510kg1 f2 h3


Last modified November 22, 2004 by Gunnar Andersson