"Peasant Revolt" Endgame Chess
Posted: Sat Sep 02, 2017 6:50 am
Inspired by this thread:
http://www.talkchess.com/forum/viewtopic.php?t=65051
and keeping the name of the variant given by Jon Fredrik Åsvang, I modified the most probably decided in favor of Black opening position to this one:
[D]1nn1k1n1/4p3/8/8/8/PPPPPPPP/8/4K3 w - - 0 1
If a player stronger than me shows how to play correctly this variant and the theoretical result of it, it would be helpful. As I am more of a patzer in deciding on the decisiveness of the Black advantage here, I left Komodo 11.2.2 play self-games. I built a a 3-mover opening book containing 1000+ positions using the same Komodo, as it randomizes better, and has the "Variety" parameter. The EPD book of 1000+ positions is here:
http://s000.tinyupload.com/?file_id=009 ... 2378736001
The unbalance here is accentuated, and there are two nice properties to this Endgame Chess variant: the position is borderline Black win versus Draw (about 50% to 50%), and is very sensitive to 6-men Syzygy (from SSD). These borderline positions are desirable in order to separate in strength engines, a slight slip can lead from draw to loss and vice-versa. Using pentanomial variance here, the errors one sees in Cutechess (trinomial errors) are really 1.7-1.8 times smaller. It is important to not have any kind of adjudication, engines misevaluate the positions during the game, and might claim adjudication, even when the game should go on.
On two threads, from that 3-mover book, Komodo 11.2.2 with Syzygy 6-men default Probe Depth=2 from SSD, the results were in 100 self-games:
10''+ 0.1''
50 Black wins, 50 draws
60''+ 0.6''
51 Black wins, 49 draws
360''+ 3.6''
44 Black wins, 56 draws
Really hard to decide if there is a trend towards one outcome or the other.
Komodo 11.2.2 (Contempt=0) beats easily Stockfish dev. in 100 games at 60''+ 0.6''. Syzygy 6-men from SSD for both:
60''+ 0.6''
Score of Stockfish dev vs Komodo 11.2.2: 30 - 65 - 105 [0.412] 200
ELO difference: -61.43 +/- 33.17
Finished match
95 Black wins, 105 draws
Also, the game is extremely sensitive to Syzygy, even better than the most sensitive to Syzygy endgame suites I had:
60''+ 0.6''
Score of Komodo Syzygy vs Komodo NO TB: 71 - 28 - 101 [0.608] 200
ELO difference: 75.88 +/- 33.89
Finished match
99 Black wins, 101 draws
The real errors here using the pentanomial are about 18-19 ELO points instead of the shown by Cutechess-Cli trinomial 33-34 ELO points.
All in all, this borderline variant helps in seeing the engines dealing with accentuated endgame disbalances and benefiting from Syzygy, 6-men from SSD in my case. It has a very high sensitivity to these two issues
http://www.talkchess.com/forum/viewtopic.php?t=65051
and keeping the name of the variant given by Jon Fredrik Åsvang, I modified the most probably decided in favor of Black opening position to this one:
[D]1nn1k1n1/4p3/8/8/8/PPPPPPPP/8/4K3 w - - 0 1
If a player stronger than me shows how to play correctly this variant and the theoretical result of it, it would be helpful. As I am more of a patzer in deciding on the decisiveness of the Black advantage here, I left Komodo 11.2.2 play self-games. I built a a 3-mover opening book containing 1000+ positions using the same Komodo, as it randomizes better, and has the "Variety" parameter. The EPD book of 1000+ positions is here:
http://s000.tinyupload.com/?file_id=009 ... 2378736001
The unbalance here is accentuated, and there are two nice properties to this Endgame Chess variant: the position is borderline Black win versus Draw (about 50% to 50%), and is very sensitive to 6-men Syzygy (from SSD). These borderline positions are desirable in order to separate in strength engines, a slight slip can lead from draw to loss and vice-versa. Using pentanomial variance here, the errors one sees in Cutechess (trinomial errors) are really 1.7-1.8 times smaller. It is important to not have any kind of adjudication, engines misevaluate the positions during the game, and might claim adjudication, even when the game should go on.
On two threads, from that 3-mover book, Komodo 11.2.2 with Syzygy 6-men default Probe Depth=2 from SSD, the results were in 100 self-games:
10''+ 0.1''
50 Black wins, 50 draws
60''+ 0.6''
51 Black wins, 49 draws
360''+ 3.6''
44 Black wins, 56 draws
Really hard to decide if there is a trend towards one outcome or the other.
Komodo 11.2.2 (Contempt=0) beats easily Stockfish dev. in 100 games at 60''+ 0.6''. Syzygy 6-men from SSD for both:
60''+ 0.6''
Score of Stockfish dev vs Komodo 11.2.2: 30 - 65 - 105 [0.412] 200
ELO difference: -61.43 +/- 33.17
Finished match
95 Black wins, 105 draws
Also, the game is extremely sensitive to Syzygy, even better than the most sensitive to Syzygy endgame suites I had:
60''+ 0.6''
Score of Komodo Syzygy vs Komodo NO TB: 71 - 28 - 101 [0.608] 200
ELO difference: 75.88 +/- 33.89
Finished match
99 Black wins, 101 draws
The real errors here using the pentanomial are about 18-19 ELO points instead of the shown by Cutechess-Cli trinomial 33-34 ELO points.
All in all, this borderline variant helps in seeing the engines dealing with accentuated endgame disbalances and benefiting from Syzygy, 6-men from SSD in my case. It has a very high sensitivity to these two issues