Matthies/RubiChess

Rubi has problems solving extreme positions

Matthies opened this issue · 1 comments

R6R/3Q4/1Q2Q3/4Q3/2QB3Q/Q4Q2/Np1Q4/1k1N1K2 w - - 3 4

Master:

position fen R6R/3Q4/1Q2Q3/4Q3/2QB3Q/Q4Q2/Np1Q4/1k1N1K2 w - - 3 4
go
info depth 1 seldepth 1 multipv 1 time 200 score mate 0 lowerbound nodes 360350 nps 1797120 tbhits 0 hashfull 0 pv f3b3
info depth 1 seldepth 1 multipv 1 time 400 score mate 0 lowerbound nodes 750216 nps 1872896 tbhits 0 hashfull 0 pv f3b3
info depth 1 seldepth 1 multipv 1 time 600 score mate 0 lowerbound nodes 1137829 nps 1895424 tbhits 0 hashfull 0 pv f3b3
info depth 1 seldepth 1 multipv 1 time 800 score mate 0 lowerbound nodes 1542821 nps 1927168 tbhits 0 hashfull 0 pv f3b3
info depth 1 seldepth 1 multipv 1 time 1000 score mate 0 lowerbound nodes 1928939 nps 1927168 tbhits 0 hashfull 0 pv f3b3
info depth 1 seldepth 1 multipv 1 time 1200 score mate 0 lowerbound nodes 2317095 nps 1929216 tbhits 0 hashfull 0 pv f3b3

1.7.3:

position fen R6R/3Q4/1Q2Q3/4Q3/2QB3Q/Q4Q2/Np1Q4/1k1N1K2 w - - 3 4
go
info depth 1 seldepth 2 multipv 1 time 3 score mate -82 nodes 183 nps 0 tbhits 0 hashfull 1 pv f3b3
info depth 2 seldepth 3 multipv 1 time 6 score mate 1 nodes 382 nps 0 tbhits 0 hashfull 1 pv a3b2
info depth 3 seldepth 3 multipv 1 time 6 score mate 1 nodes 563 nps 0 tbhits 0 hashfull 1 pv a3b2
...
info depth 27 seldepth 40 multipv 1 time 217 score mate 1 nodes 431497 nps 1984512 tbhits 0 hashfull 49 pv a3b2
info depth 28 seldepth 37 multipv 1 time 376 score mate -117 nodes 730397 nps 1938432 tbhits 0 hashfull 114 pv d2d3  b1a1  d3b1  a1b1  c4d3  b1a1  d3b1  a1b1  d1b2  b1c2  d4e3  c2b1  b2a4  b1c2  e3d4  c2d2  f1f2  d2c2  f2g2  c2d2  h4h2  d2c2  d7d6  c2d2  a8g8  d2c2  h8h6  c2d2  g8g7  d2c2  h6g6  c2d2  g7h7
info depth 29 seldepth 34 multipv 1 time 659 score mate -113 nodes 1317468 nps 1996800 tbhits 0 hashfull 185 pv d2d3  b1a1  d3b1  a1b1  c4d3  b1a1  d3b1  a1b1  d1b2  b1c2  d4e3  c2b1  b2a4  b1c2  e3d4  c2d2  f1f2  d2c2  f2g2  c2d2  h4h2  d2c2  d7d6  c2d2  a8g8  d2c2  h8h6  c2d2  g8g7  d2c2  h6g6  c2d2  g7h7
info depth 29 seldepth 36 multipv 1 time 919 score mate -110 nodes 1891274 nps 2056192 tbhits 0 hashfull 240 pv d2d3  b1a1  d3b1  a1b1  a3d3  b1a1  d3b1  a1b1  d1b2  b1a1  c4a6  a1b1  a6d3  b1a1  d3a3  a1b1  b2a4  b1c2  a8a6  c2d2  f1f2  d2c2  a6a8  c2d2  h4h2  d2c2  d7d6  c2d2  f2g2  d2c2  a8a6  c2d2  a6a5  d2c2  a5a8

It seems to be an overflow problem in kingdanger evaluation. If I remove some of the white queens and use this position
R6R/3Q4/8/4Q3/3B3Q/5Q2/Np1Q4/1k1N1K2 w - - 3 4
Rubi finds mate in 1.