Koivisto 8.6 has tablebase issue

[Jouni]

[fen]8/5R2/7P/8/6k1/1K4p1/6r1/8 b - - 0 99[/fen]

No tablebases:
...
99...Rh2 100.h7 g2 101.Rg7+ Kf3 102.Rf7+ Ke3 103.Rg7 Kf2 104.Rf7+ Kg1 105.Rg7 Rh4 106.Rg8 Rxh7 107.Kc3 Kh2 108.Kd3 g1Q 109.Rxg1 Kxg1 110.Ke3 Kg2 111.Ke4 Kf2 112.Kd5 Rh4 113.Ke6 Kg3 114.Kd6 Kg2 115.Kd5 Kf3 116.Kc5
-+ (-4.72) Depth: 30/51 00:00:04 8906kN
99...Rh2 100.h7 g2 101.Rg7+ Kf3 102.Rf7+ Ke3 103.Rg7 Kf2 104.Rf7+ Kg1 105.Rg7 Rh4 106.Rg8 Rxh7 107.Kc3 Kh2 108.Kd3 g1Q 109.Rxg1 Kxg1 110.Ke3 Kg2 111.Ke4 Kf2 112.Kd5 Rh4 113.Kc5 Rh5+ 114.Kb4 Rh6 115.Kc3 Kg2 116.Kc4 Kh2 117.Kb3 Kh3 118.Kc4
-+ (-4.72) Depth: 31/51 00:00:04 9083kN
99...Rh2 100.h7 g2 101.Rg7+ Kf3 102.Rf7+ Ke3 103.Rg7 Kf2 104.Rf7+ Kg1 105.Rg7 Rh4 106.Rg8 Rxh7 107.Kc3 Kh2 108.Kd3 g1Q 109.Rxg1 Kxg1 110.Ke3 Rd7 111.Ke2 Kh1 112.Ke3 Kg2 113.Ke4 Kf2 114.Ke5 Kf3 115.Ke6 Rd1 116.Ke5 Rd2 117.Ke6 Ke4 118.Kf6
-+ (-4.81) Depth: 32/57 00:00:06 12147kN

Tablebases:

99...Rh2
= (0.01) Depth: 9 00:00:00 0kN

Yes I don't use full 6 piece set, but engine should work correctly. I use KRPKRP.

[Colin-G]

Koivisto 7.10 has no problem with this position, quickly getting to > +70 score using 5 man tablebases and linux xboard.

[Jouni]

Koivisto 7 has same problem. Most engines like SF and Komodo analyze OK with 5 piece + partial 6 piece. This was fixed some 8 years ago in SF.

[Luecx]

[fen]8/5R2/7P/8/6k1/1K4p1/6r1/8 b - - 0 99[/fen]

No tablebases:
...
99...Rh2 100.h7 g2 101.Rg7+ Kf3 102.Rf7+ Ke3 103.Rg7 Kf2 104.Rf7+ Kg1 105.Rg7 Rh4 106.Rg8 Rxh7 107.Kc3 Kh2 108.Kd3 g1Q 109.Rxg1 Kxg1 110.Ke3 Kg2 111.Ke4 Kf2 112.Kd5 Rh4 113.Ke6 Kg3 114.Kd6 Kg2 115.Kd5 Kf3 116.Kc5
-+ (-4.72) Depth: 30/51 00:00:04 8906kN
99...Rh2 100.h7 g2 101.Rg7+ Kf3 102.Rf7+ Ke3 103.Rg7 Kf2 104.Rf7+ Kg1 105.Rg7 Rh4 106.Rg8 Rxh7 107.Kc3 Kh2 108.Kd3 g1Q 109.Rxg1 Kxg1 110.Ke3 Kg2 111.Ke4 Kf2 112.Kd5 Rh4 113.Kc5 Rh5+ 114.Kb4 Rh6 115.Kc3 Kg2 116.Kc4 Kh2 117.Kb3 Kh3 118.Kc4
-+ (-4.72) Depth: 31/51 00:00:04 9083kN
99...Rh2 100.h7 g2 101.Rg7+ Kf3 102.Rf7+ Ke3 103.Rg7 Kf2 104.Rf7+ Kg1 105.Rg7 Rh4 106.Rg8 Rxh7 107.Kc3 Kh2 108.Kd3 g1Q 109.Rxg1 Kxg1 110.Ke3 Rd7 111.Ke2 Kh1 112.Ke3 Kg2 113.Ke4 Kf2 114.Ke5 Kf3 115.Ke6 Rd1 116.Ke5 Rd2 117.Ke6 Ke4 118.Kf6
-+ (-4.81) Depth: 32/57 00:00:06 12147kN

Tablebases:

99...Rh2
= (0.01) Depth: 9 00:00:00 0kN

Yes I don't use full 6 piece set, but engine should work correctly. I use KRPKRP.

I am not sure I correctly understand the problem. Running the search without tablebases gives me a similar output to yours. I only have 5-men TB.

Code: Select all

info depth 1 seldepth 1 score cp 102 nodes 35 nps 35000 time 0 hashfull 0 pv g2f2
info depth 2 seldepth 2 score cp 109 nodes 69 nps 69000 time 0 hashfull 0 pv g2f2 f7g7
info depth 3 seldepth 3 score cp 109 tbhits 1 nodes 119 nps 4250 time 27 hashfull 0 pv g2f2 f7g7 g4f3
info depth 4 seldepth 4 score cp 181 tbhits 9 nodes 431 nps 8288 time 51 hashfull 0 pv g2h2 f7g7 g4f3 g7f7
info depth 5 seldepth 5 score cp 181 tbhits 11 nodes 479 nps 9211 time 51 hashfull 0 pv g2h2 f7g7 g4f3 g7f7 f3e4
info depth 6 seldepth 7 score cp 221 tbhits 17 nodes 742 nps 10305 time 71 hashfull 0 pv g2h2 f7g7 g4f3 g7f7
info depth 7 seldepth 9 score cp 206 tbhits 46 nodes 1453 nps 16144 time 89 hashfull 0 pv g2h2 h6h7 h2h4 b3c2 g4h3 c2d1
info depth 8 seldepth 9 score cp 197 tbhits 59 nodes 2199 nps 24164 time 90 hashfull 0 pv g2h2 h6h7 h2h4 b3c4 g3g2 f7g7 g4f3 c4c3
info depth 9 seldepth 15 score cp 207 tbhits 156 nodes 4729 nps 32390 time 145 hashfull 0 pv g2h2 h6h7 h2h4 b3c4 g3g2 f7g7 g4h3 c4d3
info depth 10 seldepth 15 score cp 190 tbhits 300 nodes 10645 nps 61531 time 172 hashfull 0 pv g2h2 f7g7 g4f3 h6h7 f3f2 b3c4 g3g2 g7f7 f2g1 f7g7
info depth 11 seldepth 15 score cp 200 tbhits 339 nodes 12260 nps 70459 time 173 hashfull 0 pv g2h2 f7g7 g4f3 h6h7 f3f2 b3c4 g3g2 g7f7 f2g1 f7g7
info depth 12 seldepth 15 score cp 253 tbhits 431 nodes 17115 nps 85149 time 200 hashfull 0 pv g2h2 f7g7 g4f3 h6h7 f3f2 b3c4 g3g2 g7f7 f2g1 f7g7 g1h1
info depth 13 seldepth 17 score cp 297 tbhits 1040 nodes 37227 nps 144290 time 257 hashfull 0 pv g2h2 h6h7 h2h4 b3c2 g4h3 f7g7 g3g2 c2b2 h3h2 b2b3 g2g1q
info depth 14 seldepth 19 score cp 364 tbhits 2096 nodes 70433 nps 220103 time 319 hashfull 0 pv g2h2 f7g7 g4f3 g7f7 f3e3 f7g7 e3f2 g7f7 f2g1 f7g7 g3g2 g7g6 h2h4 b3c2 g1h2
info depth 15 seldepth 20 score cp 497 tbhits 3348 nodes 105789 nps 239341 time 441 hashfull 10 pv g2h2 f7g7 g4f3 g7f7 f3e3 f7g7 e3f2 g7f7 f2g1 f7g7 g3g2 g7g6 h2h4 b3c2 g1h1 c2b1 h1h2
info depth 16 seldepth 20 score cp 687 tbhits 5952 nodes 180455 nps 329899 time 546 hashfull 10 pv g2h2 f7g7 g4f3 g7f7 f3e3 f7g7 e3f2 g7f7 f2g1 h6h7 g3g2 f7b7 h2h3 b3b2 g1h2 b7g7 h3h4 b2a2
info depth 17 seldepth 24 score cp 761 tbhits 11465 nodes 325412 nps 487143 time 667 hashfull 30 pv g2h2 h6h7 h2h4 b3b4 g3g2 b4a5 g4h3 f7g7 h3h2 g7c7 g2g1q c7c2 h2g3 c2c3
info depth 18 seldepth 26 score cp 831 tbhits 17964 nodes 488931 nps 644177 time 758 hashfull 40 pv g2h2 f7g7 g4f3 g7f7 f3e3 f7g7 e3f2 h6h7 g3g2 g7f7 f2g1 f7a7 h2h3 b3a4 g1h2 a7c7 h3h4 a4a5 g2g1q c7c2 h2g3 c2c3
info depth 19 seldepth 31 score cp 901 tbhits 30782 nodes 840214 nps 890057 time 943 hashfull 50 pv g2h2 f7g7 g4f3 g7f7 f3e3 f7g7 g3g2 h6h7 e3f2 g7f7 f2g1 f7a7 h2h3 b3b4 g1h1 a7c7 h3h4 b4b5 g2g1q b5c6 g1e3 c6b7
info depth 20 seldepth 31 score cp 1007 tbhits 52754 nodes 1371196 nps 1124853 time 1218 hashfull 110 pv g2h2 f7g7 g4f3 g7f7 f3e3 f7g7 g3g2 h6h7 e3f2 g7f7 f2g1 f7a7 h2h3 b3c4 g1h1 c4d5 g2g1q a7f7 g1e3 f7g7 h3h6 d5c4
info depth 21 seldepth 35 score cp 1713 tbhits 73692 nodes 1882632 nps 1291242 time 1457 hashfull 160 pv g2h2 f7g7 g4f3 g7f7 f3e3 f7g7 g3g2 b3b4 e3f2 g7f7 f2g1 f7f6 g1h1 b4c3 h2h4 c3d2 g2g1q d2e2 h4h2 e2d3 g1g3 d3e4
info depth 22 seldepth 35 score cp 7916 tbhits 86690 nodes 2033835 nps 1294611 time 1570 hashfull 170 pv g2h2 f7h7 g3g2 h7g7 g4f3 g7f7 f3g3 f7g7 g3f2 g7f7 f2g1 f7f6 g1h1 b3a4 g2g1q f6d6 h2h4 a4a5 g1g5 a5a6 h4h6
info depth 23 seldepth 38 score cp 7918 tbhits 101371 nodes 2204611 nps 1289246 time 1709 hashfull 230 pv g2h2 f7h7 g3g2 h7g7 g4f3 g7f7 f3g3 f7g7 g3f2 g7f7 f2g1 f7f6 g1h1 b3a4 g2g1q f6d6 g1a1 a4b4 h2h6
info depth 24 seldepth 38 score cp 7918 tbhits 103511 nodes 2228102 nps 1293152 time 1722 hashfull 250 pv g2h2 f7h7 g3g2 h7g7 g4f3 g7f7 f3g3 f7g7 g3f2 g7f7 f2g1 f7f6 g1h1 b3a4 g2g1q f6d6 g1a1 a4b4 h2h6
info depth 25 seldepth 38 score cp 7918 tbhits 107766 nodes 2265733 nps 1275032 time 1776 hashfull 270 pv g2h2 f7h7 g3g2 h7g7 g4f3 g7f7 f3g3 f7g7 g3f2 g7f7 f2g1 f7f6 g1h1 b3a4 g2g1q f6d6 g1a1 a4b4 h2h6
info depth 26 seldepth 38 score cp 7918 tbhits 111575 nodes 2308585 nps 1281123 time 1801 hashfull 270 pv g2h2 f7h7 g3g2 h7g7 g4f3 g7f7 f3g3 f7g7 g3f2 g7f7 f2g1 f7f6 g1h1 b3a4 g2g1q f6d6 g1a1 a4b4 h2h6
info depth 27 seldepth 40 score cp 7918 tbhits 117454 nodes 2367154 nps 1267213 time 1867 hashfull 280 pv g2h2 f7h7 g3g2 h7g7 g4f3 g7f7 f3g3 f7g7 g3f2 g7f7 f2g1 f7f6 g1h1 b3a4 g2g1q f6d6 g1a1 a4b4 h2h6
info depth 28 seldepth 42 score cp 7918 tbhits 130946 nodes 2495739 nps 1264305 time 1973 hashfull 310 pv g2h2 f7h7 g3g2 h7g7 g4f3 g7f7 f3g3 f7g7 g3f2 g7f7 f2g1 f7f6 g1h1 b3a4 g2g1q f6d6 g1a1 a4b4 h2h6
info depth 29 seldepth 46 score cp 7918 tbhits 184383 nodes 3135672 nps 1309804 time 2393 hashfull 350 pv g2h2 f7h7 g3g2 h7g7 g4f3 g7f7 f3g3 f7g7 g3f2 g7f7 f2g1 f7f6 g1h1 b3a4 g2g1q f6d6 g1a1 a4b4 h2h6
info depth 30 seldepth 46 score cp 7918 tbhits 188882 nodes 3187546 nps 1316623 time 2420 hashfull 360 pv g2h2 f7h7 g3g2 h7g7 g4f3 g7f7 f3g3 f7g7 g3f2 g7f7 f2g1 f7f6 g1h1 b3a4 g2g1q f6d6 g1a1 a4b4 h2h6
bestmove g2h2

So are you saying the problem might be related to using PARTIAL 6 men-TB?
For your information the "Depth" in your case comes from probing the DTZ. The Depth coresponds the amount of moves until zeroing will happen. Its odd that the score is infact 0 but the correct move should still be played.

Edit:

I checked and the score returned is +1. This coresponds to a CURSED WIN for some reason. I am not sure why thats returned by your tablebases.

[Jouni]

Position is mate in 25 for black says Nalimov.