This is a position from Rodin in ICC. Rodin did not saw this was a draw.
[D]6k1/5p1p/4pPpP/3p4/1r1P4/1R6/2K2P2/8 b - - 0 41
I have put the position to various engines (of differents strength) and it is a very curious thing. Same see the draw at very low depth, while others need to go to 26 depth, same fast, same not so fast.
My engine need a lot of time to see a draw. As I know, my TT works well, so what do you think it is needed to see a quick draw here?
Thanks
A draw position
Moderators: hgm, Rebel, chrisw
-
Kempelen
- Posts: 620
- Joined: Fri Feb 08, 2008 10:44 am
- Location: Madrid - Spain
-
Kempelen
- Posts: 620
- Joined: Fri Feb 08, 2008 10:44 am
- Location: Madrid - Spain
Re: A draw position
I copy some analysis from Rodin:
Code: Select all
18 149 307 1769763 1... Rc4+ 2.Kd3 Rc8 3.Ke3 Kf8 4.Kf4 Ra8 5.Rb7 Kg8 6.K
e5 Rc8 7.f3 Kf8 8.f4 Ra8 9.Rb3 Ke8 10.Rb7
19 151 421 2689960 1... Rc4+ 2.Kd3 Rc8 3.Ke3 Kf8 4.Kf4 Ra8 5.Rb7 Kg8 6.K
e5 Rc8 7.f3 Kf8 8.f4 Ra8 9.Rc7 Rb8 10.Ra7 Ke8 11.Rc7 Rb3
20 130 603 4181219 1... Rc4+ 2.Rc3 Rxc3+ 3.Kxc3 Kf8 4.Kb4 Ke8 5.Kc5 Kd8
6.Kc6 g5 7.Kd6 g4 8.Ke5 Kd7 9.Kf4 Kd6 10.Kxg4 e5 11.dxe5+ Kxe5 12.Kf3 d4 13.Ke2
Ke4 14.f3+ Kf4
21 26 1628 12955065 1... Rc4+ 2.Rc3 Kf8 3.Rxc4 dxc4 4.Kc3 Ke8 5.Kxc4 Kd7
6.Kc5 Kc7 7.d5 exd5 8.Kxd5 Kd7
22 20 2129 17844703 1... Rc4+ 2.Rc3 Kf8 3.Rxc4 dxc4 4.Kc3 Ke8 5.Kxc4 Kd7
6.Kc5 Kc7 7.d5 exd5 8.Kxd5 Kd7 9.f3 Kc7 10.Kc5 Kc8 11.Kd6 Kd8 12.f4
23 40 3338 28801787 1... Rc4+ 2.Rc3 Kf8 3.Rxc4 dxc4 4.Kc3 Ke8 5.Kxc4 Kd8
6.d5 Kd7 7.Kc5 g5 8.f3 e5 9.Kb4 e4 10.fxe4 g4 11.e5 g3 12.e6+ fxe6 13.f7 Ke7 14.
dxe6 Kxe6
24 36 5896 52571692 1... Rc4+ 2.Rc3 Kf8 3.Rxc4 dxc4 4.Kc3 Ke8 5.Kxc4 Kd8
6.Kd3 Kd7 7.f4 Kd6 8.Kd2 Kc6
25 30 15115 135874573 1... Rc4+ 2.Rc3 Kf8 3.Rxc4 dxc4 4.Kc3 Ke8 5.Kxc4 Kd8
6.f4 Kc8 7.Kb4 Kd7 8.Kb5
26 9 19484 170558687 1... Rc4+ 2.Rc3 Kf8 3.Rxc4 dxc4 4.Kc3 Ke8 5.Kxc4 g5 6
.d5 Kd7 7.Kc5 e5 8.f3 Kc7 9.Kb5 Kd6 10.Kc4-
Dann Corbit
- Posts: 12538
- Joined: Wed Mar 08, 2006 8:57 pm
- Location: Redmond, WA USA
Re: A draw position
Crafty up to version 20 had some code to detect draws caused by pawn wall formations. It did not increase the overall strength of crafty so Dr. Hyatt removed it. However, for analyzing positions of this type, I have a version of crafty compiled with DETECT_DRAW enabled and this is my favorite engine for closed positions, bar none. It took well under one second for crafty to see that this position was a draw. You can download the old code by Jeremiah Pennery in Crafty 20.0 to take a look.
Code: Select all
Q:\>craftydd
Initializing multiple threads.
System is SMP, not NUMA.
EPD Kit revision date: 1996.04.21
unable to open book file [./book.bin].
book is disabled
unable to open book file [./books.bin].
Crafty v20.0 (1 cpus)
White(1): mt 4
max threads set to 4
White(1): hash 1024M
hash table memory = 768M bytes.
White(1): hashp 64M
pawn hash table memory = 48M bytes.
White(1): st 180
search time set to 180.00.
White(1): epdpfga draw.epd draw-analyzed.epd
PFGA: EPD record: 1
time surplus 0.00 time limit 3:00 (3:00)
depth time score variation (1)
starting thread 1
starting thread 2
starting thread 3
17-> 0.28 -0.82 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4.
Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Ke5
Kd7 8. Kf4 Kd6 9. Kxg4 e5 10. dxe5+
Kxe5 11. Kg5 (s=2)
18 0.35 +1 1. ... Rxb3
18 0.41 -0.06 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4.
Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6
Kc8 8. Kd6 Kb7 9. Ke5 Kc6 10. Kf4 Kb5
11. Kxg4
18-> 0.63 -0.06 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4.
Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6
Kc8 8. Kd6 Kb7 9. Ke5 Kc6 10. Kf4 Kb5
11. Kxg4 (s=2)
19 0.72 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4.
Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6
Kc8 8. Kd6 Kb7 9. Ke5 Kc6 10. Kf4 Kb5
11. Kg3
19-> 1.14 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4.
Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6
Kc8 8. Kd6 Kb7 9. Ke5 Kc6 10. Kf4 Kb5
11. Kg3 (s=2)
20 1.23 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4.
Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6
Kc8 8. Kd6 Kd8
20-> 1.86 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4.
Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6
Kc8 8. Kd6 Kd8 (s=2)
21 1.98 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4.
Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6
Kc8 8. Kd6 Kd8
21-> 3.40 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4.
Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6
Kc8 8. Kd6 Kd8 (s=2)
22 3.56 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4.
Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6
Kc8 8. Kd6 Kd8
22-> 6.46 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4.
Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6
Kc8 8. Kd6 Kd8 (s=2)
23 6.69 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4.
Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6
Kc8 8. Kd6 Kd8
23-> 12.36 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4.
Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6
Kc8 8. Kd6 Kd8 (s=2)
24 12.64 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4.
Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6
Kc8 8. Kd6 Kd8
24-> 24.53 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4.
Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6
Kc8 8. Kd6 Kd8 (s=5)
25 25.06 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4.
Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6
Kc8 8. Kd6 Kd8 (s=4)
25-> 51.54 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4.
Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6
Kc8 8. Kd6 Kd8 (s=5)
26 52.40 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4.
Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6
Kc8 8. Kd6 Kd8 (s=4)
26-> 1:48 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4.
Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6
Kc8 8. Kd6 Kd8 (s=6)
27 1:50 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4.
Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6
Kc8 8. Kd6 Kd8 (s=5)
time=3:00 mat=-1 n=2156151372 fh=89% nps=11.97M
ext-> chk=128940286 cap=4661233 1rep=5030717 mate=1424241
predicted=0 nodes=2156151372 evals=1047112963 50move=0
endgame tablebase-> probes=0 hits=0
hashing-> 47%(raw) 43%(depth) 14%(sat) 99%(pawn)
hashing-> 0%(exact) 39%(lower) 3%(upper)
SMP-> split=721331 stop=56265 data=32/32768 elap=3:00
Black(1):
-
Kempelen
- Posts: 620
- Joined: Fri Feb 08, 2008 10:44 am
- Location: Madrid - Spain
Re: A draw position
This evening I have dissabled an old code I have to not store 3-rep possitions in TT and this accelerate a bit the detection, but still need many time.Dann Corbit wrote:Crafty up to version 20 had some code to detect draws caused by pawn wall formations. It did not increase the overall strength of crafty so Dr. Hyatt removed it. However, for analyzing positions of this type, I have a version of crafty compiled with DETECT_DRAW enabled and this is my favorite engine for closed positions, bar none. It took well under one second for crafty to see that this position was a draw. You can download the old code by Jeremiah Pennery in Crafty 20.0 to take a look.Code: Select all
Q:\>craftydd Initializing multiple threads. System is SMP, not NUMA. EPD Kit revision date: 1996.04.21 unable to open book file [./book.bin]. book is disabled unable to open book file [./books.bin]. Crafty v20.0 (1 cpus) White(1): mt 4 max threads set to 4 White(1): hash 1024M hash table memory = 768M bytes. White(1): hashp 64M pawn hash table memory = 48M bytes. White(1): st 180 search time set to 180.00. White(1): epdpfga draw.epd draw-analyzed.epd PFGA: EPD record: 1 time surplus 0.00 time limit 3:00 (3:00) depth time score variation (1) starting thread 1 starting thread 2 starting thread 3 17-> 0.28 -0.82 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4. Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Ke5 Kd7 8. Kf4 Kd6 9. Kxg4 e5 10. dxe5+ Kxe5 11. Kg5 (s=2) 18 0.35 +1 1. ... Rxb3 18 0.41 -0.06 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4. Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6 Kc8 8. Kd6 Kb7 9. Ke5 Kc6 10. Kf4 Kb5 11. Kxg4 18-> 0.63 -0.06 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4. Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6 Kc8 8. Kd6 Kb7 9. Ke5 Kc6 10. Kf4 Kb5 11. Kxg4 (s=2) 19 0.72 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4. Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6 Kc8 8. Kd6 Kb7 9. Ke5 Kc6 10. Kf4 Kb5 11. Kg3 19-> 1.14 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4. Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6 Kc8 8. Kd6 Kb7 9. Ke5 Kc6 10. Kf4 Kb5 11. Kg3 (s=2) 20 1.23 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4. Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6 Kc8 8. Kd6 Kd8 20-> 1.86 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4. Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6 Kc8 8. Kd6 Kd8 (s=2) 21 1.98 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4. Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6 Kc8 8. Kd6 Kd8 21-> 3.40 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4. Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6 Kc8 8. Kd6 Kd8 (s=2) 22 3.56 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4. Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6 Kc8 8. Kd6 Kd8 22-> 6.46 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4. Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6 Kc8 8. Kd6 Kd8 (s=2) 23 6.69 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4. Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6 Kc8 8. Kd6 Kd8 23-> 12.36 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4. Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6 Kc8 8. Kd6 Kd8 (s=2) 24 12.64 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4. Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6 Kc8 8. Kd6 Kd8 24-> 24.53 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4. Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6 Kc8 8. Kd6 Kd8 (s=5) 25 25.06 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4. Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6 Kc8 8. Kd6 Kd8 (s=4) 25-> 51.54 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4. Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6 Kc8 8. Kd6 Kd8 (s=5) 26 52.40 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4. Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6 Kc8 8. Kd6 Kd8 (s=4) 26-> 1:48 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4. Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6 Kc8 8. Kd6 Kd8 (s=6) 27 1:50 -0.01 1. ... Rxb3 2. Kxb3 g5 3. Kb4 Kf8 4. Kc5 Ke8 5. Kc6 Kd8 6. Kd6 g4 7. Kc6 Kc8 8. Kd6 Kd8 (s=5) time=3:00 mat=-1 n=2156151372 fh=89% nps=11.97M ext-> chk=128940286 cap=4661233 1rep=5030717 mate=1424241 predicted=0 nodes=2156151372 evals=1047112963 50move=0 endgame tablebase-> probes=0 hits=0 hashing-> 47%(raw) 43%(depth) 14%(sat) 99%(pawn) hashing-> 0%(exact) 39%(lower) 3%(upper) SMP-> split=721331 stop=56265 data=32/32768 elap=3:00 Black(1):
one more question, why Craty says -0.01 and not 0.00 ??
-
UncombedCoconut
- Posts: 319
- Joined: Fri Dec 18, 2009 11:40 am
- Location: Naperville, IL
Re: A draw position
Black is better if White defends incorrectly. A tiny negative score will make sure the engine won't throw away its extra pawn for no reason. (This is a general principle. In this position White would have to be stupid to lose, but the engine doesn't know that.)Kempelen wrote:one more question, why Craty says -0.01 and not 0.00 ??
-
Kempelen
- Posts: 620
- Joined: Fri Feb 08, 2008 10:44 am
- Location: Madrid - Spain
Re: A draw position
Are you sure is v 20.0 ??? I cant find any reference in the codeDann Corbit wrote:Crafty up to version 20 had some code to detect draws caused by pawn wall formations. It did not increase the overall strength of crafty so Dr. Hyatt removed it. However, for analyzing positions of this type, I have a version of crafty compiled with DETECT_DRAW enabled and this is my favorite engine for closed positions, bar none. It took well under one second for crafty to see that this position was a draw. You can download the old code by Jeremiah Pennery in Crafty 20.0 to take a look.