Hamsters randomizer in action

Discussion of chess software programming and technical issues.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Alessandro Scotti

Hamsters randomizer in action

Post by Alessandro Scotti » Sun Jun 03, 2007 7:24 pm

After the comments from Michael and Ron I've managed to spend some time with Hamsters. As a result, my wife won't talk to me. As a side effect though, I think I've found something.

In this position (best move is Kb1):

[D]6k1/3p4/3p4/3P4/4P1p1/6P1/8/K7 w - - 0 1

I get the following PV:

Code: Select all

28/46 0:03.45 0 3582109 Kb2 Kh8 Kc3 Kg7 Kd4 Kf6 Ke3 Kg5 Kd4 Kf6
28/46 0:04.22 190 4472111 Kb1 Kg7 Kc1 Kh7 Kc2 Kg7 Kc3 Kh6 Kd4 Kg6 e5 dxe5+ Kxe5 Kf7 Kf5 Ke7 Kxg4 Kd6 Kf5 Ke7 g4...
29/46 0:04.99 0 5317284 Kb1 Kg7 Kc1 Kh7 Kc2 Kg7 Kc3 Kf7 Kd4 Kf6 Ke3 Kg5 Kd4 Kf6
30/48 0:06.27 240 6471878 Kb1 Kg7 Kc1 Kh7 Kc2 Kg7 Kc3 Kh7 Kd4 Kh6 e5 Kg6 Ke4 dxe5 Kxe5 Kf7 Kf5 Ke7 Kxg4 Kf6 Kf4 d6 g4...
31/48 0:07.91 255 8076986 Kb1 Kg7 Kc1 Kh7 Kc2 Kg6 Kd3 Kf6 Kd4 Kg6 e5 Kf7 exd6 Kf6 Ke4 Kg6 Kf4 Kf7 Kxg4 Kf6 Kh5 Kf7 g4...
Or variations thereof by playing with some parameters:

Code: Select all

28/46 0:02.69 37 2754395 Kb1 Kg7 Kc1 Kh6 Kd2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 Kf7 exd6 Kf6 Ke4 Kg6 Kf4 Kf6 Kxg4...
29/47 0:03.73 125 3680504 Kb1 Kg7 Kc1 Kh6 Kd2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 Kg5 Ke4 Kg6 exd6 Kg5 Ke5 Kg6 Kf4 Kf6 Kxg4...
30/49 0:05.83 170 5676962 Kb1 Kg7 Kc1 Kh6 Kd2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 Kf7 exd6 Kf6 Ke4 Kg6 Kf4 Kf6 Kxg4...
31/49 0:08.61 51 8711860 Kb1 Kg7 Kc1 Kh6 Kd2 Kg6 Kd3 Kh6 Kd4 Kh5 e5 Kg6 Ke4 Kf7 exd6 Kg6 Kf4 Kf6 Kxg4...
32/54 0:12.11 260 12086482 Kb1 Kg7 Kc1 Kh6 Kd2 Kg6 Kd3 Kh6 Kd4 Kg6 e5 Kf5 exd6 Kf6 Ke4 Kg6 Kf4 Kf6 Kxg4...
There are huge score drops in some plies that suggest a serious bug somewhere. I've spent a few hours poking around to no avail, and yet I think this time asking for suggestions on the bug would be too much even for the collective wisdom of the CCC! :-)
So my question is... how do I debug something like this that happens million nodes away from the root?!? :o

Gerd Isenberg
Posts: 2127
Joined: Wed Mar 08, 2006 7:47 pm
Location: Hattingen, Germany

Re: Hamsters randomizer in action

Post by Gerd Isenberg » Sun Jun 03, 2007 8:25 pm

Isn't it more a deterministic odd-even effect rather than a randomizer? You have to analyze the end-positions in the pvs, causing the oscillating scores - your eval might be very sensitive by side to move. For instance, you may try to favour same color to move at the leaves in pv-nodes by doing/not doing pv-extensions near the tips.

Alessandro Scotti

Re: Hamsters randomizer in action

Post by Alessandro Scotti » Sun Jun 03, 2007 8:33 pm

To be honest, I don't have a clue what's happening. However, I've selectively enabled/disabled/changed most eval and search terms and it seems the problem is always there, although at different ply, score, etc.
In the first example it seems a 3-rep draw is found at ply 29... but then why it wasn't found at ply 28?

Gerd Isenberg
Posts: 2127
Joined: Wed Mar 08, 2006 7:47 pm
Location: Hattingen, Germany

Re: Hamsters randomizer in action

Post by Gerd Isenberg » Sun Jun 03, 2007 8:43 pm

The weaker side standing pat at the leaves might be crucial here and probably worth an extra ply if in zugzwang, where only losing king moves or losing pawn moves are possible.

Michael Sherwin
Posts: 3041
Joined: Fri May 26, 2006 1:00 am
Location: WY, USA
Full name: Michael Sherwin

Re: Hamsters randomizer in action

Post by Michael Sherwin » Mon Jun 04, 2007 2:04 am

I suspect a hashtable bug. I suggest testing Hamsters while only using the hashtables for move ordering to see if the bug is still there.
I hate if statements. Pawns demand if statements. Therefore I hate pawns.

Pradu
Posts: 287
Joined: Sat Mar 11, 2006 2:19 am
Location: Atlanta, GA
Contact:

Re: Hamsters randomizer in action

Post by Pradu » Mon Jun 04, 2007 2:33 am

Alessandro Scotti wrote:After the comments from Michael and Ron I've managed to spend some time with Hamsters. As a result, my wife won't talk to me. As a side effect though, I think I've found something.
Hi! Your search/eval is quite good here as Hamsters seems to take way less nodes than Buzz here to reach the same ply:

Code: Select all

1 -139 0 2 e5
1 15 0 4 Kb1
1 15 0 6 Kb1
2 15 0 12 Kb1 Kf7
2 15 0 18 Kb1 Kf7
3 15 1 37 Kb1 Kf7 Ka1
3 15 1 64 Kb1 Kf7 Ka1
4 15 1 144 Kb1 Kf7 Kc1 Kf6
4 15 1 179 Kb1 Kf7 Kc1 Kf6
5 15 1 288 Kb1 Kf7 Kc1 Kf6 Kb1
5 15 1 413 Kb1 Kf7 Kc1 Kf6 Kb1
6 15 1 631 Kb1 Kf7 Kc1 Kf6 Kd1 Kf7
6 18 3 909 Kb2 Kf7 Kc3 Kf6 Kd4 Kf7
6 18 3 909 Kb2 Kf7 Kc3 Kf6 Kd4 Kf7
7 15 3 1660 Kb2 Kf7 Kc3 Kg7 Kd4 Kf7 Kd3
7 15 3 1858 Kb2 Kf7 Kc3 Kg7 Kd4 Kf7 Kd3
8 15 3 2712 Kb2 Kf7 Kc3 Kg7 Kd4 Kf7 Kd3 Kg8 Kc4 Kf7 Kd4 Kf6 Kc4 Kg6 Kd3 Kf6 Kd4
Kf7
8 15 4 3112 Kb2 Kf7 Kc3 Kg7 Kd4 Kf7 Kd3 Kg8 Kc4 Kf7 Kd4 Kf6 Kc4 Kg6 Kd3 Kf6 Kd4
Kf7
9 18 4 4715 Kb2 Kf7 Kc3 Kg7 Kd4 Kf7 Ke3 Kg6 Kf4
9 18 4 5175 Kb2 Kf7 Kc3 Kg7 Kd4 Kf7 Ke3 Kg6 Kf4
10 18 6 7475 Kb2 Kf7 Kc3 Kg7 Kd4 Kf7 Ke3 Kg6 Kf4 Kh5
10 18 6 8202 Kb2 Kf7 Kc3 Kg7 Kd4 Kf7 Ke3 Kg6 Kf4 Kh5
11 18 7 11011 Kb2 Kf7 Kc3 Kg7 Kd4 Kf7 Ke3 Kg6 Kf4 Kh5 Kf5
11 18 7 11896 Kb2 Kf7 Kc3 Kg7 Kd4 Kf7 Ke3 Kg6 Kf4 Kh5 Kf5
12 15 9 25975 Kb2 Kh8 Kc2 Kh7 Kd1 Kg6 Ke1 Kg7 Kd2 Kh6 Ke2 Kh5 Kd1 Kg6
12 15 10 27388 Kb2 Kh8 Kc2 Kh7 Kd1 Kg6 Ke1 Kg7 Kd2 Kh6 Ke2 Kh5 Kd1 Kg6
13 15 10 32393 Kb2 Kh8 Kc2 Kh7 Kd1 Kg6 Ke1 Kg7 Kd2 Kh6 Ke2 Kh5 Ke3 Kg5 Kd4 Kf6 K
d3 Kg6 Ke3 Kg5
13 15 12 37401 Kb2 Kh8 Kc2 Kh7 Kd1 Kg6 Ke1 Kg7 Kd2 Kh6 Ke2 Kh5 Ke3 Kg5 Kd4 Kf6 K
d3 Kg6 Ke3 Kg5
14 22 14 44569 Kb2 Kh8 Kc2 Kh7 Kd1 Kg6 Ke1 Kg7 Kd2 Kh6 Ke2 Kh5 Kd2 Kh6
14 22 15 50194 Kb2 Kh8 Kc2 Kh7 Kd1 Kg6 Ke1 Kg7 Kd2 Kh6 Ke2 Kh5 Kd2 Kh6
15 18 18 68950 Kb2 Kh8 Kb3 Kg8 Kb4 Kf8 Ka5 Kf7 Kb6 Kf6 Kc7 Ke7 Kc8 Ke8 Kc7 Ke7
15 18 20 77771 Kb2 Kh8 Kb3 Kg8 Kb4 Kf8 Ka5 Kf7 Kb6 Kf6 Kc7 Ke7 Kc8 Ke8 Kc7 Ke7
16 22 21 92926 Kb2 Kh8 Kb3 Kg8 Kb4 Kf8 Kc4 Kf7 Kd4 Kf6 Kc4 Kf7
16 22 23 103859 Kb2 Kh8 Kb3 Kg8 Kb4 Kf8 Kc4 Kf7 Kd4 Kf6 Ke3 Kg5 Kd3 Kg6 Kd4 Kf6
17 18 28 125374 Kb2 Kh8 Kb3 Kg8 Kb4 Kf8 Kc4 Kf7 Kd4 Kf6 Ke3 Kg5 Kd4 Kf6
17 18 29 139839 Kb2 Kh8 Kb3 Kg8 Kb4 Kf8 Kc4 Kf7 Kd4 Kf6 Ke3 Kg5 Kd4 Kf6
18 15 35 183449 Kb2 Kh8 Kb3 Kg8 Kb4 Kf8 Kc4 Kf7 Kd4 Kf6 Kd3 Kg6 Ke3 Kg5 Ke2 Kh5
Kd2 Kh6 Ke3 Kg5
18 15 40 210670 Kb2 Kh8 Kb3 Kg8 Kb4 Kf8 Kc4 Kf7 Kd4 Kf6 Ke3 Kg5 Ke2 Kh5 Kd2 Kh6
Kc3 Kg7 Kc4 Kf7
19 15 48 254455 Kb2 Kh8 Kb3 Kg8 Kb4 Kf8 Kc4 Kf7 Kd4 Kf6 Ke3 Kg5 Ke2 Kh5 Kd2 Kh6
Kc1 Kg7 Kd1 Kg6 Ke1 Kg5 Kd2 Kh6
19 15 54 286213 Kb2 Kh8 Kb3 Kg8 Kb4 Kf8 Kc4 Kf7 Kd4 Kf6 Ke3 Kg5 Ke2 Kh5 Kd2 Kh6
Kc1 Kg7 Kd1 Kg6 Ke1 Kg5 Kd2 Kh6
20 15 65 344824 Kb2 Kh8 Kb3 Kg8 Kb4 Kf8 Kc4 Kf7 Kd4 Kf6 Ke3 Kg5 Ke2 Kh5 Kd1 Kg6
Ke1 Kg5 Kf1 Kg6 Kg1 Kf7 Kh1 Kf6 Kg2 Kf7 Kf1 Kg6
20 15 73 383824 Kb2 Kh8 Kb3 Kg8 Kb4 Kf8 Kc4 Kf7 e5 dxe5 Kc5 Kg8 Kd6 e4 Kxd7 e3 K
c6 Kf7 d6 e2 d7 e1=Q Kc7 Qxg3+ Kc8
21 15 135 707573 Kb2 Kh8 Kb3 Kg8 Kb4 Kf8 Kc4 Kf7 Kd4 Kf6 Ke3 Kg5 Kd4 Kf6
21 15 142 757340 Kb2 Kh8 Kb3 Kg8 Kb4 Kf8 Kc4 Kf7 Kd4 Kf6 e5+ dxe5+ Kd3 Kf5 Ke3 e
4 Ke2 Ke5 Kd1 Kf5 d6 e3 Kc2 Ke4 Kd1 e2+ Kc2 Ke5 Kd3 e1=Q Kc4 Qxg3
22 15 200 1055628 Kb2 Kh8 Kb3 Kg8 Kb4 Kf8 Kc4 Kf7 Kd4 Kf6 Ke3 Kg5 Kd2 Kh6 Kc1 Kg
7 Kb1 Kg6 Kc2 Kh7 Kb3 Kg6 Kc3 Kg7 Kd4 Kf6
22 15 206 1105113 Kb2 Kh8 Kb3 Kg8 Kb4 Kf8 Kc4 Kf7 Kd4 Kf6 Ke3 Kg5 Kf2 Kg6 Kf1 Kh
6 Kg1 Kg6 Kh1 Kh5 Kg2 Kg6 Kf1
23 15 312 1589693 Kb2 Kh8 Kb3 Kg8 Kb4 Kf8 Kc4 Kf7 Kd4 Kf6 Ke3 Kg5 Kf2 Kg6 Kf1 Kh
6 Ke2 Kh5 Ke1 Kg5 Kd2 Kh6 Kc1 Kg7 Kc2 Kh7 Kc3 Kg7 Kd4 Kf6
23 15 320 1644222 Kb2 Kh8 Kb3 Kg8 Kb4 Kf8 Kc4 Kf7 Kd4 Kf6 Ke3 Kg5 Kf2 Kg6 Kf1 Kh
6 Ke2 Kh5 Ke1 Kg5 Kd2 Kh6 Kc1 Kg7 Kc2 Kh7 Kc3 Kg7 Kd4 Kf6
24 15 523 2554487 Kb2 Kh8 Kb3 Kg8 Kb4 Kf8 Kc4 Kf7 Kd4 Kf6 Ke3 Kg5 Kf2 Kg6 Kf1 Kh
6 Ke2 Kh5 Ke1 Kg5 Kd2 Kh6 Kc1 Kg7 Kc2 Kh7 Kd3 Kg6 Ke3 Kg5
24 32 535 2655402 Kb1 Kg7 Kc1 Kg6 Kd1 Kg5 Kc2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 dxe5+ Kxe5
Kf7 Kf5 Ke7 Kxg4 Kd6 Kf5 Kxd5 g4 Kc5 g5 d5 g6 Kd6 Kf4 Ke7 Kg4
24 32 537 2659983 Kb1 Kg7 Kc1 Kg6 Kd1 Kg5 Kc2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 dxe5+ Kxe5
Kf7 Kf5 Ke7 Kxg4 Kd6 Kf5 Kxd5 g4 Kc5 g5 d5 g6 Kd6 Kf4 Ke7 Kg4
25 32 546 2726782 Kb1 Kg7 Kc1 Kg6 Kd1 Kg5 Kc2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 dxe5+ Kxe5
Kf7 Kf5 Ke7 Kxg4 Kd6 Kf5 Kxd5 g4 Kc5 g5 d5 g6 Kd6 Kf4 Ke7 Ke5 Kf8 Kd4 Kg8
25 32 934 4469943 Kb1 Kg7 Kc1 Kg6 Kd1 Kg5 Kc2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 dxe5+ Kxe5
Kf7 Kf5 Ke7 Kxg4 Kd6 Kf5 Kxd5 g4 Kc5 g5 d5 g6 Kd6 Kf4 Ke7 Ke5 Kf8 Kd4 Kg8
26 32 948 4558810 Kb1 Kg7 Kc1 Kg6 Kd1 Kg5 Kc2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 dxe5+ Kxe5
Kf7 Kf5 Ke7 Kxg4 Kd6 Kf5 Ke7 g4 Kd6 g5 Ke7 d6+ Kf7 Kf4 Ke6 Kg4 Kxd6 g6 Ke6 Kh5 d
5 Kg5 Kd7 Kf4 Ke7 Ke5 Kf8 Kd4 Kg8
26 32 1643 7609236 Kb1 Kg7 Kc1 Kg6 Kd1 Kg5 Kc2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 dxe5+ Kxe5
 Kf7 Kf5 Ke7 Kxg4 Kd6 Kf5 Ke7 g4 Kd6 g5 Ke7 d6+ Kf7 Kf4 Ke6 Kg4 Kxd6 g6 Ke6 Kh5
d5 Kg5 Kd7 Kf4 Ke7 Ke5 Kf8 Kd4 Kg8
27 30 1717 8047855 Kb1 Kg7 Kc1 Kg6 Kd1 Kg5 Kc2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 Kf7 Ke4 dx
e5 Kxe5 Ke7 d6+ Kf7 Kf5 Kg7 Kxg4 Kh6 Kf4 Kg6 g4 Kf6 Ke4 Kf7 g5 Ke6 Kf4 Kxd6 g6 K
e7 Kg5 d6 Kf5 d5 Ke5 Kf8 Kd4 Kg8
27 30 2990 13688236 Kb1 Kg7 Kc1 Kg6 Kd1 Kg5 Kc2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 Kf7 Ke4 d
xe5 Kxe5 Ke7 d6+ Kf7 Kf5 Kg7 Kxg4 Kh6 Kf4 Kg6 g4 Kf6 Ke4 Kf7 g5 Ke6 Kf4 Kxd6 g6
Ke7 Kg5 d6 Kf5 d5 Ke5 Kf8 Kd4 Kg8
28 40 3021 13895202 Kb1 Kg7 Kc1 Kg6 Kd1 Kg5 Kc2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 Kf5 exd6
Kf6 Ke4 Kg6 Kf4 Kf6 Kxg4 Ke5 Kh5 Kxd6 g4 Kxd5 g5 d6 g6 Kc6 Kg5 Kb6
28 40 4940 22489923 Kb1 Kg7 Kc1 Kg6 Kd1 Kg5 Kc2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 Kf5 exd6
Kf6 Ke4 Kg6 Kf4 Kf6 Kxg4 Ke5 Kh5 Kxd6 g4 Kxd5 g5 d6 g6 Kc6 Kg5 Kb6
29 40 4979 22724589 Kb1 Kg7 Kc1 Kg6 Kd1 Kg5 Kc2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 Kf5 exd6
Kf6 Ke4 Kg6 Kf4 Kf6 Kxg4 Ke5 Kh5 Kxd6 g4 Kxd5 g5 d6 g6 Kc6 Kg5 Kd7 Kf5 Ke7 Kg5 d
5 g7 Kf7 Kh6 Ke7 Kh7 d4 g8=Q
29 40 8382 38644084 Kb1 Kg7 Kc1 Kg6 Kd1 Kg5 Kc2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 Kf5 exd6
Kf6 Ke4 Kg6 Kf4 Kf6 Kxg4 Ke5 Kh5 Kxd6 g4 Kxd5 g5 d6 g6 Kc6 Kg5 Kd7 Kf5 Ke7 Kg5 d
5 g7 Kf7 Kh6 Ke7 Kh7
30 140 8440 38999665 Kb1 Kg7 Kc1 Kg6 Kd1 Kg5 Kc2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 Kf5 exd6
 Kg6 Kc5 Kf5 Kb6 Ke4 Kc7 Kf3 Kxd7 Kxg3 Kc7 Kf2 d7 g3 d6 g2 d8=Q g1=Q
30 140 10415 48794923 Kb1 Kg7 Kc1 Kg6 Kd1 Kg5 Kc2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 Kf5 exd
6 Kg6 Kc5 Kf5 Kb6 Ke4 Kc7 Kf3 Kxd7 Kxg3 Kc7 Kf2 d7 g3 d6 g2 d8=Q g1=Q
31 134 10731 50616592 Kb1 Kg7 Kc1 Kg6 Kd1 Kg5 Kc2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 Kf5 exd
6 Kg6 Kc5 Kf5 Kb6 Ke4 Kc7 Kf3 Kxd7 Kxg3 Ke7 Kf3 d7 g3 d8=Q Kf4 Qb6
31 134 14051 67086644 Kb1 Kg7 Kc1 Kg6 Kd1 Kg5 Kc2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 Kf5 exd
6 Kg6 Kc5 Kf5 Kb6 Ke4 Kc7 Kf3 Kxd7 Kxg3 Ke7 Kf3 d7 g3 d8=Q Kf4 Qb6
32 143 14307 68329148 Kb1 Kg7 Kc1 Kg6 Kd1 Kg5 Kc2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 Kf5 exd
6 Kg6 Kc5 Kf5 Kb6 Ke4 Kc7 Kf3 Kxd7 Kxg3 Kc7 Kf2 d7 Ke2 d6 g3 d8=Q Kf2
32 143 16568 80961639 Kb1 Kg7 Kc1 Kg6 Kd1 Kg5 Kc2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 Kf5 exd
6 Kg6 Kc5 Kf5 Kb6 Ke4 Kc7 Kf3 Kxd7 Kxg3 Kc7 Kf2 d7 Ke2 d6 g3 d8=Q Kf2
33 156 17025 83325734 Kb1 Kg7 Kc1 Kg6 Kd1 Kg5 Kc2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 Kf5 exd
6 Kf6 Ke4 Kg5 Ke5 Kg6 Kf4 Kf6 Kxg4 Ke5 Kg5 Kxd6 Kf5 Ke7 g4 Kf7 g5 d6 g6+ Kg8 Kf4
 Kf8 Kg4 Kg8 Kf5 Kf8 Kf4 Kg8 Kg4 Kf8 Kf5 Kg8
33 156 21232 105679551 Kb1 Kg7 Kc1 Kg6 Kd1 Kg5 Kc2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 Kf5 ex
d6 Kf6 Ke4 Kg5 Ke5 Kg6 Kf4 Kf6 Kxg4 Ke5 Kg5 Kxd6 Kf5 Ke7 g4 Kf7 g5 d6 g6+ Kg8 Kf
4 Kf8 Kg4 Kg8 Kf5 Kf8 Kf4 Kg8 Kg4 Kf8 Kf5 Kg8
34 249 21631 107881990 Kb1 Kg7 Kc1 Kg6 Kd1 Kg5 Kc2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 Kf5 ex
d6 Kf6 Ke4 Kg5 Ke5 Kg6 Kf4 Kh5 Kf5 Kh6 Kxg4 Kg6 Kf4 Kf7 g4 Kg8 g5 Kf8 Kg4 Ke8 Kf
5 Kf7 g6+ Kg7 Kg5 Kg8 Kf6 Kh8 g7+ Kh7 Kf7 Kh6 g8=Q Kh5
34 249 27323 139270497 Kb1 Kg7 Kc1 Kg6 Kd1 Kg5 Kc2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 Kf5 ex
d6 Kf6 Ke4 Kg5 Ke5 Kg6 Kf4 Kh5 Kf5 Kh6 Kxg4 Kg6 Kf4 Kf7 g4 Kg8 g5 Kf8 Kg4 Ke8 Kf
5 Kf7 g6+ Kg7 Kg5 Kg8 Kf6 Kh8 g7+ Kh7 Kf7 Kh6 g8=Q Kh5
35 249 27853 141990586 Kb1 Kg7 Kc1 Kg6 Kd1 Kg5 Kc2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 Kf5 ex
d6 Kf6 Ke4 Kg5 Ke5 Kg6 Kf4 Kf7 Kxg4 Kg6 Kf4 Kg7 g4 Kg8 Kg5 Kf7 Kh6 Ke8 Kh5 Kf7 g
5 Kf8 g6 Kg7 Kg5 Kf8 Kh6 Kg8 g7 Kf7 Kh7 Kf6 g8=Q Ke5 Qg2
35 249 37871 196654059 Kb1 Kg7 Kc1 Kg6 Kd1 Kg5 Kc2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 Kf5 ex
d6 Kf6 Ke4 Kg5 Ke5 Kg6 Kf4 Kf7 Kxg4 Kg6 Kf4 Kg7 g4 Kg8 Kg5 Kf7 Kh6 Ke8 Kh5 Kf7 g
5 Kf8
36 856 40260 209519491 Kb1 Kg7 Kc1 Kg6 Kd1 Kg5 Kc2 Kh5 Kc3 Kh6 Kd4 Kg6 e5 Kf5 ex
d6 Kf6 Ke4 Kg5 Ke5 Kg6 Kf4 Kf6 Kxg4 Ke5 Kh5 Kxd6 g4 Kxd5 g5 Ke6 g6 d5 Kh6 d4 g7
Kf7 Kh7 d3 g8=Q+ Ke7
Do you use a depth based replacement scheme for the hash table? If so try replace always if hashkey==entry->hashKey and see if that fixes the problem.

Tony

Re: Hamsters randomizer in action

Post by Tony » Mon Jun 04, 2007 3:35 am

Alessandro Scotti wrote:After the comments from Michael and Ron I've managed to spend some time with Hamsters. As a result, my wife won't talk to me. As a side effect though, I think I've found something.

In this position (best move is Kb1):

[D]6k1/3p4/3p4/3P4/4P1p1/6P1/8/K7 w - - 0 1

I get the following PV:
....
There are huge score drops in some plies that suggest a serious bug somewhere. I've spent a few hours poking around to no avail, and yet I think this time asking for suggestions on the bug would be too much even for the collective wisdom of the CCC! :-)
So my question is... how do I debug something like this that happens million nodes away from the root?!? :o
My guess,

if you have some kingInsideTheSquareOfThePawn code, check it
else add it.

A no presence of this code can give this kind of beheavior.

Tony

Pradu
Posts: 287
Joined: Sat Mar 11, 2006 2:19 am
Location: Atlanta, GA
Contact:

Re: Hamsters randomizer in action

Post by Pradu » Mon Jun 04, 2007 6:24 am

Tony wrote:if you have some kingInsideTheSquareOfThePawn code...
A no presence of this code can give this kind of beheavior.
Tony
Actually it doesn't. Buzz doesn't have this kind of code but scores seem stable. Buzz's eval is very simple here too, pretty much basic pawn eval (doubled, isolated...) + passers. There's no king proximity to the pawn or any bonus for the king being closer to the center squares. Only a difference in pawn structure can change the score here for Buzz...although might be another reason for Buzz's sluglishness to get good ply in this position.

Tony

Re: Hamsters randomizer in action

Post by Tony » Mon Jun 04, 2007 6:47 am

Pradu wrote:
Tony wrote:if you have some kingInsideTheSquareOfThePawn code...
A no presence of this code can give this kind of beheavior.
Tony
Actually it doesn't. Buzz doesn't have this kind of code but scores seem stable. Buzz's eval is very simple here too, pretty much basic pawn eval (doubled, isolated...) + passers. There's no king proximity to the pawn or any bonus for the king being closer to the center squares. Only a difference in pawn structure can change the score here for Buzz...although might be another reason for Buzz's sluglishness to get good ply in this position.
Well.., yes.

Not having this code means that search will find out that the pawn promotes, meaning most of the time it becomes very inefficient.

Having it implemented wrong (fe an off by 1 wrt stm) would give more of a "I'm winning", "No I'm not" beheavior every other iteration, while the former can hide it with horizon moves (and will) and give less score jumps.

Tony

Gerd Isenberg
Posts: 2127
Joined: Wed Mar 08, 2006 7:47 pm
Location: Hattingen, Germany

Re: Hamsters randomizer in action

Post by Gerd Isenberg » Mon Jun 04, 2007 7:12 am

Tony wrote: Well.., yes.

Not having this code means that search will find out that the pawn promotes, meaning most of the time it becomes very inefficient.

Having it implemented wrong (fe an off by 1 wrt stm) would give more of a "I'm winning", "No I'm not" beheavior every other iteration, while the former can hide it with horizon moves (and will) and give less score jumps.

Tony
I have this zugzwang in mind:

[D] 8/3p4/3p4/3P3k/4PKp1/6P1/8/8 b - -

If your eval recognizes this as drawish, your search will almost try to reach this position as a leaf with black to move. It interacts with path-dependencies from hash. Avoidrep-pruning becomes quite efficient.

Beside possible passer and square of king issues, co-ordinating squares might be even more important in this kind of endings.

Alessandro may have some bugs in eval or search (hashing) - therefor it is important to get the wrong evaluated positions from the primary variations.

Gerd

Post Reply