Grand Chess

[hgm]

An even simpler case, where tried to play a KIKCC end-game

8/8/8/8/8/c1c5/2I1k3/2K5 w - - 0 1
[d]8/8/8/8/8/b1b5/2R1k3/2K5 w - - 0 1
(The Bishops represent Chameleons, the Rook an Immobilizer)

Black claims a win in this position. There must be something wrong with the way it thinks Chameleons interact with Kings. Earier it also shied away from capturing the immobilized and nprotected Chameleon with its King.

[Daniel Shawul]

Indeed there was a stupid bug in attacks yet again. It was checking all the surrounding squares of the king to see if a chamleon is attacking it, instead of just checking if a chameleon is there. Fixed it now. The suicide moves are also changed to 0@c5. https://sites.google.com/site/dshawul/N ... ects=0&d=1

Btw it is good that you are only using it from the endgame because I noticed that there is large tree explosion in qsearch...sometimes it searches only 3 plies. Well ofcourse that is due to MVV/LVA which in this case is completely useless since I still have the code for the chess kind of capture. I don't know how to do that now since I don't have the "side effects" captures during move generation. The make move takes care of all that..so I need to think about that.

I was writing a simple "selfplay" code when I realized the two engines share everything. Hashtables, eval parameters etc..I hope that will not bias the piece value determination in any way, can it ?

Code: Select all

setboard 8/8/8/8/8/c1c5/2I1k3/2K5 w - - 0 1
d

            a b c d e f g h
            * * * * * * * * * * * * * * * * * * * * * * * *
            * * * * * * * * * * * * * * * * * * * * * * * *
            * * * * * * * * * * * * * * * * * * * * * * * *
            * * * * * * * * * * * * * * * * * * * * * * * *
          8 . . . . . . . . * * * * * * * * * * * * * * * * 8
          7 . . . . . . . . * * * * * * * * * * * * * * * * 7
          6 . . . . . . . . * * * * * * * * * * * * * * * * 6
          5 . . . . . . . . * * * * * * * * * * * * * * * * 5
          4 . . . . . . . . * * * * * * * * * * * * * * * * 4
          3 c . c . . . . . * * * * * * * * * * * * * * * * 3
          2 . . I . k . . . * * * * * * * * * * * * * * * * 2
          1 . . K . . . . . * * * * * * * * * * * * * * * * 1
            * * * * * * * * * * * * * * * * * * * * * * * *
            * * * * * * * * * * * * * * * * * * * * * * * *
            * * * * * * * * * * * * * * * * * * * * * * * *
            * * * * * * * * * * * * * * * * * * * * * * * *
            a b c d e f g h

        8/8/8/8/8/c1c5/2I1k3/2K5 w - - 0 1

analyze
3 -29997 0 236  0@c2 c3b2
4 -540 0 757  c1b1 e2f3 b1c1 a3c5
5 -535 0 1309  c1b1 e2e3 b1c1 a3c5 c1b2
5 -535 0 1311  c1b1 e2e3 b1c1 a3c5 c1b2
6 -545 0 1741  c1b1 e2e3 b1c1 a3c5 c1d1 e3d4
6 -545 0 1743  c1b1 e2e3 b1c1 a3c5 c1d1 e3d4
7 -545 1 2050  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1
7 -545 1 2052  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1
8 -550 1 2668  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 e3d4
8 -550 1 2670  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 e3d4
9 -545 1 3284  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1
9 -545 3 3286  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1
10 -550 3 4168  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 e3e4
10 -550 3 4170  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 e3e4
11 -545 3 5678  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1
11 -545 4 5682  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1
12 -550 4 8181  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e3d4
12 -550 4 8185  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e3d4
13 -545 6 12365  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4e5 e1d
1
13 -545 6 12369  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4e5 e1d
1
14 -550 7 19726  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4e5 e1d
1 e3d4
14 -550 7 19730  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4e5 e1d
1 e3d4
15 -545 9 30262  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4e5 e1d
1 e5d4 d1e1
15 -545 9 30266  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4e5 e1d
1 e5d4 d1e1
16 -550 10 37699  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4e5 e1
d1 e5d4 d1e1 e3e4
16 -550 10 37703  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4e5 e1
d1 e5d4 d1e1 e3e4
17 -545 12 50233  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4e5 e1
d1 e5d4 d1e1 d4e4 e1d1
17 -545 12 50237  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4e5 e1
d1 e5d4 d1e1 d4e4 e1d1
18 -550 15 83248  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4e5 e1
d1 e5d4 d1e1 d4e4 e1d1 e3d4
18 -550 17 83252  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4e5 e1
d1 e5d4 d1e1 d4e4 e1d1 e3d4
19 -545 18 111373  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4e5 e
1d1 e5d4 d1e1 d4e4 e1d1 e4e5 d1e1
19 -545 20 111390  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4e5 e
1d1 e5d4 d1e1 d4e4 e1d1 e4e5 d1e1
20 -550 21 134164  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4e5 e
1d1 e5d4 d1e1 d4e4 e1d1 e4e5 d1e1 e3d4
20 -550 23 134180  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4e5 e
1d1 e5d4 d1e1 d4e4 e1d1 e4e5 d1e1 e3d4
21 -545 28 189575  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4e5 e
1d1 e3d4 d1d2 e5e3 d2e1 d4e4 e1d1 e4d5 d1e1
21 -545 28 189596  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4e5 e
1d1 e3d4 d1d2 e5e3 d2e1 d4e4 e1d1 e4d5 d1e1
22 -550 37 296672  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4e5 e
1d1 e3d4 d1d2 e5e3 d2e1 d4e4 e1d1 e4d5 d1e1 e3e4
22 -550 37 296694  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4e5 e
1d1 e3d4 d1d2 e5e3 d2e1 d4e4 e1d1 e4d5 d1e1 e3e4
23 -550 43 364625  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4g2 e
1d1 g2e2 d1c1 e3d4 c1b2 e2e5 b2b3 e5c5 b3b2 c5c4 b2c1
23 -550 45 364650  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4g2 e
1d1 g2e2 d1c1 e3d4 c1b2 e2e5 b2b3 e5c5 b3b2 c5c4 b2c1
24 -550 59 521594  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4g2 e
1d1 g2e2 d1c1 e3d4 c1b2 e2e5 b2b3 e5c5 b3a3 c5c4 a3a4 c4d5
24 -550 59 521616  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4g2 e
1d1 g2e2 d1c1 e3d4 c1b2 e2e5 b2b3 e5c5 b3a3 c5c4 a3a4 c4d5
25 -545 82 789568  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4g2 e
1d1 g2e2 d1c1 e3d4 c1b2 e2e5 b2b3 e5c5 b3a3 c5c4 a3a4 d4e4 a4a5
25 -545 82 789590  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4g2 e
1d1 g2e2 d1c1 e3d4 c1b2 e2e5 b2b3 e5c5 b3a3 c5c4 a3a4 d4e4 a4a5
26 -550 93 890931  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4g2 e
1d1 g2e2 d1c1 e3d4 c1b2 e2e5 b2b3 e5c5 b3a3 c5c4 a3a4 d4e4 a4a5 c4d5
26 -550 93 890956  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4g2 e
1d1 g2e2 d1c1 e3d4 c1b2 e2e5 b2b3 e5c5 b3a3 c5c4 a3a4 d4e4 a4a5 c4d5
27 -545 117 1160611  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4g2
 e1d1 g2e2 d1c1 e3d4 c1b2 e2e5 b2c1 e5d5 c1d2 d5e4 d2e2 e4e3 e2d1 d4d5 d1e1
27 -545 118 1160636  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4g2
 e1d1 g2e2 d1c1 e3d4 c1b2 e2e5 b2c1 e5d5 c1d2 d5e4 d2e2 e4e3 e2d1 d4d5 d1e1
28 -545 143 1454272  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4g2
 e1d1 g2e2 d1c1 e3d4 c1b2 e2e5 b2c1 e5d5 c1b2 d5c4 b2c1 d4e3 c1b2 e3e2 b2a3 e2e3

28 -545 145 1454294  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4g2
 e1d1 g2e2 d1c1 e3d4 c1b2 e2e5 b2c1 e5d5 c1b2 d5c4 b2c1 d4e3 c1b2 e3e2 b2a3 e2e3

29 -545 178 1842121  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4g2
 e1d1 g2e2 d1c1 e3d4 c1b2 e2e5 b2c1 e5d5 c1b2 d5c4 b2c1 c4e6 c1d2 e6e3 d2e1 d4e4
 e1d1
29 -545 179 1842143  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4g2
 e1d1 g2e2 d1c1 e3d4 c1b2 e2e5 b2c1 e5d5 c1b2 d5c4 b2c1 c4e6 c1d2 e6e3 d2e1 d4e4
 e1d1
30 -545 223 2357032  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4g2
 e1d1 g2e2 d1c1 e3d4 c1b2 e2e5 b2a2 e5d5 a2b3 d5c4 b3b2 c4c5 b2a2 d4e3 a2a3 c5c4
 a3a4 e3d4
30 -545 225 2357077  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4g2
 e1d1 g2e2 d1c1 e3d4 c1b2 e2e5 b2a2 e5d5 a2b3 d5c4 b3b2 c4c5 b2a2 d4e3 a2a3 c5c4
 a3a4 e3d4
31 -545 289 3080357  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4g2
 e1d1 g2e2 d1c1 e3d4 c1b2 e2e5 b2a2 e5d5 a2b3 d5c4 b3b2 c4c5 b2a2 d4e3 a2a3 c5c4
 a3a4 e3d4 a4a5
31 -545 289 3080402  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4g2
 e1d1 g2e2 d1c1 e3d4 c1b2 e2e5 b2a2 e5d5 a2b3 d5c4 b3b2 c4c5 b2a2 d4e3 a2a3 c5c4
 a3a4 e3d4 a4a5
32 -545 381 4129966  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4g2
 e1d1 g2e2 d1c1 e3d4 c1b2 e2e5 b2c1 e5e3 c1d1 d4e4 d1c1 e4d5 c1d1 d5c4 d1e1 c4c5
 e1f1 c5d5 f1g2 e3e4
32 -545 382 4130011  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4g2
 e1d1 g2e2 d1c1 e3d4 c1b2 e2e5 b2c1 e5e3 c1d1 d4e4 d1c1 e4d5 c1d1 d5c4 d1e1 c4c5
 e1f1 c5d5 f1g2 e3e4
33 -545 495 5406881  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4g2
 e1d1 g2e2 d1c1 e3d4 c1b2 e2e5 b2c1 e5e3 c1d1 d4e4 d1c1 e4d5 c1d1 d5c4 d1e1 c4c5
 e1f1 c5d4 f1e1 d4e4 e1d1
33 -545 496 5406931  c1b1 e2e3 b1c1 a3c5 c1d1 c5d5 d1e1 d5d4 e1d1 d4e4 d1e1 e4g2
 e1d1 g2e2 d1c1 e3d4 c1b2 e2e5 b2c1 e5e3 c1d1 d4e4 d1c1 e4d5 c1d1 d5c4 d1e1 c4c5
 e1f1 c5d4 f1e1 d4e4 e1d1
exit
nodes = 5530487 <31 qnodes> time = 5734ms nps = 964507
splits = 0 badsplits = 0

[Daniel Shawul]

Yesterday I tried to use nullmove to detect possible check extension. I had a mate threat extension already but that was triggered if there was exactly a mate on the next ply. So I changed that to whatever big score we set for king captures and other threats as well...it didn't work that well. But I did not try the stalemate trick though.

[Daniel Shawul]

Indeed, end-game values cn be determned through the same method, but did not do very many of them. Just to make sure the surprisingly high Archbishop value was not game-stage dependent. (Sometimes you have this, that a piece that is essentially worthless in the end-game has a very good forking power on a crowded board, so that you can almost always force a trade for something intrinsically more valuable. For instance with a Camel (= (3,1) leaper, where Knight = (2,1) leaper).)

I am digesting the rest of your post. It was also surprising for me that even A and C were equal. Orthodox chess mentality has it that bishops only attack once color but in this case it only takes a tempo for A to switch colors. If the bishops or archbishops had a move like the lieutenants in spartan chess (i.e sideways without capture) I bet they would be equal to their "rookish" counterparts.

[hgm]

The suicide moves are also changed to 0@c5. https://sites.google.com/site/dshawul/N ... ects=0&d=1

OK, I still have to add that null move. Problem is that I have too many WinBoard branches now, which differ in the place that needs to be changed. (Namely the parser. And i also have to merge the Grand Chess patch with the Alien branch.)

Btw it is good that you are only using it from the endgame because I noticed that there is large tree explosion in qsearch...sometimes it searches only 3 plies. Well ofcourse that is due to MVV/LVA which in this case is completely useless since I still have the code for the chess kind of capture. I don't know how to do that now since I don't have the "side effects" captures during move generation. The make move takes care of all that..so I need to think about that.

Indeed, I noticed the drop in depth too. MVV works so well n Chess because the most valuable piece is also the one that can go on the rampage with captures. In Ultima it is not so obvious that you would want to capture the Immobilizer first. It cannot capture anything anyway. But it is still effective in raisng beta for the opponent, of course, so that he is more likely to run out of non-futile captures.

I would still recommend the Fairy-Max way: Do IID during QS, and preceed QS (which is a 1-ply capture search with QS reply) by a 1-ply capture search with stand-pat reply (perhaps based on lazy eval). In this first 'iteration' you would calculate the material gains from the captures (and suppress beta cutoffs, so that all captures are 'searched'), and the best move would be sorted to front, and tried first in the real QS iteration that follows. That is equivalent to MVV, but correctly takes account of summing victims of multiple captures. LVA might still make sense in Ultima, even though in general the capturer moves to a different square than where the vctim was: it is still on a new square, and might present a vulnerability that was not there before, so it is better to risk pieces of low value.

I was writing a simple "selfplay" code when I realized the two engines share everything. Hashtables, eval parameters etc..I hope that will not bias the piece value determination in any way, can it ?

I don't think so. It would be if you play different tunings against each other, or different search routines that share the hash table. But if they are really identical, usng the knowlege gathered by the opponent is equivalent to always having a hash hit when pondering the position.

[George Tsavdaris]

Ayee...that hurts. But it keeps the motivation burning inside Allow me to blame this loss on untuned material values.
cheers

Be my guest to tune as long as you want the material. Result will still be the same. Gothic Chess as also generally Capablanca Random Chess programs can't cope with my style on this game(CRC and variations) at blitz/rapid time controls(i lose at long time controls and always win at correspondence time controls).

Ed Trice tried to fix this countless times in his Gothic Vortex program but still didn't manage to find a way to prevent me from winning.

I noticed the Nebiyu version i downloaded(1.2) doesn't castle even when it has the chance. Does it know how to castle? Or i just downloaded an older version?

Anyway with the following hard fought(by both sides) draw the score is now 1.5-0.5 in favor of me.

That was an amazing game!!!! Nebiyu thought initially it is better, then that it is a lot better while i knew that my attack would be unstoppable. A classic theme. And indeed after launching my attack Nebiyu realized it is losing badly.
But in move 41 it found an amazing defensive resource that i didn't think of, so it turned the table and it was winning i guess. But i found a very clever endgame theme which i lead Nebiyu to it and it fall to it so it was losing again!
But i misplayed 1-2 moves so the result was Nebiyu's Archbishop against 3 Pawns and a Bishop of mine which was a draw! Really nice and educative game....
I should have won in the first place with the initial attack, but then Nebiyu should have won after move 41 and after this failure of it, i should have won the won endgame. We both failed to convert so a draw was a fair result i guess.

I have to say that i should give the throne of the best positional CRC(Capablanca ranom Chess and variations) engine to Nebiyu and dethrone Smirf that was the best positional engine until now. Not that it was any good, all CRC engines suck big time at this, but it was better than the others.
Nebiyu is a kind of Shredder in this. Aggressive when it needs to be but generally not to forcing, good placement of its pieces and i noticed a very very good Pawn play and a good handle of the center.
A tough opponent. I will play 8 more games to see how it really would do.

[Event "Computer Chess Game"]
[Site "George"]
[Date "2011.03.21"]
[Round "2"]
[White "NebiyuChess_1.2"]
[Black "George"]
[Result "1/2-1/2"]
[TimeControl "40/300"]
[Variant "gothic"]
[Annotator "1. +0.15"]

1. Nh3 {+0.15/15 2.4} h6 2. Nc3 {+0.15/14 1.7} c6 3. g3 {+0.35/13 5} g6 4.
e4 {+0.50/14 12} Nj6 5. Bf3 {+0.55/13 1.4} i6 6. d4 {+0.65/13 15} Bi7 7.
Kg2 {+0.85/12 7} Ah7 8. Be3 {+0.90/13 7} O-O 9. Qd2 {+1.20/14 13} g5 10.
Ae2 {+1.05/13 12} Cg8 11. g4 {+1.10/12 9} Na6 12. Rf1 {+1.20/12 9} d6 13.
j4 {+1.20/11 19} Ni4 14. Cd3 {+1.25/12 8} Be6 15. d5 {+1.15/14 9} cxd5 16.
exd5 {+0.85/14 8} Bd7 17. Bd4 {+1.55/15 12} f6 18. Ce3 {+1.25/13 10} Aj6
19. i3 {+1.20/12 26} Kj8 20. j5 {+0.45/10 12} i5 21. Rae1 {+0.75/12 22} Qe8
22. a3 {+1.25/10 17} Qf7 23. b4 {+0.95/11 17} Rh7 24. Be4 {+1.10/12 20} Rh8
25. f3 {+0.65/10 7} h5 26. Nd1 {+0.40/11 5} hxg4 27. fxg4 {-1.70/15 1.5}
Nxh2 28. Kxh2 {-1.10/13 0.4} Qj3 29. Rh1 {-1.75/13 0.5} Ai4+ 30. Kg1
{-2.45/15 0.4} Qxi3+ 31. Kf2 {-2.35/16 0.4} Rxh3 32. Rxh3 {-2.25/17 0.5}
Axh3+ 33. Cxh3 {-2.20/16 0.7} Qxh3 34. Rh1 {-2.25/15 0.4} Qxj5 35. Ke1
{-2.25/14 0.4} i4 36. Qc3 {-2.20/13 0.4} Qj2 37. Nf2 {-2.50/15 0.3} Cc8 38.
Qf3 {-1.95/14 0.5} Bj6 39. Af1 {-1.65/14 0.4} Qi2 40. Ae3 {-1.85/14 0.5}
Nc7 41. Bxf6 {+2.10/18 26} exf6 42. Qxf6 {+2.50/20 8} Qxh1+ 43. Bxh1
{+2.50/18 15} Bi7 44. Be4 {+2.20/18 18} Nxd5 45. Axd5 {+2.10/18 13} Bc6 46.
Ae3 {+1.80/19 43} Bxe4 47. Nxe4 {+1.05/18 0.8} i3 48. Ng3 {+0.80/18 26} Ch8
49. Kd1 {+1.10/18 51} Ch3 50. Af1 {+1.25/17 42} Ch4 51. Qf2 {+0.85/16 3}
Rh8 52. Qe2 {+1.20/17 0.8} j5 53. c3 {+1.65/14 0.5} j4 54. Qf2
{+0.35/16 1.0} j3 55. Nh5 {-2.20/19 4} Rxh5 56. gxh5 {-1.55/19 0.8} i2 57.
Qxh4 {+3.50/20 0.9} gxh4 58. Ah3 {+3.70/18 0.8} Bh6 59. b5 {+2.95/18 0.9}
Kj7 60. Ke2 {+3.60/18 1.7} b6 61. Kf3 {+3.20/19 1.6} Bg7 62. h6
{+3.55/21 0.7} Be5 63. h7 {+0.00/22 0.8} Ki6 64. Ag5+ {+0.00/23 0.9} Ki5
65. h8=C {+0.00/23 1.4} Bxh8 66. Ah3+ {+0.00/27 0.5} Ki6 67. Aj4+
{+0.00/27 0.6} Ki7 68. Ah5+ {+0.00/28 0.8} Kj8 69. Af4+ {+0.00/30 0.8} Bi7
70. Ah3 {+0.00/29 6}
1/2-1/2

Almost final position before the draw by repetition become unavoidable:

[Daniel Shawul]

George, it is good you are having fun. But you should know Nebiyu does not have any kind of positional terms in eval except for piece square tables. Its search is deep though so that fills up lots of voids. It also centralizes everything so if you can make passers on the wing , you can beat it easily.
Enjoy. I will follow your games and hope to see you beaten