Komodo 9 update (9.01)

[gordonr]

Quick question, did you have the full Win Loss Draw plus DTZ 6 piece table bases installed?

Yes. My syzygy folder is 149 GB in total. It has subfolders for dtz, dtz-1, dtz-2, dtz-3, dtz-4, dtz-5 (amongst other folders).

I've been using 6-piece syzygy for a while, including with Komodo 8, etc. and I've never seen a won endgame failing to be progressed. I'll experiment with other engines for the same endgame.

[gordonr]

Using Komodo 9.01, I can't reproduce the issue if I just analyse one of the positions from the endgame. I need to step around the game while analysing, in order for it to produce some unconvincing analysis. I think clearing hash helps reduce the chance of reproducing; not sure.

After seeing this in Arena 3.5, I tried Chessbase 12 and was able to reproduce it. After stepping around, I got Kb8 suggested for the following position.

[d]

If I then start from scratch with a new game and paste this position, I see the better Rf4 suggested.

Then opening the complete game again, clicking around move 112 while analysing, and stepping forward a move, and I see moves such as Kb8 or Kb7?!

[mjlef]

It is a bit strange since Komodo would normally return a score of +/- 1000 for a known win/loss at the root. the +/-250 score are for wins discovered during the search. Thanks for the positions. I have to reinstall 6 piece Syzygy on something to try and duplicate this behavior.

[gordonr]

It is a bit strange since Komodo would normally return a score of +/- 1000 for a known win/loss at the root. the +/-250 score are for wins discovered during the search. Thanks for the positions. I have to reinstall 6 piece Syzygy on something to try and duplicate this behavior.

Looking at the game score, during the 6 piece endgame, Komodo was usually using 250 or 318. I know that syzygy is different from Nalimov tablebases, but what is the expectation in terms of playing out won positions? Should Komodo play an optimal line every time or just something close enough to it?

Further experimenting with analysing the game, and I'm seeing that it is quite unpredicatable. Open game; start Komodo as kibitzer; click on 112.Rf6... then combinations of step forward and back; sometimes clearing hash inbetween. Sometimes, after 112...Ra1, I see Rf4 or Rf5, Rc6, Kb7, Kb8. All maintain the win but not equal distance according to shredderchess.com.

[MikeB]

It is a bit strange since Komodo would normally return a score of +/- 1000 for a known win/loss at the root. the +/-250 score are for wins discovered during the search. Thanks for the positions. I have to reinstall 6 piece Syzygy on something to try and duplicate this behavior.

Analysis mode in xboard with 4 core i& iMac. On my machine Komoko was seeing 17.9M nps.

45 +250.00 5.33G 5:04.23 Rf4 Rd1 Kc6 Rd8 Kc5 Rd1 Rd4 Rb1 Kd6 Kg6 Rf4 Re1 Rf6+ Kg7 h5 Rh1 h6+ Kh7 Ke6 Rg1 Rf7+ Kg8 Rg7+ Kh8 Rg6 Kh7 Kf7 Rb1 Rf6 Rb7+ Ke6 Rb6+ Kd7 Rb7+ Kc6 Rb4 Rf7+ Kg6 Rg7+ Kf5 h7 Rc4+ Kd7 Rd4+ Ke7 Re4+ Kf8 Rh4 Rd7 Kg6 Rf7 Rxh7
44 +250.00 5.05G 4:50.00 Rf4 Rd1 Kc6 Rd8 Kc5 Rd1 Rd4 Rb1 Kd6 Kg6 Kd7 Kg7 Rf4 Kg6 Rf6+ Kg7 h5 Rh1 Rg6+ Kf7 Rh6 Kg8 Ke7 Ra1 Rc6 Kg7 Rg6+ Kh7 Rd6 Kg7 Rd5 Re1+ Kd7 Kg8 Kd6 Kh7 Rc5 Re4 Ra5 Rc4 Rf5 Kg7 g6 Rd4+ Ke6 Rd3 Ra5 Re3+ Re5 Rb3 Kf5
43 +250.00 4.99G 4:47.12 Rf4 Rd1 Kc6 Rd8 Kc5 Rd1 Rd4 Rb1 Kd6 Kg6 Kd7 Kg7 Rf4 Kg6 Rf6+ Kg7 h5 Rg1 Rf5 Rh1 g6 Rd1+ Ke6 Re1+ Kd6 Rd1+ Ke7 Rg1 Rf7+ Kg8 Ke6 Re1+ Kd7 Rd1+ Ke8 Re1+ Re7 Rd1 Re5 Rh1 Rf5 Re1+ Kd7 Rb1 Ke6 Rb6+ Ke5 Rb8 Kf6 Rf8+ Ke6 Re8+ Kd7 Re1
43 +250.00 4.82G 4:37.46 Kd7 Rd1+ Ke7 Re1+ Re6 Rh1 Re4 Ra1 Rf4 Re1+ Kd6 Re8 Kd5 Re1 Re4 Rb1 Re6 Kf7 Rf6+ Kg7 Rh6 Re1 Kd6 Kf7 Rf6+ Kg8 Re6 Rh1 Rh6 Re1 Rf6 Rh1 Rf4 Kg7 Ke6 Kg6 Rf6+ Kh5 Rh6+ Kg4 Kf6 Re1 Kg7 Re4 g6 Kf5 Kf7 Ra4 g7 Ra7+ Kg8 Ke5
42 +250.00 4.65G 4:28.20 Kd7 Rd1+ Ke7 Re1+ Re6 Rh1 Re4 Ra1 Rf4 Re1+ Kd7 Re5 Kd6 Re2 Rf6 Ra2 Kd5 Ra5+ Ke6 Ra1 h5 Re1+ Kd7 Rg1 Rf5 Rh1 g6 Rd1+ Ke7 Rg1 Rd5 Ra1 Rd7 Ra6 Rc7 Kg8 Kd7 Kg7 Kc8+ Kg8 Kb7 Rd6 Rc6 Rd7+ Kb6 Rd5 h6 Kf8 Rf6+ Kg8 h7+ Kg7 Rf7+ Kxg6
42 +250.00 4.52G 4:21.71 Kd7 Rd1+
41 +250.00 4.04G 3:55.30 Kd7 Rd1+ Rd6 Rh1 Rh6 Ra1 Re6 Ra5 Ke7 Rf5 Kd6 Kf7 Re5 Rf1 h5 Rd1+ Rd5 Rf1 Kd7 Re1 Kd8 Ke6 g6 Kf6 Rd7 Ra1 Rf7+ Kg5 Ke7 Ra8 Rh7 Kf4 g7 Kg3 h6 Ra7+ Ke8 Ra8+ Kd7 Ra7+ Kc6 Kf2 g8=Q Ra6+ Kc7 Ra7+ Kc8 Ra8+ Kb7 Rxg8
41 +250.00 3.51G 3:26.36 Kd7
40 +250.00 3.41G 3:21.33 Kd7 Rd1+ Ke8 Re1+ Kd8 Rd1+ Ke7 Rg1 Rf7+ Kg8 Rf4 Kg7 Ke6 Kg6 Kd7 Rg2 Ke8 Re2+ Kf8 Re6 Rd4 Rb6 Rd7 Rb8+ Ke7 Rh8 Rd4 Rh7+ Ke6 Ra7 Rd6 Ra5 Ke7+ Kg7 Rf6 Ra7+ Ke6 Kg8 Ke5 Rh7 Rf4 Re7+ Kf5 Rf7+ Kg4 Ra7 Rf6 Ra4+ Kg3 Ra5 Rf3 Ra7
40 +250.00 3.27G 3:14.40 Kd7
39 +250.00 3.14G 3:07.29 Kd7 Rg1 Ke6 Re1+ Kf5 Rf1+ Ke5 Re1+ Kd6 Rh1 Rh6 Re1 h5 Rg1 Rg6+ Kf7 Rf6+ Kg7 Rf5 Rd1+ Rd5 Re1 Rc5 Rd1+ Ke6 Re1+ Kf5 Rf1+ Ke4 Re1+ Kf3 Rf1+ Kg2 Ra1 Rc7+ Kg8 Kf3 Ra5 Kf4 Ra4+ Ke5 Ra5+ Kf6 Ra6+ Ke7 Kg7 g6 Ra5 Kd6+ Kf6 Rf7+ Kg5 Rh7
39 +250.00 2.95G 2:57.59 Kd8 Rg1
39 +250.00 2.81G 2:50.17 Kd8 Rg1
38 +250.00 2.77G 2:48.38 Kd8 Rg1 Rf4 Re1 Kd7 Kg8 Kd6 Kg7 Kd5 Kg6 Rf6+ Kg7 h5 Rh1 h6+ Kh7 Ke6 Rg1 Rf7+ Kg8 Rg7+ Kh8 Rg6 Kh7 Kf7 Rc1 Rf6 Rc7+ Ke8 Rc8+ Ke7 Rc7+ Kd8 Rc5 Rf7+ Kg6 Rg7+ Kf5 h7 Ra5 g6 Ra8+ Kd7 Kg5 Rg8 Ra7+ Kc6 Ra6+ Kc5 Kg4 h8=Q Ra7
38 +250.00 2.71G 2:44.96 Kd8
37 +250.00 2.67G 2:43.15 Kd8 Rh1 Rf4 Re1 Kd7 Re2 Kd6 Re1 Kd5 Kg6 Rf6+ Kh5 Rh6+ Kg4 g6 Kf5 h5 Re5+ Kd6 Re6+ Kd7 Ra6 Rh7 Kf6 Rf7+ Kg5 Ke7 Ra8 Rh7 Ra7+ Kd6 Ra6+ Kc5 Ra8 g7 Kf6 h6 Ke5 Kb6 Rb8+ Kc6 Rc8+ Kd7 Rb8 Rh8 Rb7+ Kc6 Rb4 g8=Q
37 +250.00 2.65G 2:42.30 Kd8 Rh1
37 +250.00 2.65G 2:42.24 Kd8 Rh1
37 +250.00 2.65G 2:42.14 Kd8 Rh1
37 +250.00 2.64G 2:41.71 Kd8 Rh1
37 +250.00 2.63G 2:41.12 Kd8 Rh1
36 +250.00 2.62G 2:40.12 Kd8 Rh1 Rf4 Re1 Kd7 Re2 Kd6 Re1 Kd5 Kg6 Rf6+ Kh5 Rh6+ Kg4 g6 Kf5 h5 Rd1+ Kc6 Kf6 Rh7 Rc1+ Kd6 Rd1+ Kc7 Rc1+ Kd8 Ra1 Rf7+ Kg5 Ke7 Ra6 Rh7 Ra7+ Kd6 Ra6+ Kc5 Ra8 g7 Rg8 h6 Kf4 Kb6 Ra8 Rh8 Ra1 g8=Q
36 +250.00 2.59G 2:38.51 Kd8 Rh1
36 +250.00 2.57G 2:37.76 Kd8 Rh1
36 +250.00 2.51G 2:33.84 Kd8 Rh1
36 +250.00 2.50G 2:33.38 Kd8 Rh1
36 +250.00 2.49G 2:33.15 Kd8 Rh1
36 +250.00 2.49G 2:32.92 Kd8 Rh1
36 +250.00 2.49G 2:32.87 Kd8 Rh1
36 +250.00 2.49G 2:32.78 Kd8 Rh1
36 +250.00 2.47G 2:31.86 Kd8 Ra5
35 +250.00 2.41G 2:28.34 Kd8 Ra5 Ke7 Rb5 Ke6 Ra5 Rf5 Ra4 h5 Ra6+ Ke7 Ra7+ Kd6 Ra6+ Kc7 Ra7+ Kb6 Re7 Kc6 Re2 Kd6 Rh2 g6 Rd2+ Ke6 Re2+ Re5 Rb2 Rd5 Re2+ Kf5 Rf2+ Ke4 Rg2 Ke5 Re2+ Kf4 Kh6 Rd8 Rf2+ Ke4 Re2+ Kf5 Rf2+ Ke6 Re2+ Kf7 Rb2 Rf8 Kg5 g7 Kxh5
35 +6.44 1.81G 1:50.25 Kd8
35 +5.95 1.77G 1:47.98 Kd8
35 +5.62 1.63G 1:39.16 Kd8
35 +5.40 1.51G 1:31.82 Kd8
35 +5.25 1.48G 1:29.73 Kd8
35 +5.15 1.40G 1:25.16 Kd8
35 +5.08 1.30G 1:18.91 Kd8
34 +5.01 961.7M 0:59.30 Kd8 Ra8+ Ke7 Rh8 Rf7+ Kg6 Rf4 Rh7+ Ke6 Rh8 Rd4 Re8+ Kd7 Rh8 Ke7 Rh7+ Ke6 Rb7 Rd6 Rb5 Ke7+ Kg7 Rf6 Re5+ Kd6 Re1 h5 Rg1 Rg6+ Kf7 Kd7 Rg4 Kc6 Rh4 Rh6 Rg4 g6+ Kg7 Rh7+ Kf6 Kc7 Ra4 Kb7 Rd4 Kb6 Rg4 Rd7 Rg1
33 +5.01 938.7M 0:57.83 Kd8 Ra8+ Ke7 Rh8 Rf7+ Kg6 Rf4 Rh7+ Ke6 Rh8 Rd4 Re8+ Kd7 Rh8 Ke7 Rh7+ Ke6 Rb7 Rd6 Rb5 Ke7+ Kg7 Rf6 Ra5 Ke6 Rb5 Rf5 Rb6+ Ke5 Rb5+ Kf4 Rb4+ Kg3 Rb3+ Kg4 Rb2 Kh5 Rg2 Rf4 Rd2 Rf6 Rd3 Rg6+ Kf8 Rc6 Ke7 Rf6
33 +5.01 928.5M 0:57.14 Kd8
33 +4.28 833.3M 0:51.08 Kd8
33 +3.79 732.3M 0:44.83 Kd8
33 +3.46 682.1M 0:41.75 Kd8
33 +3.45 607.2M 0:37.07 Re6 Rh1 Rh6 Ra1 Kb6 Rc1 h5 Ra1 Re6 Kh8 Re8+ Kg7 Re7+ Kf8 Re5 Kg7 g6 Kh6 Kc6 Ra7 Kd6 Ra8 Ke7 Ra2 Rd5 Re2+ Kf6 Rf2+ Rf5 Ra2 Kf7 Ra7+ Kg8 Rg7+ Kf8 Rg8+ Ke7 Ra8 Rd5 Kg7 Rd7 Ra6 Kd8+
33 +3.24 539.0M 0:32.94 Re6
33 +3.09 458.4M 0:28.20 Re6
33 +3.00 408.6M 0:25.32 Kb7 Re1 h5 Rh1 Rg6+ Kh7 Rh6+ Kg8 Kc6 Ra1 Kb5 Rh1 Kc4 Ra1 Rg6+ Kh7 Rc6 Rg1 g6+ Kh6 Rc5 Rb1 Rd5 Rc1+ Kd4 Rg1 Rf5 Rd1+ Ke5 Re1+ Kf6 Ra1 Rd5 Ra6+ Ke7 Rc6 Rd7 Kg7 Kd8+ Kg8 Re7 Ra6 Kd7 Ra7+ Kd6 Ra5
33 +2.99 337.7M 0:21.12 Kb7
33 +2.92 284.2M 0:18.06 Kb7
32 +2.85 223.7M 0:14.57 Kb7 Rh1 Rh6 Re1 Kc6 Re5 Kd6 Ra5 Re6 Kf7 Rf6+ Kg8 Ke6 Kg7 Rf5 Ra4 h5 Ra6+ Kd7 Ra7+ Kc6 Ra6+ Kd5 Ra5+ Ke4 Ra4+ Ke5 Ra1 Kd6 Ra6+ Kc7 Ra1 Re5 Rd1 Re7+ Kg8 g6 Rh1 Re5 Kg7 Rf5 Rc1+ Kd6 Rd1+ Ke5 Re1+ Kd4 Rd1+ Kc5 Kh6
32 +2.84 184.3M 0:12.08 Kb7
32 +2.77 135.8M 0:09.16 Re6 Rh1 Rh6 Ra1 Kb7 Ra4 Kb6 Rd4 Kc5 Rd1 h5 Rb1 Rg6+ Kh7 Rb6 Rd1 Rh6+ Kg7 Rg6+ Kh7 Rd6 Rg1 g6+ Kh6 Rd5 Rh1 Kc6 Rh3 Rf5 Rc3+ Kd6 Rc8 Ke7 Rc7+ Ke6 Rc6+ Kd7 Ra6 Rd5 Ra1 Kd6
31 +2.77 98.9M 0:06.80 Re6 Rh1 Rh6 Ra1 Kb7 Ra4 Kb6 Rd4 Kc5 Rd1 h5 Ra1 Rb6 Rh1 Rg6+ Kf7 Rh6 Kg8 Kd6 Rg1 Rg6+ Kf7 Ke5 Re1+ Kf5 Rf1+ Kg4 Rg1+ Kf4 Rf1+ Kg3 Rg1+ Kf3 Rf1+ Kg2 Rf4 Kg3 Rf5 Kg4 Rf1 Rb6 Rg1+ Kf4 Rf1+ Ke4 Re1+ Kf3 Rf1+ Kg2 Rc1
30 +2.76 80.0M 0:05.61 Re6 Rh1 Rh6 Ra1 Kb7 Ra4 Kb6 Rd4 Kc5 Rd1 h5 Ra1 Rb6 Rh1 Rg6+ Kh7 Rh6+ Kg7 Kd6 Rg1 Rg6+ Kf7 Ke5 Re1+ Kf5 Rf1+ Ke4 Re1+ Kf4 Rf1+ Kg4 Rg1+ Kf3 Rf1+ Kg2 Rf4 Kg3 Rf1 Rc6 Rg1+ Kf4 Rf1+ Ke3 Re1+ Kf3 Rf1+ Kg2 Rf5
29 +2.76 70.7M 0:05.00 Re6 Rh1 Rh6 Ra1 Kb7 Ra4 Kb6 Rd4 Kc5 Rd1 Kc6 Kf7 h5 Kg7 Rg6+ Kh7 Rf6 Rg1 Rf7+ Kg8 g6 Rh1 Rh7 Rd1 Kc7 Rd2 Kb7 Rd5 Kb6 Rd1 Kc6 Rd2 Kc5 Rd1
28 +2.76 63.0M 0:04.48 Re6 Rh1 Rh6 Ra1 Kb7 Ra4 Kb6 Rd4 Kc5 Rd1 Kc6 Rd4 h5 Ra4 Rd6 Ra1 Rd7+ Kg8 g6 Rh1 Rh7 Rd1 Kc7 Rd2 Kb7 Rd5 Kb6 Rd1 Rc7 Rh1 Rc5 Kg7
27 +2.76 46.6M 0:03.33 Re6 Rh1 Rh6 Ra1 Kb6 Re1 Kc6 Re5 Kd6 Ra5 Re6 Kf7 Rf6+ Kg7 Kd7 Ra7+ Ke6 Ra5 Rf5 Ra6+ Ke5 Ra1 Kf4 Ra4+ Kg3 Ra1 Kg4 Rg1+ Kh5 Rc1 Rf6 Re1 Rd6 Re7 Kg4
26 +2.76 42.7M 0:03.06 Re6 Rh1 Rh6 Ra1 Kb6 Re1 Kc6 Re5 Kd6 Ra5 Re6 Kf7 Rf6+ Kg7 Ke6 Kh7 Rf5 Ra6+ Ke5 Ra4 Rf4 Ra5+ Kf6 Ra6+ Kf5 Rb6 Kg4 Rb1 Rf6 Rb4+ Kh5 Rb7 Rg6
25 +2.76 39.1M 0:02.84 Re6 Rh1 Rh6 Ra1 Kb6 Re1 Kc6 Re5 Kd6 Re1 h5 Rg1 Rg6+ Kf7 Kc7 Rh1 Rh6 Rg1 Rh7+ Kg8 g6 Rd1 h6 Rc1+ Kd7 Rd1+ Ke6 Re1+ Kf6 Rf1+ Ke5 Re1+ Kf4 Rf1+ Kg4 Rg1+ Kh5 Rh1+ Kg5 Rg1+ Kf5
24 +2.76 36.5M 0:02.65 Re6 Rh1 Rh6 Ra1 Kb6 Re1 Kc6 Re5 Kd6 Rf5 Ke6 Ra5 Rf6 Ra1 h5 Re1+ Kd6 Rg1 Rg6+ Kf7 Kc7 Rh1 Rh6 Kg7 Kb6 Rh2 Kc5 Rh1 Kd6 Rg1
23 +2.76 34.9M 0:02.54 Re6 Rh1 Rh6 Ra1 Kb6 Re1 Kc5 Rf1 Kd4 Rf5 Ke4 Rc5 Re6 Kf7 Rd6 Rb5 Rh6 Rb4+ Kf5 Rb5+ Kg4 Rb4+ Kh5 Kg7 Rg6+ Kh7 Re6 Kg7 Re7+ Kg8 Kg6
23 +2.76 33.2M 0:02.41 Re6
23 +2.54 30.4M 0:02.19 Re6
23 +2.30 23.0M 0:01.62 Rh6 Ra5
23 +2.39 22.5M 0:01.58 Rh6
23 +2.32 20.1M 0:01.39 Rh6 Ra5
23 +2.39 19.5M 0:01.34 Rh6
22 +2.32 10.6M 0:00.76 Rh6 Ra4 Kb6 Rc4 h5 Rc1 Rg6+ Kh7 Rf6 Rh1 Rh6+ Kg7 Kc7 Rc1+ Rc6 Rg1 Rg6+ Kf7 Kd7 Rg4 Rf6+ Kg7 Rf5 Rh4 g6 Kh6 Ke7 Re4+ Kf7 Rd4 Kf6 Rg4 Rb5 Rg2
21 +2.36 8.18M 0:00.62 Rh6 Ra4 Kb6 Rc4 h5 Rc1 Rg6+ Kh7 Rf6 Rh1 Rh6+ Kg8 g6 Rb1+ Kc7 Rc1+ Kd6 Rd1+ Ke6 Re1+ Kf6 Rf1+ Kg5 Rg1+ Kf5 Rf1+ Kg4 Rg1+ Kh3 Rh1+ Kg2 Rc1 Rh7 Rc3 Ra7 Rc5 Rh7
20 +2.36 7.73M 0:00.58 Rh6 Ra4 Kb6 Rc4 h5 Rc1 Rg6+ Kh7 Rf6 Rh1 Rh6+ Kg8 g6 Rb1+ Kc7 Rc1+ Kd6 Rd1+ Ke6 Re1+ Kf6 Rf1+ Kg5 Rg1+ Kf5 Rf1+ Kg4 Rg1+ Kh3 Rh1+ Kg2 Rc1 Rh7 Rc3 Ra7 Rc5 Rh7
19 +2.36 7.31M 0:00.56 Rh6 Ra4 Kb6 Rc4 h5 Rc1 Rg6+ Kh7 Rf6 Rh1 Rh6+ Kg8 g6 Rb1+ Kc7 Rc1+ Kd6 Rd1+ Ke6 Re1+ Kf6 Rf1+ Kg5 Rg1+ Kf5 Rf1+ Kg4 Rg1+ Kh3 Rh1+ Kg2 Rc1 Rh7 Rc3 Ra7 Rc5 Rh7
19 +2.32 6.24M 0:00.48 Rh6
19 +2.09 5.69M 0:00.44 Rd6 Rh1
19 +2.24 5.18M 0:00.40 Rd6 Rh1
19 +2.31 5.11M 0:00.40 Rd6
19 +2.24 4.80M 0:00.38 Rd6 Rh1
18 +2.31 4.35M 0:00.34 Rd6 Rh1 Rh6 Ra1 Kb6 Ra4 Kc5 Ra5+ Kd4 Rb5 Ke4 Rb2 Kf4 Rh2 Kf5 Rf2+ Ke6 Re2+ Kd6 Re1 Kd7 Re5 Kc6
17 +2.31 3.40M 0:00.27 Rd6 Rh1 Rh6 Ra1 Kb6 Ra4 h5 Rg4 Rg6+ Kh7 Kc6 Rh4 Rh6+ Kg7 Kc5 Rf4 Rg6+ Kh7 Rd6 Rg4 Rd7+ Kg8 Rd5 Rh4
16 +2.31 2.63M 0:00.22 Rd6 Rh1 Rh6 Ra1 Kb6 Ra4 h5 Rg4 Rg6+ Kh7 Kc5 Rh4 Rh6+ Kg7 Kd6 Rh1 Ke5 Rg1 Rg6+ Kh7 Kd6 Rg4 Ke6
16 +2.21 2.25M 0:00.19 Rd6
16 +2.04 1.21M 0:00.11 Kd6
16 +2.03 1.21M 0:00.11 h5 Ra5
16 +2.16 1.06M 0:00.10 h5
16 +2.08 972243 0:00.09 h5 Rh1
16 +2.18 883234 0:00.08 h5 Rh1
16 +2.25 796366 0:00.08 h5 Rg1
16 +1.70 1.76M 0:00.15 Kd6 Ra4
15 +2.32 719086 0:00.07 h5 Rg1 Rf5 Rh1 g6 Rc1+ Kd6 Rd1+ Rd5 Re1 Re5 Rd1+ Ke7 Ra1 Rd5 Rf1 Kd7 Rg1 Kd8 Kh6 Ke8
14 +2.32 649713 0:00.06 h5 Rg1 Rf5 Rh1 g6 Rc1+ Kd6 Rd1+ Rd5 Re1 Re5 Rd1+ Ke7 Ra1 Rd5 Rf1 Kd7 Ra1 Ke8
14 +2.31 625290 0:00.06 h5
13 +2.24 531471 0:00.06 h5 Rg1 Rf5 Rh1 g6 Rc1+ Kd6 Rd1+ Rd5 Re1 Re5 Rd1+ Ke7 Rb1 Kd7 Kh6 Rf5 Re1 Kd8
13 +2.18 510282 0:00.05 h5
13 +2.11 444670 0:00.05 h5 Rg1
12 +2.18 422188 0:00.05 h5 Rg1 Rf5 Rh1 g6 Rc1+ Kd6 Rd1+ Rd5 Rg1 Re5 Kh6 Rc5 Rd1+ Ke6 Re1+ Re5 Rg1
12 +2.12 366800 0:00.04 h5 Rg1
12 +2.24 362446 0:00.04 h5 Kh7
12 +2.31 293337 0:00.04 h5
11 +2.24 263562 0:00.03 h5 Ra7+ Kb6 Re7 Rd6 Re5 h6+ Kh7 Rd7+ Kh8 Rg7 Rd5 g6 Rh5 Rh7+ Kg8 Kc6
11 +2.21 259389 0:00.03 h5 Ra7+
11 +2.34 228357 0:00.03 h5
11 +2.25 225470 0:00.03 h5 Kh7
11 +2.32 177026 0:00.02 h5
11 +2.25 175490 0:00.02 h5
10 +2.18 162337 0:00.02 h5 Ra7+ Kd8 Kh7 Ke8 Ra5 Rf7+ Kg8 Rf8+ Kg7 h6+ Kg6 Rg8+ Kh7 Rg7+ Kh8 Rg6
10 +2.23 131180 0:00.02 h5
10 +2.16 125426 0:00.02 h5 Rc1+
9 +2.23 116954 0:00.02 h5 Rc1+ Kd7 Rd1+ Rd6 Rg1 Rd5 Re1 Kc6 Rc1+ Kb5 Rb1+ Kc5 Rb7
9 +2.19 113401 0:00.02 h5
9 +2.07 109432 0:00.02 h5 Rc1+
9 +2.16 97935 0:00.02 h5
9 +2.09 94500 0:00.02 h5 Rc1+
9 +2.16 86014 0:00.02 h5
8 +2.09 74258 0:00.02 h5 Rc1+ Kd6 Rd1+ Ke5 Re1+ Kd5 Rd1+ Ke6 Re1+ Kd7 Rd1+ Ke7 Re1+ Re6 Rg1
8 +2.13 70155 0:00.02 h5 Rc1+
8 +2.20 56862 0:00.01 h5
8 +2.13 50703 0:00.01 h5 Kh7
7 +2.20 37047 0:00.01 h5 Kh7 Rh6+ Kg8 Kd6 Ra5 g6 Ra7 Ke5
7 +2.08 29335 0:00.01 h5
7 +1.96 27963 0:00.01 h5 Kh7
7 +2.06 24063 0:00.01 h5 Kh7
7 +2.14 23268 0:00.01 h5 Kh7
6 +2.22 19340 0:00.01 h5 Kh7 Rf7+ Kg8 Re7 Ra5 Re8+ Kg7
5 +2.22 13108 0:00.01 h5 Rh1 h6+ Kh7 Rf7+ Kg6 Rg7+ Kf5 Kd6
5 +2.16 8828 0:00.01 h5
5 +2.06 8662 0:00.01 h5
5 +1.94 7765 0:00.01 Kd6 Rh1 Rh6 Re1 Rf6
5 +1.94 7477 0:00.01 Kd6
4 +1.87 3926 0:00.01 Kd6 Rh1 Rh6 Rc1
3 +1.96 3015 0:00.01 Kd6 Rh1 Rf4
2 +2.05 1483 0:00.01 Kd6 Ra8
2 +1.91 828 0:00.01 h5 Rh1 h6+
1 +2.16 299 0:00.01 h5
0 #

[mjlef]

It is a bit strange since Komodo would normally return a score of +/- 1000 for a known win/loss at the root. the +/-250 score are for wins discovered during the search. Thanks for the positions. I have to reinstall 6 piece Syzygy on something to try and duplicate this behavior.

Looking at the game score, during the 6 piece endgame, Komodo was usually using 250 or 318. I know that syzygy is different from Nalimov tablebases, but what is the expectation in terms of playing out won positions? Should Komodo play an optimal line every time or just something close enough to it?

Further experimenting with analysing the game, and I'm seeing that it is quite unpredicatable. Open game; start Komodo as kibitzer; click on 112.Rf6... then combinations of step forward and back; sometimes clearing hash inbetween. Sometimes, after 112...Ra1, I see Rf4 or Rf5, Rc6, Kb7, Kb8. All maintain the win but not equal distance according to shredderchess.com.

Syzygy do not have distance to win (or loss or draw). They use two sets of tables. One has an indication if the position will lead to a win, a loss or a draw with perfect play. The second table (DTZ) indicates how many moves to a conversion (a capture, a loss, a draw a win..). So, there is not enough information to play perfectly. Once a program gets down to a 6 piece engine, it should be able to at least play the move(s) that lead to the fastest conversion. Once it converts, it does the same thing. But this leads to kinda weird play where the program often makes strange looking (but equally good in terms of moves to conversion) moves. One way to get "better looking" moves is to combine the root move lookup with some searching. But even with these kinds of additions, the move selection can still seem a bit strange. Not that full move to win tablebases are much larger and not fast enough on present machines to be used very deep in the search. Even ram drives are nowhere near the search speeds of current programs. The Syzygy scheme is a clever compromise to shrink the data enough to be useful.

So saying this, I am not the expert on Syzygy. I will ask Jeremy to take a look (he added the Syzygy TB to Komodo).

[shrapnel]

Hi Mark
I saw someone using an Engine called Komodo 9.2 RC or something like that yesterday in the Infinity Chess engine Room.
When will your loyal customers get the update ?

[mjlef]

Hi Mark
I saw someone using an Engine called Komodo 9.2 RC or something like that yesterday in the Infinity Chess engine Room.
When will your loyal customers get the update ?

I have not made a Komodo 9.2 RC, so I am unsure where that name is coming from. We sometimes give copies to some testers to test, of course, and help us know if a version is an actual improvement. Changes of a few elo take many thousands of games, and we do not plan on releasing anything until we are convinced it is a significant improvement, or the release fixes some bug or issue. For bug fix releases, the timing will depend on how many people it effects or how bag the bug is. For strength improvements, we have a target of about 10 elo better than the last release.

Note the Komodo 9.01 update does help quite a bit in sudden death (no increment) play, but changes nothing for increment play.

[shrapnel]

Hi Mark
I saw someone using an Engine called Komodo 9.2 RC or something like that yesterday in the Infinity Chess engine Room.
When will your loyal customers get the update ?

I have not made a Komodo 9.2 RC, so I am unsure where that name is coming from. We sometimes give copies to some testers to test, of course, and help us know if a version is an actual improvement. Changes of a few elo take many thousands of games, and we do not plan on releasing anything until we are convinced it is a significant improvement, or the release fixes some bug or issue. For bug fix releases, the timing will depend on how many people it effects or how bag the bug is. For strength improvements, we have a target of about 10 elo better than the last release.

Note the Komodo 9.01 update does help quite a bit in sudden death (no increment) play, but changes nothing for increment play.

Thanks for the quick response.
Well, it seemed genuine at the time because the player using it had a high Rating and the letters "RC" seemed to indicate it was a Release Candidate. Maybe some joker with a Hex Editor.
Thank You