The Real Test for Komodo 4

[lkaufman]

For me personally the only thing that matters about an engine is the quality of its analysis. Any modern engine will crush almost any human player, there's no need for new engines just to play against. My biggest motivation for Komodo was to create a superior analysis tool. I believe that Komodo produces the most accurate analysis of any one core program now (given a minute or so to think), but I don't expect anyone to take my word for it, as I am hardly neutral. That's why we focus on the Elo rating, it's the only generally accepted measure of how well a program plays chess, which people take as the same as how well it will analyze positions. I would love to see a better objective measure of this.

What is your metric for assessing accuracy? Positional evaluation confirming your own preference for a position, with the tactics filtered out by the search? Given the fact that you have developed the program yourself, it could be impossible for any other program to be more accurate, simply because the definition of accuracy is your own! (this is not to be cynical, it just shows the difficulty of defining any objectivity for taste).

I guess my metric would be which program more often chooses the same move that a majority of the world's top ten players would choose if playing correspondence chess without computer assistance but with plenty of time and incentive. Of course there's no way to determine this in actual practice. We can try to match move choices of top correspondence players but they are not truly "top" players, and they rely heavily on whatever engine was strongest when the game was played.

[Robert Flesher]

To date I have only seen evaluations based in how much more or less elo points than Houdini -or any other- Komodo have.
Forgive me, but that is irrelevant. To be 3000 Elo or 3012 is the very same thing.
In fact there are dozens of engine above 2400, more than enough to kill almost anybody here. Sheers strength does not matter, AT LEAST you are a freak that prefer play engines that to play them or like those guys that are worried of the quality of the sound of his music gear instead of what music they listen.
So the question is: how much good for playing HUMANS Komodo is? How much more interesting is his style of playing? how much or less fun we can get from it? How more or less lessons we can extract of his playing ways? How good time we get with it?

For that I am still waiting for some human being here telling us that because HE played the thing.


best
Fern, Human.

Heya Fern, your question it based on what you or any particular "human" intends to achieve. For example, some people study opening theory, play many games, and seek to improve their game through analysis and dedicated study. For those people super strong engines, or for that matter many of them are great tools. Of course this is not the only method, but clearly popular as shown by the top players in the world openly
stating they like to use Houdini, Rybka, Stockfish, Komodo, etc, etc in their analysis.

However, for those who just wish to play for the fun of it and do not seek to improve their game in the above mention methods, then perhaps any 2400+ engine will do.

[fern]

Of course, Robert, I consider also those other ways to play an engine, not only to get fun, as me. But in all cases, fun or learning, the matter is not anymore sheer strength because that is already available. Naturally is not the same to use a 1890 Elo or a 2900 Elo to learn, but it is the same thing if the engines are 3000 or 3012. In this last case, what I say is that the core of the matter is style.

Fern
learning from chess challenger 7 regards

[lech]

For me personally the only thing that matters about an engine is the quality of its analysis. Any modern engine will crush almost any human player, there's no need for new engines just to play against. My biggest motivation for Komodo was to create a superior analysis tool. I believe that Komodo produces the most accurate analysis of any one core program now (given a minute or so to think), but I don't expect anyone to take my word for it, as I am hardly neutral. That's why we focus on the Elo rating, it's the only generally accepted measure of how well a program plays chess, which people take as the same as how well it will analyze positions. I would love to see a better objective measure of this.

A wonderful news!
I see that my idea to make a human-engines just found a feedback!
I hope Komodo 4 solves this extremaly example.
[d]r5k1/1qrnbppp/3p4/3Pp1PP/pp2N1R1/3QBP2/PPP5/1K1R4 w - - 0 1
1. g6!! fg6 2 hg6 h6 3 Bxh6 or 2...h5 3 Sg3
Sting SF 2.0 can do it (PIV 2 threads )

Code: Select all

1	00:02	       1.109	470	-0,04	f3f4
   2	00:02	       1.593	670	 0,00	f3f4 Ra8c8
   3	00:02	       3.205	1.341	-0,56	f3f4 b4b3 a2xb3 a4xb3 Qd3xb3
   3	00:02	       4.012	1.667	-0,08	g5g6 h7xg6 h5xg6
   4	00:02	       4.683	1.946	-0,20	g5g6 f7xg6 h5xg6 h7h5
   5	00:02	       5.637	2.327	+0,04	g5g6 f7xg6 h5xg6 h7h5 Rg4g2
   6	00:02	       6.829	2.783	+0,12	g5g6 f7xg6 h5xg6 h7h5 Rg4g2 Ra8c8
   7-	00:02	      14.556	5.751	-0,08	g5g6 f7xg6 h5xg6 h7h5 Rg4g2 h5h4 Ne4g5 Be7xg5 Be3xg5 h4h3
   7-	00:02	      18.023	7.034	-0,28	g5g6 f7xg6 h5xg6 h7h5 Rg4g2 h5h4 Ne4g5 Be7xg5 Be3xg5 h4h3
   7	00:02	      24.991	9.520	-0,60	g5g6 f7xg6 h5xg6 h7h5 Rg4g2 h5h4 Ne4g5 Be7xg5 Rg2xg5
   7	00:02	      32.749	12.330	-0,24	h5h6 g7g6 f3f4 b4b3 a2xb3 a4xb3 Qd3xb3
   8	00:02	      43.496	15.997	-0,16	h5h6 g7g6 f3f4 b4b3 c2xb3 a4xb3 Qd3xb3 Qb7xb3 a2xb3 f7f5 g5xf6/ep Nd7xf6 Ne4xf6+ Be7xf6 f4xe5 Bf6xe5
   9-	00:02	      63.491	22.699	-0,44	h5h6 g7g6 f3f4 b4b3 c2xb3 a4xb3 Qd3xb3 Qb7xb3 a2xb3 Ra8b8 Kb1a2 Rc7b7
   9	00:02	      87.944	29.943	 0,00	h5h6 g7g6 f3f4 b4b3 c2xb3 a4xb3 Qd3xb3 Qb7xb3 a2xb3 Ra8b8 Kb1a2 Rb8a8+ Ka2b1
  10-	00:02	      91.189	30.713	-0,20	h5h6 g7g6 f3f4 b4b3 c2xb3 a4xb3 Qd3xb3 Qb7xb3 a2xb3 Ra8b8 Kb1a2 f7f5 g5xf6/ep Nd7xf6 Ne4xf6+ Be7xf6 f4xe5 Bf6xe5
  10-	00:03	     103.320	34.087	-0,40	h5h6 g7g6 f3f4 b4b3 c2xb3 a4xb3 Qd3xb3 Qb7xb3 a2xb3 Ra8b8 Kb1a2 Rc7b7
  10	00:03	     112.706	36.427	-0,24	h5h6 g7g6 f3f4 b4b3 c2xb3 a4xb3 Qd3xb3 Qb7xb3 a2xb3 Ra8b8 Kb1a2 Rc7b7 f4xe5 Rb8a8+ Ka2b1 Nd7xe5
  11-	00:03	     140.824	43.747	-0,56	h5h6 g7g6 f3f4 b4b3 c2xb3 Ra8b8 Kb1a1 a4xb3 a2a4 Rb8a8
  11	00:03	     199.490	57.506	-0,56	h5h6 g7g6 f3f4 b4b3 c2xb3 Ra8b8 Rd1c1 a4xb3 Qd3xb3 Qb7xb3 a2xb3 Rc7xc1+ Be3xc1 Rb8xb3 Kb1c2 Rb3b4
  12	00:04	     434.465	100.757	-0,28	h5h6 g7g6 Rg4g2 b4b3 c2xb3 a4xb3 Qd3xb3 Qb7xb3 a2xb3 Rc7b7 Rg2c2 Rb7xb3 Rc2c7 Rb3xe3 Rc7xd7 Kg8f8 Rd1f1
  13	00:05	     688.356	130.741	-0,44	h5h6 g7g6 Rg4g2 Ra8b8 Rd1c1 b4b3 a2xb3 a4xb3 c2xb3 Rc7xc1+ Kb1xc1 Qb7xb3 Qd3xb3 Rb8xb3 Ne4c3 f7f5 Kc1c2
  14	00:08	   1.646.617	194.796	-0,56	h5h6 g7g6 Rg4g2 Ra8b8 f3f4 b4b3 a2xb3 a4xb3 c2xb3 Rb8c8 Rd1c1 Rc7xc1+ Be3xc1 Nd7c5 Ne4xc5 Rc8xc5 f4xe5 d6xe5
  14+	00:09	   1.827.324	202.698	-0,12	Ne4g3 g7g6 h5xg6 f7xg6 Rg4h4 b4b3 c2xb3 a4a3 Rd1h1
  14	00:09	   1.918.605	206.368	-0,16	Ne4g3 g7g6 h5xg6 f7xg6 Rg4h4 b4b3 c2xb3 a4xb3 Qd3xb3 Qb7a6 a2a4 h7h5 Qb3b5 Qa6xb5 a4xb5 Kg8f7 f3f4 Rc7c4
  15-	00:12	   3.011.903	241.861	-0,52	Ne4g3 g7g6 h5xg6 f7xg6 Rg4h4 b4b3 c2xb3 a4xb3 Qd3xb3 Qb7xb3 a2xb3 h7h5 Rh4a4 Ra8f8
  15+	00:23	   6.516.732	277.130	+0,20	Ne4g3 g7g6 h5xg6 f7xg6 Rg4c4 Rc7xc4 Qd3xc4 b4b3 c2xb3 a4xb3 Qc4xb3 Qb7xb3 a2xb3 Ra8f8 Rd1f1 Nd7c5 b3b4
  15	00:24	   6.782.277	278.784	-0,20	Ne4g3 g7g6 h5xg6 f7xg6 Rg4c4 Rc7xc4 Qd3xc4 Ra8c8 Qc4g4 Nd7f8 f3f4 b4b3 a2xb3 a4xb3 c2c3 Rc8a8 f4xe5 d6xe5
  16	00:29	   8.351.268	286.129	-0,24	Ne4g3 a4a3 Ng3f5 Be7f8 b2b3 g7g6 h5xg6 f7xg6 Nf5g3 Ra8c8 Rd1d2 Qb7a8 Qd3b5 Rc7b7 Qb5e2 Bf8e7 Ng3e4
  17+	00:32	   9.346.776	287.867	-0,16	Ne4g3 a4a3 Ng3f5 Be7f8 b2b3 g7g6 h5xg6 f7xg6 Nf5g3 Ra8c8 Rd1d2 Qb7a8 Qd3b5 Rc7b7 Qb5e2 Bf8e7 f3f4
  17	00:38	  11.288.718	291.442	-0,16	Ne4g3 a4a3 Ng3f5 Be7f8 b2b3 g7g6 h5xg6 f7xg6 Nf5g3 Ra8c8 Rd1d2 Rc7c3 Qd3e2 Bf8e7 Ng3e4 Rc3c7 f3f4 Qb7a8
  18+	00:48	  14.368.158	297.594	-0,04	Ne4g3 a4a3 Ng3f5 Be7f8 Nf5h6+ g7xh6 g5xh6+ Kg8h8 Qd3f5 b4b3 c2xb3
  18-	00:55	  17.113.765	307.751	-0,28	Ne4g3 b4b3 c2xb3 a4xb3 Qd3xb3 Qb7a6 Ng3f5 Be7f8 a2a4 Qa6e2 g5g6 f7xg6 h5xg6 h7h5
  18-	01:15	  24.110.383	319.211	-0,40	Ne4g3 Rc7c3 Qd3e2 Ra8c8 Ng3f5 Rc3xc2 Nf5xe7+ Kg8f8
  18+	02:33	  49.522.920	321.740	+0,08	Ne4g3 Rc7c3 Qd3d2 Ra8c8 Ng3f5 Kg8f8 Rd1c1 a4a3 b2b3 g7g6 h5xg6 f7xg6 Nf5xe7 Kf8xe7 Rg4h4
  18	02:37	  50.473.779	321.329	 0,00	Ne4g3 Rc7c3 Qd3d2 Ra8c8 Ng3f5 Be7f8 Nf5h6+ Kg8h8 Nh6xf7+ Kh8g8 g5g6 Nd7f6 g6xh7+ Kg8xh7 Rg4g1 Kh7g8 Nf7h6+ Kg8h8 Nh6f7+ Kh8g8
  19	02:47	  53.846.879	322.075	-0,01	Ne4g3 Rc7c3 Qd3d2 Ra8c8 Ng3f5 Be7f8 Nf5h6+ Kg8h8 Nh6xf7+ Kh8g8 g5g6 Nd7f6 g6xh7+ Kg8xh7 Rg4g1 Kh7g8 Nf7h6+ Kg8h8 Nh6f7+ Qb7xf7 b2xc3 b4xc3 Qd2f2 Nf6xd5 Kb1a1 Nd5b4
  20-	03:07	  61.282.833	326.705	-0,13	Ne4g3 Rc7c3 Qd3d2 Ra8c8 Ng3f5 Be7f8 Nf5h6+ Kg8h8 Nh6xf7+ Kh8g8 g5g6 h7h6 Rd1c1 Nd7f6 Rg4g3 a4a3
  20	03:40	  71.969.458	326.856	 0,00	Ne4g3 Rc7c3 Qd3d2 Ra8c8 Rd1c1 a4a3 b2b3 Kg8h8 Ng3f5 Be7f8 f3f4 Rc8e8 Rc1h1 Re8c8 Rh1c1
  21	04:18	  85.585.681	330.946	 0,00	Ne4g3 Rc7c3 Qd3d2 Ra8c8 Rd1c1 a4a3 b2b3 Kg8h8 Ng3f5 Be7f8 f3f4 Rc8e8 Rc1h1 Re8c8 Rh1c1
  22	05:02	 101.792.104	336.242	 0,00	Ne4g3 Rc7c3 Qd3d2 Ra8c8 Rd1c1 a4a3 b2b3 Kg8h8 Ng3f5 Be7f8 f3f4 Rc8e8 Rc1h1 Re8c8 Rh1c1
  23	06:12	 127.672.567	343.205	 0,00	Ne4g3 Rc7c3 Qd3d2 Ra8c8 Rd1c1 a4a3 b2b3 Kg8h8 Ng3f5 Be7f8 f3f4 Rc8e8 Rc1h1 Re8c8 Rh1c1
  24	07:46	 163.053.550	349.654	 0,00	Ne4g3 Rc7c3 Qd3d2 Ra8c8 Rd1c1 a4a3 b2b3 Kg8h8 Ng3f5 Be7f8 f3f4 Rc8e8 Rc1h1 Re8c8 Rh1c1
  25	12:18	 263.710.941	357.143	 0,00	Ne4g3 Rc7c3 Qd3d2 Ra8c8 Rd1c1 Be7f8 Rg4h4 Bf8e7 Ng3f5 Be7f8 Nf5g3
  26	17:42	 391.813.627	368.798	 0,00	Ne4g3 Rc7c3 Qd3d2 Ra8c8 Rd1c1 Be7f8 Rg4h4 Bf8e7 Ng3f5 Be7f8 Nf5g3
  27	53:13	1.230.918.802	385.399	 0,00	Ne4g3 Be7f8 Ng3e4 Bf8e7
  28	1:06:53	1.566.577.070	390.357	 0,00	Ne4g3 Be7f8 Ng3e4 Bf8e7
  28+	1:37:55	2.398.574 904	408.258	+0,08	g5g6 f7xg6 h5xg6 h7h5 Ne4g3 h5xg4 Qd3f5 Be7h4 Qf5e6+ Kg8h8 Qe6xg4 Bh4xg3 Rd1h1+ Kh8g8 Qg4e6+
  28+	2:02:44	3.069.752 922	416.815	+0,16	g5g6 f7xg6 h5xg6 h7h6 Be3xh6 g7xh6 Qd3e3 Be7g5
  28	2:12:30	3.330.297 947	418.890	 0,00	g5g6 f7xg6 h5xg6 h7h6 Be3xh6 g7xh6 Qd3e3 Be7g5 Rg4xg5 h6xg5 Qe3xg5 b4b3 c2xb3 a4xb3

i7 is ~28 x faster

[Robert Flesher]

Of course, Robert, I consider also those other ways to play an engine, not only to get fun, as me. But in all cases, fun or learning, the matter is not anymore sheer strength because that is already available. Naturally is not the same to use a 1890 Elo or a 2900 Elo to learn, but it is the same thing if the engines are 3000 or 3012. In this last case, what I say is that the core of the matter is style.

Fern
learning from chess challenger 7 regards

Gotcha!

[IGarcia]

Fern, Do you have a game against this monsters to post? One of the kind you can learn (no a simple blunder tactical kill)

[A Distel]

Where can i download this "Sting SF 2.0" Engine" ?

[fern]

I do not know. Maybe there is one or several, but I do not look at them after my defeats and many times also I do not save them. Once a game is over, the fun is also over and I forget the all thing-. I am NOT a serious player...

Fern

[lech]

Where can i download this "Sting SF 2.0" Engine" ?

Sting SF 2.0 is not ready yet.
I want to do it faster, stronger and wiser. It is not easy.

[Master Om]

Damn Marek.....you are spot on!!