Look who's number 1 at 40/2 on CEGT!

[lkaufman]

I would just say that although Watanabe is unquestionably a genius at shogi, I don't know if he is knowledgeable about computer shogi, nor do we know his opinion on the matter after the Yonenaga defeat. Moriuchi, also among the main title contenders, told me about two years ago that he considered even the best program to be quite weak by the standards of title contenders. But these great players are unlikely to know much about the rate of improvement of both software and hardware or how progress in this field will occur. I can't imagine that it is credible that a program could be stronger than Yonenaga now but not stronger than Habu or Watanabe two years from now. I've told the pros that by about 2020 the best program will be able to give the top human pro a lance handicap with success. They don't believe me, but we shall see. I'm sure human GMs in chess would have also ridiculed the notion of programs giving them pawn odds, but this has happened.

[Don]

I would just say that although Watanabe is unquestionably a genius at shogi, I don't know if he is knowledgeable about computer shogi, nor do we know his opinion on the matter after the Yonenaga defeat. Moriuchi, also among the main title contenders, told me about two years ago that he considered even the best program to be quite weak by the standards of title contenders. But these great players are unlikely to know much about the rate of improvement of both software and hardware or how progress in this field will occur. I can't imagine that it is credible that a program could be stronger than Yonenaga now but not stronger than Habu or Watanabe two years from now. I've told the pros that by about 2020 the best program will be able to give the top human pro a lance handicap with success. They don't believe me, but we shall see. I'm sure human GMs in chess would have also ridiculed the notion of programs giving them pawn odds, but this has happened.

Strong players very often underestimate computers anyway. They see a game and they see the weak moves or the computer style but don't always consider the compensating strengths.

[S.Taylor]

Is Komodo L.Kaufmans own program?
Excuse my ignorance.

At any rate, does it look like Houdini will not be top, in the near future?

Isn't Houdini nearing the end of it's waiting period for us to see the next upgrade?

If so, then an additional 30 elo may keep it on top for a long time? or not?

[lkaufman]

Don Dailey and I are partners on Komodo. Don is the main programmer; until this new year I could only propose changes for Don to implement or make changes to some parameters. We both do testing. Now I've actually learned enough C programming to make simple changes to the code myself, which I've been doing while Don works on MP. Of course Don reviews any changes I make and if possible codes them better for the official Komodo.

I predict that our next release (which should be quite soon) will equal or surpass Houdini at the longer time limits (like 40/40min. or perhaps at 40/20min), though we'll only be second at blitz time limits. As to whether Houdini will release a stronger version, I have no inside information on that, I can only point out that the last release, despite a long wait, failed to prove to be any improvement at all at the longer time limits used by the testers.

[Robert Flesher]

Well if an engine can see mates this deep (kidding) , then no wonder it is #1.

Great job guys!


Guest1833339 - Guest1833412, Friendly Game, 3m + 0s 2012
8/1K6/1P6/1q6/8/7k/8/8 w - - 0 1

Analysis by Komodo64 SSE Version 4:

50.Kc7 Kg3 51.b7 Qc5+ 52.Kd8 Qb6+ 53.Kd7 Qxb7+ 54.Ke6 Qc6+ 55.Kf5
-+ (-7.33) Depth: 6 00:00:00 4kN
50.Kc7 Kg3 51.b7 Qc5+ 52.Kd8 Qb6+ 53.Kd7 Qxb7+ 54.Ke6 Qc6+ 55.Kf5
-+ (-7.33) Depth: 6 00:00:00 4kN
50.Kc7 Kg3 51.b7 Qc5+ 52.Kd8 Qb6+ 53.Kd7 Qxb7+ 54.Ke6 Qc6+ 55.Kf5
-+ (-7.33) Depth: 7 00:00:00 6kN
50.Kc7 Kg3 51.b7 Qc5+ 52.Kd8 Qb6+ 53.Kd7 Qxb7+ 54.Ke6 Qc6+ 55.Kf5
-+ (-7.33) Depth: 7 00:00:00 6kN
50.Kc7 Qc5+ 51.Kd7 Qxb6 52.Ke7 Kg4 53.Kf8 Qf6+ 54.Kg8 Qf5
-+ (-7.68) Depth: 8 00:00:00 11kN
50.Kc7 Qc5+ 51.Kd7 Qxb6 52.Ke7 Kg4 53.Kf8 Qf6+ 54.Kg8 Qf5
-+ (-7.68) Depth: 8 00:00:00 11kN
50.Kc7 Qc5+
-+ (-7.85) Depth: 9 00:00:00 12kN
50.Kc7 Qc5+
-+ (-8.35) Depth: 9 00:00:00 12kN
50.Kc7 Qc5+
-+ (-9.10) Depth: 9 00:00:00 12kN
50.Kc7 Qc5+
-+ (-10.23) Depth: 9 00:00:00 13kN
50.Kc7 Qc5+
-+ (-11.93) Depth: 9 00:00:00 13kN
50.Kc7 Qc5+
-+ (-14.48) Depth: 9 00:00:00 13kN
50.Kc7 Qc5+
-+ (-18.31) Depth: 9 00:00:00 13kN
50.Kc7 Qc5+
-+ (-24.06) Depth: 9 00:00:00 13kN
50.Kc7 Qc5+
-+ (-32.69) Depth: 9 00:00:00 13kN
50.Kc7 Qc5+
-+ (-45.64) Depth: 9 00:00:00 13kN
50.Kc7 Qc5+
-+ (-65.07) Depth: 9 00:00:00 13kN
50.Kc7 Qc5+
-+ (-94.22) Depth: 9 00:00:00 13kN
50.Kc7 Qc5+
-+ (-137.95) Depth: 9 00:00:00 13kN
50.Kc7 Qc5+
-+ (-203.55) Depth: 9 00:00:00 13kN
50.Kc7 Qc5+ 51.Kd7 Kg3 52.Ke8 Kg4 53.Kd7 Qd4+ 54.Kc6 Qc4+ 55.Kd6 Qb5
-+ (-#1332) Depth: 9 00:00:00 17kN
50.Kc7 Qc5+ 51.Kd7 Kg3 52.Ke8 Kg4 53.Kd7 Qd4+ 54.Kc6 Qc4+ 55.Kd6 Qb5
-+ (-#1332) Depth: 9 00:00:00 17kN
50.Kc7 Qc5+ 51.Kb7 Kg3 52.Ka8 Qc8+ 53.Ka7 Qc6 54.Ka6 Qc4+ 55.Kb7 Kf3 56.Ka7 Qa4+ 57.Kb8
+- (#1716) Depth: 10 00:00:00 27kN
50.Kc7 Qc5+ 51.Kb7 Kg3 52.Ka8 Qc8+ 53.Ka7 Qc6 54.Ka6 Qc4+ 55.Kb7 Kf3 56.Ka7 Qa4+ 57.Kb8
+- (#1716) Depth: 10 00:00:00 27kN
50.Kc7
-+ (-162.42) Depth: 11 00:00:00 49kN
50.Kc7 Kg3 51.b7 Qc5+ 52.Kd8 Qd6+ 53.Ke8 Qb8+ 54.Ke7 Kg4 55.Kd7 Qxb7+ 56.Ke6 Qe4+ 57.Kd6 Qd4+ 58.Kc7 Qc5+ 59.Kb7 Qb5+ 60.Ka8
-+ (-7.72) Depth: 11 00:00:00 114kN
50.Kc7 Kg3 51.b7 Qc5+ 52.Kd8 Qd6+ 53.Ke8 Qb8+ 54.Ke7 Kg4 55.Kd7 Qxb7+ 56.Ke6 Qe4+ 57.Kd6 Qd4+ 58.Kc7 Qc5+ 59.Kb7 Qb5+ 60.Ka8
-+ (-7.72) Depth: 11 00:00:00 114kN
50.Kc7 Kg3 51.b7 Qc5+ 52.Kd8 Qd6+ 53.Ke8 Qb8+ 54.Ke7 Kh3 55.Kd7 Qxb7+ 56.Ke6 Kg3 57.Kf6 Qd5 58.Kg7 Kg4
-+ (-7.87) Depth: 12 00:00:00 162kN
50.Kc7 Kg3 51.b7 Qc5+ 52.Kd8 Qd6+ 53.Ke8 Qb8+ 54.Ke7 Kh3 55.Kd7 Qxb7+ 56.Ke6 Kg3 57.Kf6 Qd5 58.Kg7 Kg4
-+ (-7.87) Depth: 12 00:00:00 163kN
50.Kc7 Qc5+
-+ (-162.66) Depth: 13 00:00:00 317kN
50.Kc7 Qc5+ 51.Kb7 Qc4 52.Ka7 Kg4 53.Kb8 Qc5 54.Ka7 Qa5+ 55.Kb7 Qb5 56.Kc7 Qc5+ 57.Kd7 Qd5+ 58.Ke7 Qc6 59.Kf8 Qf6+ 60.Kg8
+- (#1786) Depth: 13 00:00:00 438kN
50.Kc7 Qc5+ 51.Kb7 Qc4 52.Ka7 Kg4 53.Kb8 Qc5 54.Ka7 Qa5+ 55.Kb7 Qb5 56.Kc7 Qc5+ 57.Kd7 Qd5+ 58.Ke7 Qc6 59.Kf8 Qf6+ 60.Kg8
+- (#1786) Depth: 13 00:00:00 438kN
50.Kc7 Qc5+ 51.Kb7 Qc4 52.Ka7 Kg4 53.Kb8 Qc5 54.Ka7 Qa5+ 55.Kb7 Qb5 56.Kc7 Qc5+ 57.Kd7 Qd5+ 58.Ke7 Qc6 59.Kf8 Qf6+ 60.Kg8 Qe5
+- (#1786) Depth: 14 00:00:00 570kN
50.Kc7 Qc5+ 51.Kb7 Qc4 52.Ka7 Kg4 53.Kb8 Qc5 54.Ka7 Qa5+ 55.Kb7 Qb5 56.Kc7 Qc5+ 57.Kd7 Qd5+ 58.Ke7 Qc6 59.Kf8 Qf6+ 60.Kg8 Qe5
+- (#1786) Depth: 14 00:00:00 572kN
50.Kc7 Qc5+ 51.Kb7 Kg3 52.Ka6 Qb4 53.Kb7 Qc4 54.Kb8 Qf4+ 55.Ka7 Qf5 56.Ka8 Kg4 57.Ka7 Qd5 58.Kb8 Qd7 59.Ka8
+- (#1786) Depth: 15 00:00:00 1254kN
50.Kc7 Qc5+ 51.Kb7 Kg3 52.Ka6 Qb4 53.Kb7 Qc4 54.Kb8 Qf4+ 55.Ka7 Qf5 56.Ka8 Kg4 57.Ka7 Qd5 58.Kb8 Qd7 59.Ka8
+- (#1786) Depth: 15 00:00:00 1259kN
50.Kc7
+- (#1770) Depth: 16 00:00:01 2125kN
50.Kc7
+- (#1723) Depth: 16 00:00:01 2911kN
50.Kc7
+- (#1652) Depth: 16 00:00:01 3650kN
50.Kc7 Qc5+ 51.Kb7 Kg3 52.Ka7 Kf4 53.Kb7 Qc4 54.Ka7 Qd5 55.Kb8 Qf7 56.Kc8 Qf5+ 57.Kb7 Qd7+ 58.Kb8 Qe7 59.Kc8 Kg3 60.Kb8 Qd8+ 61.Kb7
+- (#1546) Depth: 16 00:00:02 4547kN
50.Kc7 Qc5+ 51.Kb7 Kg3 52.Ka7 Kf4 53.Kb7 Qc4 54.Ka7 Qd5 55.Kb8 Qf7 56.Kc8 Qf5+ 57.Kb7 Qd7+ 58.Kb8 Qe7 59.Kc8 Kg3 60.Kb8 Qd8+ 61.Kb7
+- (#1546) Depth: 16 00:00:02 4640kN
50.Kc7 Qc5+
+- (#1666) Depth: 17 00:00:02 5095kN
50.Kc7 Qc5+ 51.Kb7 Qd6 52.Ka7 Qa3+ 53.Kb7 Qc5 54.Ka7 Kg4 55.Ka6 Qc4+ 56.Kb7 Qb5 57.Ka7 Qd5 58.Kb8 Qd6+ 59.Ka7 Kg3 60.Ka8 Qd5+ 61.Kb8 Kf4 62.Kc7
+- (#1796) Depth: 17 00:00:04 8956kN
50.Kc7 Qc5+ 51.Kb7 Qd6 52.Ka7 Qa3+ 53.Kb7 Qc5 54.Ka7 Kg4 55.Ka6 Qc4+ 56.Kb7 Qb5 57.Ka7 Qd5 58.Kb8 Qd6+ 59.Ka7 Kg3 60.Ka8 Qd5+ 61.Kb8 Kf4 62.Kc7
+- (#1796) Depth: 17 00:00:04 8957kN
50.Kc7
+- (#1551) Depth: 18 00:00:08 16492kN
50.Kc7 Qc5+
+- (#1552) Depth: 18 00:00:09 19385kN

(, Microsoft 05.02.2012)



[d]8/1K6/1P6/1q6/8/7k/8/8 w - - 0 50

[tomgdrums]

Don Dailey and I are partners on Komodo. Don is the main programmer; until this new year I could only propose changes for Don to implement or make changes to some parameters. We both do testing. Now I've actually learned enough C programming to make simple changes to the code myself, which I've been doing while Don works on MP. Of course Don reviews any changes I make and if possible codes them better for the official Komodo.

I predict that our next release (which should be quite soon) will equal or surpass Houdini at the longer time limits (like 40/40min. or perhaps at 40/20min), though we'll only be second at blitz time limits. As to whether Houdini will release a stronger version, I have no inside information on that, I can only point out that the last release, despite a long wait, failed to prove to be any improvement at all at the longer time limits used by the testers.

Will the next release be an update or a new release?

What do you of Junior at long time controls. Junior seems to be holding its own with Houdini 1.5a at a long classical time control.

[S.Taylor]

So i suppose some people may be thinking that he (Houdart/Houdini) hit his limit, with outside help perhaps.

[lkaufman]

Will the next release be an update or a new release?

The plan is to release the new version as an update for those who bought SP, while of course the MP will be a new release as we haven't had MP so far. We would only make the new version a new release if the elo gain is sufficiently large to justify this, which is unlikely (but not impossible) in the anticipated short time until MP is ready.

What do you of Junior at long time controls. Junior seems to be holding its own with Houdini 1.5a at a long classical time control.

I hadn't seen that result. Was it based on enough games to be reliable? (Edit: I just saw that it is based on 8 games, with Junior down one. Let me know when a couple hundred have been played!) It would seem to me unlikely that better scaling could offset the huge elo gap between Junior and Houdini in blitz. I think better scaling can offset fifty elo or so at blitz, but not much more than that.