60 games Komodo 5 against Top4 at 120m+3s

Dr.Wael Deeb · Post by **Dr.Wael Deeb** » Thu Jul 26, 2012 2:40 pm

Houdini wrote:After some of the worst, also some of the very best.
Game 47, King's Indian Defense.
The key position occurs after White's move 29.Qd2.

[D]1rq5/2p1nrk1/2Pp1np1/B2Ppb2/R3p3/2N3P1/3QBP2/4R1K1 b - - 2 29

Houdini uses nearly 12 minutes to play the sacrifice 29...Qh8!! and after 30.Bxc7 the truly outstanding 30..Kg8!! with a winning attack (Houdini eval -3.45).

Code: Select all

&#91;Event "Komodo5-Houdini"&#93;
&#91;Site "?"&#93;
&#91;Date "2012.07.26"&#93;
&#91;Round "47.1"&#93;
&#91;White "Komodo 5 64-bit"&#93;
&#91;Black "Houdini 2.0c Pro x64"&#93;
&#91;Result "0-1"&#93;
&#91;ECO "E97"&#93;
&#91;Annotator "0.19;-0.02"&#93;
&#91;PlyCount "74"&#93;
&#91;EventDate "2012.07.19"&#93;
&#91;EventType "tourn"&#93;

&#123;AMD Phenom&#40;tm&#41; II X6 1045T Processor 2999 MHz  W=21.4 plies; 1.055kN/s  B=21.
6 plies; 1.833kN/s&#125; 1. d4 Nf6 2. c4 g6 3. Nc3 Bg7 4. e4 d6 5. Nf3 O-O 6. Be2 e5
7. O-O Nc6 8. d5 Ne7 9. b4 a5 10. Ba3 b6 11. bxa5 Nh5 12. g3 &#123;&#91;%eval 19,22&#93; 
&#91;%emt 0&#58;01&#58;48&#93;&#125; Rxa5 &#123;&#91;%eval -2,21&#93; &#91;%emt 0&#58;02&#58;43&#93; &#40;f5&#41;&#125; 13. Bb4 &#123;&#91;%eval 2,20&#93;
&#91;%emt 0&#58;00&#58;32&#93;&#125; Ra6 &#123;&#91;%eval 3,23&#93; &#91;%emt 0&#58;02&#58;49&#93;&#125; 14. a4 &#123;&#91;%eval -1,23&#93; &#91;%emt
0&#58;13&#58;54&#93;&#125; f5 &#123;&#91;%eval 0,24&#93; &#91;%emt 0&#58;00&#58;00&#93;&#125; 15. a5 &#123;&#91;%eval 3,23&#93; &#91;%emt 0&#58;03&#58;42&#93;
&#40;Ng5&#41;&#125; Nf6 &#123;&#91;%eval 4,22&#93; &#91;%emt 0&#58;04&#58;59&#93;&#125; 16. axb6 &#123;&#91;%eval 11,25&#93; &#91;%emt 0&#58;06&#58;51&#93;
&#125; Rxb6 &#123;&#91;%eval 2,24&#93; &#91;%emt 0&#58;00&#58;00&#93;&#125; 17. Ba5 &#123;&#91;%eval 12,22&#93; &#91;%emt 0&#58;02&#58;15&#93;&#125; Rb7
&#123;&#91;%eval -4,23&#93; &#91;%emt 0&#58;04&#58;27&#93; &#40;fxe4&#41;&#125; 18. Nd2 &#123;&#91;%eval 5,23&#93; &#91;%emt 0&#58;06&#58;37&#93;&#125; Bh6
&#123;&#91;%eval 0,23&#93; &#91;%emt 0&#58;00&#58;40&#93; &#40;Nxe4&#41;&#125; 19. c5 &#123;&#91;%eval 31,21&#93; &#91;%emt 0&#58;01&#58;38&#93;&#125; Bxd2
&#123;&#91;%eval 8,22&#93; &#91;%emt 0&#58;01&#58;16&#93;&#125; 20. Qxd2 &#123;&#91;%eval 40,21&#93; &#91;%emt 0&#58;00&#58;07&#93;&#125; fxe4 &#123;
&#91;%eval 4,22&#93; &#91;%emt 0&#58;03&#58;01&#93;&#125; 21. c6 &#123;&#91;%eval 27,22&#93; &#91;%emt 0&#58;00&#58;49&#93; &#40;cxd6&#41;&#125; Rb8 &#123;
&#91;%eval -5,22&#93; &#91;%emt 0&#58;02&#58;18&#93;&#125; 22. Ra4 &#123;&#91;%eval 23,22&#93; &#91;%emt 0&#58;01&#58;25&#93;&#125; Bf5 &#123;
&#91;%eval -9,23&#93; &#91;%emt 0&#58;06&#58;38&#93; &#40;Kg7&#41;&#125; 23. Rfa1 &#123;&#91;%eval 17,21&#93; &#91;%emt 0&#58;04&#58;27&#93;&#125; Rf7
&#123;&#91;%eval -5,23&#93; &#91;%emt 0&#58;00&#58;00&#93;&#125; 24. Qa2 &#123;&#91;%eval 15,21&#93; &#91;%emt 0&#58;03&#58;45&#93; &#40;Rd1&#41;&#125; Kg7
&#123;&#91;%eval -4,22&#93; &#91;%emt 0&#58;03&#58;11&#93;&#125; 25. Re1 &#123;&#91;%eval 15,22&#93; &#91;%emt 0&#58;00&#58;26&#93; &#40;Rd1&#41;&#125; h5
&#123;&#91;%eval -17,22&#93; &#91;%emt 0&#58;01&#58;49&#93; &#40;g5&#41;&#125; 26. Rd1 &#123;&#91;%eval -13,21&#93; &#91;%emt 0&#58;05&#58;30&#93; 
&#40;h4&#41;&#125; h4 &#123;&#91;%eval -34,21&#93; &#91;%emt 0&#58;04&#58;10&#93; &#40;Ng4&#41;&#125; 27. Re1 &#123;&#91;%eval -14,20&#93; &#91;%emt 0&#58;
02&#58;40&#93;&#125; hxg3 &#123;&#91;%eval -56,20&#93; &#91;%emt 0&#58;00&#58;40&#93;&#125; 28. hxg3 &#123;&#91;%eval -64,22&#93; &#91;%emt 0&#58;
06&#58;06&#93; &#40;fxg3&#41;&#125; Qc8 &#123;&#91;%eval -57,20&#93; &#91;%emt 0&#58;02&#58;09&#93;&#125; 29. Qd2 &#123;&#91;%eval -69,21&#93; 
&#91;%emt 0&#58;01&#58;32&#93;&#125; Qh8 &#123;&#91;%eval -147,22&#93; &#91;%emt 0&#58;11&#58;55&#93; &#40;Bg4&#41;&#125; 30. Bxc7 &#123;&#91;%eval
-59,20&#93; &#91;%emt 0&#58;01&#58;02&#93;&#125; Kg8 &#123;&#91;%eval -345,18&#93; &#91;%emt 0&#58;00&#58;00&#93; &#40;e3&#41;&#125; 31. Bf1 &#123;
&#91;%eval -204,20&#93; &#91;%emt 0&#58;04&#58;02&#93;&#125; Ng4 &#123;&#91;%eval -422,23&#93; &#91;%emt 0&#58;00&#58;00&#93;&#125; 32. f3 &#123;
&#91;%eval -228,20&#93; &#91;%emt 0&#58;00&#58;49&#93; &#40;f4&#41;&#125; exf3 &#123;&#91;%eval -427,19&#93; &#91;%emt 0&#58;00&#58;26&#93;&#125; 33.
Rxg4 &#123;&#91;%eval -218,23&#93; &#91;%emt 0&#58;02&#58;59&#93;&#125; f2+ &#123;&#91;%eval -473,23&#93; &#91;%emt 0&#58;08&#58;26&#93; 
&#40;Bxg4&#41;&#125; 34. Qxf2 &#123;&#91;%eval -190,19&#93; &#91;%emt 0&#58;00&#58;55&#93;&#125; Bxg4 &#123;&#91;%eval -618,19&#93; &#91;%emt
0&#58;00&#58;00&#93;&#125; 35. Qd2 &#123;&#91;%eval -240,21&#93; &#91;%emt 0&#58;04&#58;23&#93; &#40;Qh2&#41;&#125; Rbf8 &#123;&#91;%eval -686,19&#93;
&#91;%emt 0&#58;00&#58;41&#93; &#40;Bf3&#41;&#125; 36. Bg2 &#123;&#91;%eval -512,20&#93; &#91;%emt 0&#58;06&#58;54&#93;&#125; Qh5 &#123;&#91;%eval
-783,22&#93; &#91;%emt 0&#58;00&#58;00&#93;&#125; 37. Bxd6 &#123;&#91;%eval -567,21&#93; &#91;%emt 0&#58;07&#58;13&#93; &#40;Be4&#41;&#125; Rh7 &#123;
&#91;%eval -1270,19&#93; &#91;%emt 0&#58;00&#58;31&#93;&#125; 0-1

Although you don't deserve it,I must admit that this is one of the most beautiful and brilliant moves I've ever encountered in my life....

That being said,I still think that you're a nasty and mean person....
Dr.D

Uri Blass · Post by **Uri Blass** » Thu Jul 26, 2012 3:22 pm

Sedat Canbaz wrote:
lkaufman wrote: I hope you continue your work. Your testing hasn't been relevant for us since we haven't yet released MP, but when we do we will of course hope to do well in your testing.
In general people who disagree with you are more likely to comment than those who agree, so don't think the comments are a representative sample.
The one thing I don't understand about your work is why you feel that it is important for reversal testing to use a "perfect" book in terms of the normal White advantage. Obviously a very good book is vital for competitive play, but for engine testing we need huge samples to overcome the error bars, and if the price we must pay is using somewhat imperfect books, that is a necessary price in my opinion.

Best regards,
Larry

Dear Larry,

Thank you...

Be sure,i will have a BIG pleasure, when i will start testing Komodo MP in SCCT

Yes,i hope to continue my work,but at least i need to pause some of my activities

And still i am not sure which one to pause:
-Chess Benchmarks by Houdini/Fritz ?
-Working over Perfect Book Series ?
-SCCT Rating list (testing the Top Chess Engines) ?
-SCCT Super Book League (based on Top Book Makers)
vs...

It seems those critics are appearing, due to i have lot of activities-that's my mistake or weakness

And for solution (for less critics and for more rest),i think the right choice will be:
-I need to pause some of them...,but the biggest problem is that: ''i love a lot of ComputerChess''

Let's wait and see...the time will tell

About the openings issue,
As far as understand, you are looking for varied book,thats ok and i give you right, because you an engine developer
Its a very good idea that you need to test Komodo in various positions,before final release,actually if i was in your shoes,probably i would do the same as you...

But however,(during my testings) i did not satisfied by the performance of some tested opening lines
Of course,i wish i could test the engines with all grandmaster openings,
but it seems,some Top engines performance suffer,in case if we test the engines with various openings ...
For me its more important the performance of the engines than to use various openings...
That's why, as far as possible i prefer the strongest openings lines for both sides White and Black (as far as possible i keep the winning lines:Withes 55 % and Blacks 45 %)
I mean,its something like:i am optimizing my opening lines for serious official tournament

For example,the below positions are based on books,based some SCCT games,
And we see the winning percentage of some popular lines suffer...(i mean less than 40 % winning percentage with Blacks)

I'd like to point out too,(in my latest neutral short books) one of the main first moves is 'Nc3'
Of course, (for official competitions or for Playchess server),i dont suggest to be used this move as main
But if the engines are using same Perfect 2012b book (for both sides) then 'Nc3' is highly recommended
The main Nc3 move increases the possibility to be played the strongest Sicilian openings

Best Wishes,
Sedat

I see no reason to justify disabling the main line of the french defence
(1.e4 e6 2.d4 d5 3.Nc3 Bb4)

The only logical reason to disable Bb4 is if we get the conclusion that
Bb4 is bad for black.

If most chess programs do not know to play that line well with black
and there are good theory moves that they fail to find for black
then it is not a good reason to punish a programmer who teach his program to play this line well with black by not increasing the elo of his program.

lech · Post by **lech** » Thu Jul 26, 2012 4:35 pm

Houdini wrote:After some of the worst, also some of the very best.

Robert, it is a real pleasure to watch Houdini's play.

Houdini is very strong in the tactical play, hence pieces like mobility, goes forward and looking for a chance. It is a very good your strategy, even if it can lose a piece.

Sedat Canbaz · Post by **Sedat Canbaz** » Thu Jul 26, 2012 5:00 pm

Uri Blass wrote: I see no reason to justify disabling the main line of the french defence
(1.e4 e6 2.d4 d5 3.Nc3 Bb4)

The only logical reason to disable Bb4 is if we get the conclusion that
Bb4 is bad for black.

If most chess programs do not know to play that line well with black
and there are good theory moves that they fail to find for black
then it is not a good reason to punish a programmer who teach his program to play this line well with black by not increasing the elo of his program.

I see a lot of reasons

about why i need to disable the below line:
1.e4 e6 2.d4 d5 3.Nc3 Bb4

1)The overall winning percentage is approx.32 %, really very disappointing results
*Note:based on 495 SCCT games,played at different time controls (Blitz and Slow) on my latest decent hardwares

2)Mostly of the Top engines (i mean for Blacks) are fighting for a draw,in many cases the game results are ending with loss

3)Perfect 2012a's opening lines are neutral,which are mainly up to 8 moves and the lines are optimized for overall Engines performance
*Almost all Perfect 2012a's lines are with winning percentage approx.55 % Whites and 45 % Blacks

More details about the current critical French line :
I noticed too that only a few engines favors from the current critical line 1.e4 e6 2.d4 d5 3.Nc3 Bb4
But anyway,for me its more important the overall opening statistics than individuals
In other words,the critical opening lines or the disadvantage openings are not allowed for nowadays
Note also that the current opening statistics are based mainly on SCCT results,not sure about Non-SCCT conditions

Who knows ? maybe i will enable Bb4, when the engines will start to play better than nowadays,until this date i will prefer to NOT allow it

Best,
Sedat

JuLieN · Post by **JuLieN** » Thu Jul 26, 2012 5:30 pm

Dr.Wael Deeb wrote:That being said,I still think that you're a nasty and mean person....
Dr.D

[moderation]
Please guys, do not use this forum to insult other people.
Saying "I don't like you" is not an insult, saying "you're an idiot" is.
[/moderation]

Sven · Post by **Sven** » Thu Jul 26, 2012 6:59 pm

Sedat Canbaz wrote:
Uri Blass wrote: I see no reason to justify disabling the main line of the french defence
(1.e4 e6 2.d4 d5 3.Nc3 Bb4)

The only logical reason to disable Bb4 is if we get the conclusion that
Bb4 is bad for black.

If most chess programs do not know to play that line well with black
and there are good theory moves that they fail to find for black
then it is not a good reason to punish a programmer who teach his program to play this line well with black by not increasing the elo of his program.
I see a lot of reasons about why i need to disable the below line:
1.e4 e6 2.d4 d5 3.Nc3 Bb4

1)The overall winning percentage is approx.32 %, really very disappointing results
*Note:based on 495 SCCT games,played at different time controls (Blitz and Slow) on my latest decent hardwares

Hi Sedat,

I have two questions on this matter:

1) Which black replies to 1.e4 e6 2.d4 d5 3.Nc3 score so significantly better than 32% for black (within your 495 games) that you decided to skip 3...Bb4?

2) If you base such decisions upon only 495 SCCT games then how many different opening lines remain that you do not exclude for reasons like in the case of 3...Bb4?

Sven

Uri Blass · Post by **Uri Blass** » Thu Jul 26, 2012 7:20 pm

Sedat Canbaz wrote:
Uri Blass wrote: I see no reason to justify disabling the main line of the french defence
(1.e4 e6 2.d4 d5 3.Nc3 Bb4)

The only logical reason to disable Bb4 is if we get the conclusion that
Bb4 is bad for black.

If most chess programs do not know to play that line well with black
and there are good theory moves that they fail to find for black
then it is not a good reason to punish a programmer who teach his program to play this line well with black by not increasing the elo of his program.
I see a lot of reasons about why i need to disable the below line:
1.e4 e6 2.d4 d5 3.Nc3 Bb4

1)The overall winning percentage is approx.32 %, really very disappointing results
*Note:based on 495 SCCT games,played at different time controls (Blitz and Slow) on my latest decent hardwares

2)Mostly of the Top engines (i mean for Blacks) are fighting for a draw,in many cases the game results are ending with loss

3)Perfect 2012a's opening lines are neutral,which are mainly up to 8 moves and the lines are optimized for overall Engines performance
*Almost all Perfect 2012a's lines are with winning percentage approx.55 % Whites and 45 % Blacks

More details about the current critical French line :
I noticed too that only a few engines favors from the current critical line 1.e4 e6 2.d4 d5 3.Nc3 Bb4
But anyway,for me its more important the overall opening statistics than individuals
In other words,the critical opening lines or the disadvantage openings are not allowed for nowadays
Note also that the current opening statistics are based mainly on SCCT results,not sure about Non-SCCT conditions

Who knows ? maybe i will enable Bb4, when the engines will start to play better than nowadays,until this date i will prefer to NOT allow it

Best,
Sedat

I dislike the idea of disabling lines that top programs do not know to play well.

Maybe some new engine that is going to be released tomorrow knows to better the Bb4 line.
If you do not test you are not going to know and that engine is going to get worse rating then it deserves only because the programmer chose to add to his program a different knowledge.

I also wonder if all engines really score clearly worse with black relative to white in the french Bb4 line.

Uri

Dr.Wael Deeb · Post by **Dr.Wael Deeb** » Thu Jul 26, 2012 7:34 pm

JuLieN wrote:
Dr.Wael Deeb wrote:That being said,I still think that you're a nasty and mean person....
Dr.D
[moderation]
Please guys, do not use this forum to insult other people.
Saying "I don't like you" is not an insult, saying "you're an idiot" is.
[/moderation]

Fair enough

Sedat Canbaz · Post by **Sedat Canbaz** » Thu Jul 26, 2012 9:11 pm

Uri Blass wrote:
Sedat Canbaz wrote:
Uri Blass wrote: I see no reason to justify disabling the main line of the french defence
(1.e4 e6 2.d4 d5 3.Nc3 Bb4)

The only logical reason to disable Bb4 is if we get the conclusion that
Bb4 is bad for black.

If most chess programs do not know to play that line well with black
and there are good theory moves that they fail to find for black
then it is not a good reason to punish a programmer who teach his program to play this line well with black by not increasing the elo of his program.
I see a lot of reasons about why i need to disable the below line:
1.e4 e6 2.d4 d5 3.Nc3 Bb4

1)The overall winning percentage is approx.32 %, really very disappointing results
*Note:based on 495 SCCT games,played at different time controls (Blitz and Slow) on my latest decent hardwares

2)Mostly of the Top engines (i mean for Blacks) are fighting for a draw,in many cases the game results are ending with loss

3)Perfect 2012a's opening lines are neutral,which are mainly up to 8 moves and the lines are optimized for overall Engines performance
*Almost all Perfect 2012a's lines are with winning percentage approx.55 % Whites and 45 % Blacks

More details about the current critical French line :
I noticed too that only a few engines favors from the current critical line 1.e4 e6 2.d4 d5 3.Nc3 Bb4
But anyway,for me its more important the overall opening statistics than individuals
In other words,the critical opening lines or the disadvantage openings are not allowed for nowadays
Note also that the current opening statistics are based mainly on SCCT results,not sure about Non-SCCT conditions

Who knows ? maybe i will enable Bb4, when the engines will start to play better than nowadays,until this date i will prefer to NOT allow it

Best,
Sedat
I dislike the idea of disabling lines that top programs do not know to play well.

Maybe some new engine that is going to be released tomorrow knows to better the Bb4 line.
If you do not test you are not going to know and that engine is going to get worse rating then it deserves only because the programmer chose to add to his program a different knowledge.

I also wonder if all engines really score clearly worse with black relative to white in the french Bb4 line.

Uri

Dear Uri,

Believe me... i am not happy too with NOT allowing the current Bb4 line
In this way, the variety of Perfect book's opening lines are going down
But unfortunately i have no other way,exception to disable it

Btw,when we add more games,based on older engine versions with different Engine Elo strength,played on older (slower) hardwares
Then the winning percentage is increasing around :41 %

But however,i am concentrated on game results,played by Top MP engines
Plus,i prefer the games statistics,which are played on my Intel 6 core machines

Best,
Sedat

Sedat Canbaz · Post by **Sedat Canbaz** » Thu Jul 26, 2012 9:46 pm

Dear Sven,

1) Which black replies to 1.e4 e6 2.d4 d5 3.Nc3 score so significantly better than 32% for black (within your 495 games) that you decided to skip 3...Bb4?

Almost all Top engines (with Blacks) perform significantly better than 32%,when theirs opponents are approx. 200-300 weaker
But in case of match with similar Engine Elo strength,then the winning percentage is even less than 30 %

2) If you base such decisions upon only 495 SCCT games then how many different opening lines remain that you do not exclude for reasons like in the case of 3...Bb4?

If you are meaning about how many opening lines did you optimize or disabled between the both versions Perfect 2012 and Perfect 2012b ?
Then my answer is:
-I did not count all positions,which i already disabled or optimized,but this number is probably more than 300 positions
For example,in the next release, C42,C67,E97,D44,D17,D15...will be not allowed too (due to passive openings or less than 40-45 % with winning percentage

One thing more about the current Bb4 line,
Here are another statistics by different books,which are deep in moves

Note:the above book statistics are based on games played mainly by Houdini or Rybka...)

So the overall statistics of the current French Bb4 move look like:
6444 games = 30 % winning percentage

Actually i can optimize and increase the current Bb4 line with winning percentage to be around 45 %
For solution, i just need to add/import many new strong moves (probably at least 10 moves more) into the current line
But then....i am afraid that Perfect book's lines will be very deep in moves (not very good idea for testing the engines strength)

Best,
Sedat

60 games Komodo 5 against Top4 at 120m+3s

Re: 60 games Komodo 5 against Top4 at 120m+3s

Re: 60 games Komodo 5 against Top4 at 120m+3s

Re: 60 games Komodo 5 against Top4 at 120m+3s

Re: 60 games Komodo 5 against Top4 at 120m+3s

Re: 60 games Komodo 5 against Top4 at 120m+3s

Re: 60 games Komodo 5 against Top4 at 120m+3s

Re: 60 games Komodo 5 against Top4 at 120m+3s

Re: 60 games Komodo 5 against Top4 at 120m+3s

Re: 60 games Komodo 5 against Top4 at 120m+3s

Re: 60 games Komodo 5 against Top4 at 120m+3s