Super finale nTCEC saison 2

[Invité]

Grosse surprise suite au classement de la demi-finale !!!
1. Stockfish 20.5 pts  
2. Komodo 18 points  
3. Houdini 17 points !!!!  

Finale entre SF et Komodo... 

A suivre... nTCEC

[admin]

c'est à quelle cadence
j'ai pas trouvé sur leur site ?

[Invité]

Alors 48 parties de 120 minutes + 30 secondes.
Si égalité, 12 parties de 25 minutes et 10 secondes.
Si de nouveau égalité, 6 blitz de 3 minutes et 2 secondes.

[Lion_aile]

c'est le tournoi le plus proche du jpc en cadence je pense meme si pour ma part, je pousse l'analyse beaucoup plus loin pour chaque coup...

[photophore]

Et quelle bibliothèque ?   

[Lion_aile]

la c'est le truc que je comprends pas trop...

[Invité]

The nTCEC Computer
CPUs: 2 x 8 core Intel Xeon E5-2689 @ 3300 MHz
CPU Coolers: 2 x Corsair H80i
Motherboard: Asus Z9PE-D8 WS
RAM: 32 GB Samsung MV-3V4G3D/US @ 9-9-9-24
PSU: Corsair AX 760
SSDs: 2 x Samsung 840 Pro 128 GB @ Raid 0
Case: Silverstone Raven RV03B-WA

[Lion_aile]

la c'est le truc que je comprends pas trop...

je n'ai pas eu le courage de lire tout ca:

[Vous devez être inscrit et connecté pour voir ce lien]

[Invité]

Bibliothèque d'ouverture :

nTCEC uses openings taken from recent strong human tournaments for each Stage. Also, the format has to be as similar as possible to the current Stage, so for example for Stage 1 the openings used will be taken from Aeroflot Open 2012 and for Stage 2 the theme is Tata Steel Chess GM groups A + B 2010-2013. For Stage 3 the themes are Candidates Tournament 2013, Sinquefield Cup 2013 and King's Tournament 2012. For Stage 4, the themes are London Chess Classic 2011-2012, Linares 2010, Capablanca Memorial 2013 and Bilbao Grand Slam 2012. The rest of the openings for the Superfinal be revealed a few days before it starts. For all games, the GUI selects a random opening and plays it out until the engines take over after move 6. So all openings are 6 moves long.

[Invité]

PARTIE N°1



Komodo en difficulté en milieu de partie qui excelle en finale pour annuler...

[Invité]

PARTIE N°2



Surprise !!! Komodo marque le point !!!!

[Invité]

Don Dailey, le développeur du module Komodo est décédé cette nuit des suites d'une longue maladie (leucémie), au moment ou son module remportait sa seconde partie en finale contre Stockfish (annonce de son associé Larry Kaufman).
Lien

Triste nouvelle...

[Lion_aile]

Ouaip triste nouvelle ...

[Invité]

PARTIE N°3



Un nouveau gain pour Komodo !!!

[Invité]

Pour la quatrième partie avec 1.b3, Komodo avec les blancs change de variante avec 7.Df3...

[Invité]

Une chose importante : les 6 premiers coups de chaque partie sont imposés en partie aller et retour.... 

[Invité]

le dernier développement de Stockfish est-il plus faible que le précédent ???  

[Lion_aile]

oui je pense qu'il est plus faible
ils ont changé beaucoup trop de parametres

[Invité]

bizarre de le mettre en finale avec quelques bugs !!! les développeurs sont ils devenus aveugles quand à la suprématie du moteur en demi-finale !!!
A suivre...

[Invité]

De plus, Komodo a mis en finale le dernier développement de son moteur : version 1142 au lieu de 1133...

Est-ce le développement parfait ??????? je le souhaite pour feu son "père"...

[photophore]

Maintenant , entre 3 mn/ coup et 1 h/coup ( ma cadence minimum , mais je peux prendre beaucoup plus ) , il y a quand meme une marge , et la hiérarchie peut etre complètement différente
En plus le choix du début a une importance primordiale : en 2004-2005 , la suprématie se disputait entre Fritz8 et Junior7
Junior7 était imbattable dans les Gambits du Roi ( quelle que soit la couleur ) , ainsi que les Najdorf , mais perdait tous les GD
Le résultat d'un match entre les 2 dépendait des fluctuations d'échantillonnage
Mais ce n'était pas gènant pour un joueur qui , comme moi , ne répond jamais d5 à 1 d4

[Invité]

L'idée d'imposer les 6 premiers coups dans cette finale me semble judicieuse pour éviter les "bibliothèques maisons".

[Lion_aile]

bizarre de le mettre en finale avec quelques bugs !!! les développeurs sont ils devenus aveugles quand à la suprématie du moteur en demi-finale !!!
A suivre...

ils font des tonnes de tests milliers de parties
mais apres cela depend de l'ouverture, la cadence...
perso, je prefere intuitivement l'ancienne version quand on regarde l'historique des changements

[photophore]

ce que je veux dire , c'est que ce sont un peu les "24 heures du Mans" des modules d'analyse , qui n'ont pas d'intéret pratique pour le choix des joueurs
Mes critères ne sont pas ceux de Jean , parce que nous n'avons pas le meme style ni le meme répertoire

[Invité]

J'ai bien peur que si Stockfish remporte la finale, la version future devienne payante...(et de ce fait Houdini gratuit  )

[Invité]

PARTIE N°4

[Invité]

Pour les parties 5 et 6, l'ouverture proposée sera une est-indienne (attaque des 4 pions, variante 6.Fe2) :

1.d4 Cf6 2.c4 g6 3.Cc3 Fg7 4.e4 d6 5.f4 0-0 6.Fe2 c5

[Lion_aile]

Pour les parties 5 et 6, l'ouverture proposée sera une est-indienne (attaque des 4 pions, variante 6.Fe2) :

1.d4 Cf6 2.c4 g6 3.Cc3 Fg7 4.e4 d6 5.f4 0-0 6.Fe2 c5

comment le sais tu?

[Invité]

Je suis la partie sur le site officiel  

[Invité]

Nelson Hernandez (Cato the Younger in the TCEC chat) gives us an explanation on his approach to create the openings for the Superfinal:

"Here's a few points on my database. The entire collection has been thoroughly dupe-checked using a multiple-stage process. Each game has been adjudicated using a proprietary program we developed which verifies (and corrects, if necessary) game outcomes by ending evaluation or 6-man tablebases at the instant each game simplified to six pieces.

As you know, everyone has a different point of view on what games ought to be included and excluded from an opening book. My approach strongly favors quantity over quality based on years of experience, and the book is not oriented toward competition as much as universal coverage. Nonetheless, for every game-type there are particular standards with respect to Elo, time-controls, game length and other qualitative factors.

The point of departure for opening selections was to recognize that “human” databases and “engine” databases have different characteristics. What works in one may not work so well in the other both in terms of success-rates and draw-rates. So I pulled statistics from both game-types separately and this helped my analysis a lot.

I operated under the working theory that book-exit positions that are noticeably drawish in both human and engine databases are likely going to be extremely drawish in a Superfinal. The way to reduce drawing tendencies was to select opening positions that were less draw-prone in both databases while at the same time avoiding any positions that were too favorable to white or black. I also picked opening positions that had been seen at least 100 times in both databases. (Over 75% of the positions were 1000+ games in aggregate.)

Here’s what I’ve accomplished. In my aggregate database I currently have a 33.8% draw-rate. The average draw-rate in the book-exit positions I’ve picked is 28.5%. In the engine-only database (which is more telling) the same comparison is 37.4% (database) versus 31.2% (book-exit positions), a 17% draw reduction proportionally speaking.

What will this mean in practical terms? Well, I suspect that a random sampling of balanced positions would have produced a Superfinal draw-rate in the 75-80% range. Instead of that I am hoping that we will see a proportional reduction of draws to something more like 60-65%. It’s not going to be dramatic change and many who will only see what actually happened as opposed to what might have happened will still not be satisfied. But given the caliber of the engines anything under 70% draws without skewing the book-exits toward one color will be a subtle but solid achievement."

Source : lien

[Invité]

Souvenir - Souvenir   

[photophore]

C'est DEAD DRAW    : j'ai eu 2 fois cette position contre Baumgartner Si j'ai perdu la 2ème fois , c'est sur erreur de saisie normalement ça devait aussi faire nulle

[Invité]

J'avoue que pour le moment le choix des ouvertures n'est pas au top !!!  

[Invité]

   

[Invité]

PARTIE N°5

[Invité]

Dans la sixième partie, Komodo joue l'autre ligne : 7.d5  

[Invité]

PARTIE N°6

[Invité]

Parties 7 et 8 sur le Gambit Roi accepté (Kieseritzky) - défense de Berlin 6.Fc4      

1.e4 e5 2.f4 exf4 3.Cf3 g5 4.h4 g4 5.Ce5 Cf6 6.Fc4 d5 ...

[Invité]

PARTIE N°7

[Invité]

PARTIE N°8

[Invité]

Egalité parfaite 4-4  

Nous continuons sur une ligne  intéressante  l'attaque Trompovsky  

1.d4 Cf6 2.Fg5 c5 3.d5 Ce4 4.Ff4 Db6 5.Fc1 e6 6.f3 Cf6

[Invité]

Je ne sais pas où ils vont chercher leurs lignes, sachant que celle-ci a été vu la dernière fois dans cette partie  

[Invité]

Le Roi Kasparov l'a joué !!!    

[Invité]

Intéressant...

"Komodo is a very strong chess engine that is consistently rated in the top three on most major chess engine rating lists. It heavily relies on search and evaluation rather than depth and scaling, and thus has a distinctive positional style. Its forte is to play when there is nothing to play. Komodo author Don Dailey describes it as such: "In positions that most engines would likely struggle or find it impossible to make progress, Komodo quietly prepares a break and ends up with the victory.""

[Invité]

Hommage de Larry Kaufman à Don Daily...

Don Daily (1956 -2013)

Don Dailey, my partner in chess programs from RexChess to Komodo spanning a quarter of a century, died tonight, Nov 22, 2013, in Roanoke Virginia at age 57 from Leukemia, just about the same time that Komodo pulled ahead of Stockfish by winning game 2 in the TCEC final. He no longer even recognized his own family on the final day, so he could not know about this, but was still somewhat lucid the day before so he did understand that Komodo had made it to the final. He is survived by his wife Mary, one brother, and both of his parents. He was raised as a Jehovah's Witness and served in recent years as an "elder" in the church.

Don's life story is rather amazing. He grew up in relative poverty in Michigan, and never even graduated from high school (though he later got a General Education Diploma via a test). He taught himself how to program computers as well as how to play chess (his best rating was in Class A, somewhere around 1900 I believe). Despite the lack of formal education, by the mid 1990s he was the head of the computing department at M.I.T. (!!!!). Probably they never imagined he had so little formal education. It is a curious coincidence that Don had this job, that I was an M.I.T. graduate, and that Vasik Rajlich whom I worked with on Rybka 3 was also an M.I.T. graduate. It is likely that Vasik and Don interacted at the computer dept. there, although neither remembered knowing the other back then.

I first met Don in the late 1980s. He had written his own chess program and competed with it, although it was not yet very strong. His chess advisor was the notorious Sam Sloan who happened to be living in the Roanoke where Don had settled. That didn't last too long, not surprising considering Don's strong ethics and Sloan's reputation. When Sam had to flee the country to avoid prosecution he introduced Don to me, and I took over as the chess advisor. Soon I persuaded Don to move to Florida where I then lived to devote himself full time to computer chess, and we developed Rexchess, Socrates, and (with Julio Kaplan) Kasparov's Gambit. Then Don took the M.I.T. job and moved to Boston, later moving on to a position with a tech startup. He told me that the stock they gave him was on paper worth five million dollars at one time, but he couldn't yet sell it and the 2001 tech crash nearly wiped out his equity. About that time he married Mary and they moved back to his favorite Roanoke, a major center for Jehovah's Witnesses. He made a living doing a variety of programming jobs for most of that decade. Then we got in touch again, Don expressed an interest in writing a new chess program, and I agreed to help him although at the time I was still working on Rybka, Vasik having no objection. Neither of us imagined that it would become a threat to Rybka's primacy, but events showed otherwise. After my Rybka work concluded we decided to aim for a commercial product. After a while I agreed to subsidize Don so he could devote full time to Komodo until it went on sale. But about the same time he learned that he had a terminal illness which would become a fatal Leukemia after a few years. He never really made a living from Komodo until the final months of his life, but was able to get by due to very low cost of living in Roanoke, some savings from the tech days, and his wife having a job. He certainly could have made far more money doing business programming, but he wanted to do what he loved more than he wanted material possessions.

At least Don lived to see Komodo surpass Houdini in every stage of TCEC, although whether Komodo or Stockfish is the new number one is yet to be determined. During the last weeks of his life, following TCEC was a prime interest for him.
I'll post more information as it becomes available.

Good Bye Don, Mark and I will do our best to see that your brainchild Komodo continues to improve and give chessplayers around the world much joy.

Larry Kaufman

[Invité]

Une superbe partie de Komodo avec un sacrifice au 33eme coups !!!

[Invité]

PARTIE N°9

[Invité]

PARTIE N°10

[Invité]

Les parties 11 et 12 seront sur le thème de la Caro-Kann, variante des deux cavaliers.

1.e4 c6 2.Cc3 d5 3.Cf3 dxe4 4.Cxe4 Cf6 5. Cxf6+ gxf6 6.d4 Fg4

[Invité]

Position identique dans cette partie :

[Contenu sponsorisé]