you can find in the table I posted before, the performance from the predecessor Version 0.70 I have tested (same tourney) ... good to compare, better as in my rating list with all games.
Version 0.71 I haven't tested. Updates from Andscacs comes very fast and I was not ready with my FCT1-2015 tourney and testing 0.70.
But now ...
Programmer gave the information ... around +50 Elo!
Andscacs is one of the engines with the biggest Elo jumpings in the last months. I will see it ...
Next one will be Cheng4 0.38 x64, will be start around Friday this week. A lot of updates for testing at the moment.
Happiness consists in contentment. For the programmer and for us. That is allways the same. But most important in testing Andscacs is the king safty. Predecessor versions have here much problems. Interesting to looking here, not on Elo only! Can be see later with the final results.
Elo goes higher and higer ... at the moment after 125 games around +150 to 0.70.
very nice games I follow here!
Interesting are the results so far ... vs. tactical stronger engines a bit problems (if we are looking on the results only: Nirvana, Hiarcs, Junior). Furthermore, problems vs. Deuterium ... here in endgame ... Deuterium is very strong in endgames.
Most interesting are the stratgic ideas by your engine!
Will give you an example, follow the game slowly without an other engine runs in background as analyze engine.
Very interesting computer chess on a high level!
Here is the game:
Full of strategic ideas after a very complicated opening for chess programs.
Chiron try to get with White pieces a bit more in the opening. Andscacs hold the position closed without any mistake. More or less your engine is in waiting position (for white mistakes).
Interesting is black move 52 (... 52. e5).
I follow the complete game and hope of that move. Complicated position, now Andscacs with a small advantage but it's not easy to win that. Nice what both programs Plays with Knights and Bishops (both programs try to find again and again better positions in the late middlegame). Nice game with a hand full pawn later only. I believe a Human can't play such a complicated game ... each move is difficult. I am sure a human lost ... not important with black or white ... to many possibilities on the board. Means such position a human have to play very exactly over a long time ... not possible.
This game you put is terrible complicated! Sure, as you say, any human will fall at some point at those tactical lines.
Frank Quisinsky wrote:
draw eval is very good in the late middlegame. Early your engine give 0,00 in clearly draw positions. That must be new ...
I think is more a search artifact than anything I had problems at some point with Andscacs finding repetitions where they had no sense, and I was only capable of solving it superficially. Anyway I can verify some positions if you have example games, to say if it's doing well telling 0.00.
In my Rating list of all games ... Version 2.01.300.s2 over 140 Elo ...
With so many participants my results are very clear after around 600-800 games. So we have to wait ...
It seems a good seria (game 101-200) a bad seria (game 201-300) and a normaly seria (game 1-100).
not easy to search the games.
I am looking a lot in the still running games and after all what I can see I have the Impression that your engine can see draw earlier as others. Over 45% draw Quote, but often your engine can make draw in bad position. Really interesting what Andscacs do.
Have a look in game 316, lost vs. Rodent ... black move 12 f6 ... not understand the opening ... not sure, perhaps I found a bad line in my book.
Have a look in game 68, lost vs. Junior ... white move 33. a3 (clearly blunder).
Nice games vs. Smarthink, game 251!
Fantastic game vs. Naum, game 62, this one you have to replay ... directly!!!
Database is full of interesting stuff!
Games vs. Texel are intersting, late middlegame your engine is very aggressive, search each time activity. I saw also games Andscacs lost such games, to many activity.
In case of Andscacs it isn't Transposition into endgame (most engines are clearly stronger here). More end of middlegame before transposition into endgame your engine seems to be very very strong in building a lot of own activity. Karpov style, started with strategic ideas in the beginning and if the position is good, the surprise comes fast.
Could be an order for the future to try that directly after the opening books moves. More aggressive pawn moves could give more activity. But it seems your engine flip the switch. Means ... later in the game, your engine is more and more aggressive. In the beginning of the game Andscacs is looking what is happen, very human style. Really, interesting chess! More human GM chess in the late middlegame but much more exactly.
After 331 games:
1 game lost before move number 49 and two games won. King safty problems, easy to see in other Andscacs version I can't see at the moment.
