Humans v Houdini chess engine (Elo 3300)

Komodo TCEC @25 ply says 12...c5. I couldn't keep the engine running for longer because it's already 35º here in Seville.

well my houdini plays rook to c8 with enouch time not the time as TCEC championship is clearly that Bc6 is not a good move coz dont let the pawn go to c5 and open the center

so dont argue for something that is unquestionable " houdini is unbeatable by any human" dont came with idiot examples that only affirm the truth

IMpfren I challenge you im going to crush you

I dont want your money only demonstrate that you are wrong

$50 for a valuable life lesson - sounds like a good deal!

Which is today´s lesson IM pfren? 

Thanks for the nice answer Mr. pfren, I haven´t see that Grischuk-Caruana game, I´ll check it! 

hmmmm...

I have a question that wasnt answered after reading through this entire thread. Can't super GMS reliably draw against these engines playing the Berlin or Petrov?

Amazing to see how persistent these threads are!  Anyway, here again is what Nakamura recently said about chess programs:

"Computer programs these days are actually stronger than human players, significantly stronger. If I played a computer in a match I would get destroyed, I would be slaughtered, and that's saying something considering where I'm at in the world of chess."  Hikaru Nakamura - Q&A at Wahington University, St Louis, MO 2/26/11

I'm not one to argue with Nakamura over this so it seems pretty settled to me.  

This isn't to say that a GM could never, ever score a point against the beasts, only that the GM's would likely fare poorly in a ten game match.  Remember those odds games some years years ago?

http://en.chessbase.com/post/the-dzindzi-rybka-3-handicap-match

The GM's involved didn't fare that well ... and that was with material odds!  Plus it was seven years ago against Rybka 3!

Are we done here? :)

That's an interesting question. A match where a draw would count as a win for the human player would probably be of some interest.

Crappov wrote:

phenix28 wrote:

I have a question that wasnt answered after reading through this entire thread. Can't super GMS reliably draw against these engines playing the Berlin or Petrov?

Amazing to see how persistent these threads are!  Anyway, here again is what Nakamura recently said about chess programs:

"Computer programs these days are actually stronger than human players, significantly stronger. If I played a computer in a match I would get destroyed, I would be slaughtered, and that's saying something considering where I'm at in the world of chess."  Hikaru Nakamura - Q&A at Wahington University, St Louis, MO 2/26/11

I'm not one to argue with Nakamura over this so it seems pretty settled to me.  

This isn't to say that a GM could never, ever score a point against the beasts, only that the GM's would likely fare poorly in a ten game match.  Remember those odds games some years years ago?

http://en.chessbase.com/post/the-dzindzi-rybka-3-handicap-match

The GM's involved didn't fare that well ... and that was with material odds!  Plus it was seven years ago against Rybka 3!

Are we done here? :)

Ironically those links actually give me more confidence in my hypothesis that draws can be secured by elite super-GMs playing super solid black defences like the Berlin or Petroff. We can also use the expertise gained by centaur chess to help with preparation so that we can guide how the game plays out 30-40 moves into the game. The only pitfalls would be motivation to play at your very best against a brick wall and avoiding blunders

Hey.

Both sides, are ridiculously saying they'd win without any concrete proof. Has all GMs combined, or engines combined whatever - completed chess? Who knows if the theory is correct? Maybe many years later - when they complete chess - many things could be correct, as well as wrong.

For those rooting humanity - 

Play against a top notch computer and post your results. You are really just taking one of the engine's weakness and just seeing it in a stupid way really. Does it matter if the engine fails to spot the winning move? Does it really have to win all the games?

No. And even if they fail to see - it doesn't really matter either. (My personal opinion : You will be simply crushed.)

For those rooting for computers - 

I'd say the same. (although the results seem obvious to me)

I root the engines. I trust the power of silicon. If the GMs have so much knowledge, so much intuition - why can't they all solve chess? Atleast some part.

For those rooting humanity:

Furthermore, why do you rely on computers for analysis if you can do it better and computers "do not give much of reliable analysis"? LOL.

I'm seriously lolling at Topalov right now... ;)

PS) Nakamura lost to Stockfish 5 even with the aid of Rybka. Could have been a tie, but however - note that the computer wasn't top notch. 

3 years after IM pfren posted that "puzzle", I have been testing the newest top engines against it. This time at 30 ply, Komodo 8 x64 thinks the position is 0.0. For Komodo, every single move pfren provided is 0.0.

It's been 3 years and engines haven't been able to solve the puzzle?

http://www.chess.com/forum/view/general/humans-v-houdini-chess-engine-elo-3300?page=10

It amazes me how sometimes these chess gods look us in the eyes and say: pick any move, they are all fine!! 

If you show Komodo 8 29. h5, at 40 ply, it will say white is +50. Before that, at 30 ply, it likes 29. Re1 +29. As I have already mentioned, it thinks at least five candidate moves are fine!

Interessting.

I don't know which kind of chess engines you had referred to but when I feeded the position to StockFish DD (on my Android), StockFish came out easily and clearly with 21....f6 as the best variant (probably a draw 0.0 point), while 21...Rfc8? was clearly classified as a bad move for black (0.34 point). In fact, this end result was quite stable from depth=14 and up and showed unmistakably from depth=22. On my little tablet, the calculation took merely 1-2 minutes.

No, I must reject your opinion. It's NOT impossible for an engine to see that several moves before. In opposite, it's quite possible when you don't optimize (pruning the futile variants) too early on. For the analysis, I set the Multi PV=4.

Maybe it's difficult for Houdini only in this particular position. But for Stockfish 4 (very old now), the engine came up with 14. hxg3 as best variant (-0.22 point) and disregard 14.Qxb4 as bad move very clearly (-0.75 point) at depth=25. So there was not a mistake of the machine but rather a misjudgement of the human player, who overridden with 14.Qxb4 and led his game ineviatbly to doom.

And StockFish didn't need a LOT of thought to come out with 13.Qb3 as you said, just around 30 s on my old laptop!
But from depth=20 onwards, StockFish favors 13.gxh4 as the best variant to neutralize the treat on g3 (~0.0). The move 13.Qb3 is punished with a heavy low score of -0.59, clearly a bad omen.
I truely can not understand how this could be explained as "following a machine advice".
There is another mistake at 17.Rae1 where the knight could take the rook on f1. StockFish proposed 17.Rfe1 as a simple escape but recognized then 17...Qh4 as a very serious threat (-1.54).

For me, it's unmistakable a misjudgment of the human player which had led to the loss of white, not of the machine!
 ( { [Stockfish 270414] 28:+0.00} 13. gxh4 Nh5 14. Qb3 Nc6 15. Bxf8
    Kxf8 16. Qxe3 Qxh4 17. Qh6+ Ke7 18. Qe3+ Kf8 )
    ( { [Stockfish 270414] 27:-0.10} 13. g4 a5 14. Bxe3 Nh7 15. d5 Na6 16.Bd4 Nc5 17. Qc2 Bd7 18. e4 Re8 19. e5 Qg5 20. Rfe1 Qh6 21. Be3 Qg7 22. exd6 )
    ( { [Stockfish 270414] 27:-0.30} 13. Bxe3 Nh7 14. d5 Qe7 15. Bh6 Re8 16. Qd2 hxg3 17. hxg3 a5 18. e4 Na6 19. Be3 Nc5 20. Rae1 Qe5 21. g4 Bd7 22. Rd1 Qf6 23. b3 )
    ( { [Stockfish 270414] 27:-0.59} 13. Qb3 hxg3 14. hxg3 Re8 15. Qxb4 Nh5 16. g4 Ng3 17. Rfe1 Qh4 18. Qb5 Bf5 )
13. ... hxg3 14. Qxb4
    ( {[Stockfish 270414] 26:-0.25} 14. hxg3 Nh5 15. Qxb4 Nxg3 16. Bxf8 Nxe2+ 17. Kh1 Ng3+ )
    ( {[Stockfish 270414] 25:-0.75} 14. Qxb4 gxh2+ 15. Kxh2 Ng4+ 16. fxg4 Qh4+ 17. Kg1 Qxh6 18. Qe1 Bxg4 19. Qg3 Qg7 20. Rae1 Rae8 21. Rf4 Be6 22. Qxg7+ Kxg7 23. b3 Bc8 24. Rc1 f5 25. c5 Rh8 26. cxd6 cxd6 )
14. ... Nh5 15. hxg3 Re8 16. g4 Ng3 17. Rae1
    ( { [Stockfish 270414] 26:-0.34} 17. Rfe1 Bxg4 18. Qb5 Be6 19. Bxe3 Bxc4 20. Qg5+ Qxg5 )
    ( { [Stockfish 270414] 25:-1.68} 17. Rae1 Nxf1 18. Rxf1 )
    ( { [Stockfish 270414] 26:-1.54} 17. Rfe1 Qh4 18. Qb5 Bf5 19. Bf4 Kg7
    20. Qb3 Re7 21. Bxe3 Rh8 22. d5 Bxg4 23. fxg4 Qh2+ 24. Kf2 Ne4+ 25.
    Kf1 Rhe8 26. Bd4+ Kg8 27. Qd3 Ng3+ 28. Kf2 Rxe2+ 29. Rxe2 Rxe2+ 30.
    Qxe2 Nxe2 31. Kxe2 Qxg2+ 32. Kd3 b5 33. cxb5 Qxd5 )
17. ... Nxf1 18. Rxf1 Qf6 19. Qb5 Be6 20. Bg5 Qxd4 21. b3 Rab8 22. Qa5 f6
23. Bh4 Re7 24. Rc1 Rf8 25. Qc3 Qxc3 26. Rxc3 Bd7 27. Bg3 f5 28. f4 Bc6
29. Bxc6 bxc6 30. g5 c5 31. Kg2 Kf7 32. Rc1 Rh8 33. Rf1 a5 34. Be1 a4 35.
Bc3 Rh4 36. Be1 Rg4+ 37. Kf3 Re4 0-1

There are enough proofs of this ,that nowadays engines can't be beaten by any human player,the best that can be reached is some laughable draw ,nothing more.(and it depends on the opening)

Human with computer can beat a computer.