Can Houdini 3 be beaten?

I look at the evaluation window to know when to resign; if it matters, I'll cover it up and present games where no knowledge of evaluation occurs.

This game was played without sight of the evaluation or what Houdini was thinking.

I lost with a Sicilian and a Caro-Kann, but then Houdini played an inferior variation of the main line of the Two Knights' Defense (It was book for the computer); as is typical, I had an edge but couldn't convert and had to settle for a draw by repetition:

Maybe IM Phren could win this, but I can't win won games. :)

1. e4 e5 2. Nf3 Nc6 3. Bc4 Nf6 4. Ng5 d5 5. exd5 Na5 6. Bb5+ c6 7. dxc6 bxc6 8.Bd3 Be7 9. O-O h6 10. Ne4 Nd5 11. Ng3 O-O 12. Bf5 g6 13. Bxc8 Rxc8 14. Nc3 h5 15. Nge2 Re8 16. Nxd5 cxd5 17. d4 exd4 18. Nxd4 Bf6 19. c3 Nc4 20. b3 Nd6 21.Bb2 Ne4 22. Qd3 Nc5 23. Qd2 Ne4 24. Qd3 Nc5 25. Qd2 * 1/2-1/2

Here is the second draw this evening; neither side can improve:

1. e4 e5 2. Nf3 Nc6 3. Bc4 Nf6 4. Ng5 d5 5. exd5 Na5 6. Bb5+ c6 7. dxc6 bxc6 8.Qf3 Be7 9. Bd3 O-O 10. O-O g6 11. Qd1 Ng4 12. Ne4 f5 13. Nec3 Bd6 14. Be2 Nxh2 15. Kxh2 Qh4+ 16. Kg1 e4 17. g3 Bxg3 18. fxg3 Qxg3+ 19. Kh1 Qh3+ 20. Kg1 Qg3+ 21. Kh1 Qh3+ *

In the previous game in a Caro-Kann,  White was better the entire game; I tripled majors on the e-file in a heavy piece ending with a dead equal position, drifted into an ending of -.68 (potentially drawable) and decided to save time and resigned.

Obviously against a human opponent I'd play for the win but decided to take a quick draw, my third out of 8 games. I took less than 7 minutes; Houdini took about 55 minutes:

1. e4 Nf6 2. e5 Nd5 3. d4 d6 4. exd6 exd6 5. c4 Nb6 6. Bd3 Nc6 7. Be3 Be7 8.Nd2 Bg5 9. Ngf3 Bxe3 10. fxe3 O-O 11. O-O Qe7 12. Re1 Nb4 13. Bf1 Nc6 14. Bd3 Nb4 15. Bf1 Nc6 16.Bd3 1/2-1/2

Tone it down Moses or I'll file a complaint with the moderator.

He must be scared.

Yeah sure, you are getting good results against Houdini all the time right? Btw I beat Carlsen every now and then just so You know... Both scenarios are very likely the way that I see it... Jerk!

Neither of them.

He is absolutely sane, a weak OTB player who does not know (or does not bother to learn) how to use his engine, and with a huge idea about himself, which, quite unfortunately, is not verified by any non-fictitious fact.

sloughterchess = next 2800+++ player:)

Iam wondering when i checked your so called games.. Its either an engine to engine game using the 4 Knight opening or your playing an engine much weaker than houdini3 cause even on my Houdini 1.5 some moves which played by "Houdini3" are really dubious and not even on his top 5 list.. And btw what do you really want to prove in here? I know that most GM would find it hard to draw Houdini even on a decent laptop even with 4 Knights,  and even with odds and that yes Houdini can be beaten not by human logic but by sheer brute force calculation that only an engine can do.. Maybe 1 or 2 people would take side with you.. but most of those here will mock you for its really hard to belive you can draw much less win against Houdini when even top GM cannot. Its a nomal human behavior to boast, so other will be jealous.. But when all evidence points the other way around.. It will just make u stupid. Not trying to offend you.. But maan common..

Can't we all just get along???  

Or better yet... just stop posting to this thread.  Every time it almost goes away, a flurry of posts hit it.  Let's all stop wasting our time trying to reason with someone who won't listen to reason.  If you read some of his early posts, he admits he changes Houdini's moves to suit himself.  If it makes him happy to cheat against his own computer, good for him.  But there's no point in feeding his ego any more when he's obviously just trying to provoke us.

I had a computer expert design the computer with the expressed purpose of getting optimum performance of Houdini 3. Roman, who uses Houdini 3 running on 6 processors calls my Houdini "baby Houdini", but when I told him my system's qualitites i.e. i7 CPU Socket 1155, 4GB memory, 500GB hard drive, he said that this was "good". So don't try to tell me my Houdini is not performing at a high level.

The last three draws posted were without sight of evaluation or whatever the computer was thinking; one draw was simply following Roman's and my analysis right to the draw. No thought was required because the computer and I went down a pathway where neither side could vary without standing worse.

Play through this game and you will realize why the move sequence Nf3/Ne1/Nd3 are all developing moves. To the best of my knowledge I am the first player to demonstrate that a Knight can gain time by returning to the first rank:

Since the Knight can gain additional time with multiple moves in the opening and middlegame, it should be obvious that multiple Knight moves gaining time are always candidate moves. In the next game each of these moves is a developing move as I define development: Nc3/Na4/b3/Nb2/Qc1/Nd1/f3/Nf2. What is remarkable is that in the game presented below, by the time White plays Nf2, White is clearly better even thought his Knights occupy “passive” squares.

What I'd like to do now is provide a more sophisticated example of development based on a game I played against Houdini 3. Moody-Houdini 3, Match 2013, Game 1: 1.e4 e5 2.Nf3 Nc6 3.Bc4 Nf6 4.Ng5 Bc5 5.Bxf7+ Ke7 6.Bc4!? (While technically this is not a Novelty, it is practically unknown; White anticipates an exchange sacrifice on f3 so the retreat Be2 dulls the effectiveness of that sacrifice:

6...Na5?! (Houdini tries to con me into winning the exchange: 7.Nf7? Qe8 9.Nxh8 Nxc4 =/+) 7.Bd3! (White demonstrates that the Knight on a5 is misplaced; White holds the e4 pawn in anticipation of h6/Nf3/Nxe4) d6 8.Nc3 h6 9.Nf3 Be6 10.Be2! Nc6 (There is nothing better. From the above examples the reader will readily assign a value of -1 to this move, but the move sequence Bd3/Be2 is worth 0 because White can reach d3 or e2 in one move.

11.d3 +/=a7 12.Na4! (Not to attack the Bishop so much; the main reason is to free the c-pawn) 12...Ba7 13.O-O Kf7 14.h3 Rf8 15.b3! (To free b2 for the Knight) 15...Kg8 (Look how many tempos Black lost by castling by hand) 16.Be3! Bd4?! 17.c3! (Houdini doesn't like this move) Bxe3 18.fxe3 b5 19.Nb2 d5! (Houdini “knows” it has to attack, yet this straightens out White's pawns) 20.exd5 Nxd5 21.Qc1 (Holding c3 and e3 and not blocking the d-file) Qd6 22.Nd1 Rab8 23.c4 bxc4 24.bxc4 Nf6 25.Nf2 Qc5 26.Rd1 (Preparing d4) Qa5 27.Rd2 Bf5 28.d4! e4 29.Nh2! Kh8 30.Nf1! Qb4 31.Nd1! Qd6 32.c5 Qe7 33.Rb2 Rxb2 34.Qxb2 Nd5? (This is a programming error; Houdini should force me to win the a-pawn, not just give it away for nothing.) 35.Bxa6 +- (The threat is Bishop retreat, a4/a5/a6, but this is over the computer's horizon so it thinks that other moves give White a lesser advantage; in a game between GM's the phrase you will hear is that, “White's pressure nets him a pawn.”)

Now that White has a clear edge, the Knights' will go on the offensive. Unfortunately I made a series of fatigue errors and lost, but this does not alter the fact that White gained ground the entire middlegame primarily because of his Knight maneuvers.

Houdini is great at holding inferior positions; here, for example, White had an extra pawn that had no hope of Queening. The computer's flawless endgame technique in simple endings is incredible.

The computer is good at exploiting poor piece coordination. One of the mistaken concepts of post members is that I "calculate". I never calculate more that 2-3 moves ahead in the middlegame because the computer will never allow  simple tactical shots but tries to play them routinely.

When playing Houdini I always connect my pieces, avoid pawn weaknesses and merely put my pieces on the squares where they belong without calculating anything; the computer makes no effort to stop me. In this game I played 14.Nd2 not because I calculated the temporary pawn sacrifice, but because I knew from playing this position 100 times that White was better and only 14.Nd2 made sense. I had no idea I was winning a second pawn.

1. e4 e5 2. Nf3 Nc6 3. Bc4 Nf6 4. Ng5 d5 5. exd5 Nd4 6. c3 b5 7. Bf1 Nxd5 8.cxd4 Qxg5 9. Bxb5+ Kd8 10. O-O Rb8 11. Bc6 exd4 12. d3 Qf5 13. Re1 Be6 14. Nd2 Qxd3 15. Nf3 Qxd1 16. Rxd1 Kc8 17. Nxd4 Bd6 18. Nxe6 fxe6 19. Bxd5 exd5 20.Rxd5 Re8 21. Kf1 Be5 22. Rd2 Rb6 23. Re2 Kd7 24. Rb1 Bd4 25. b3 Rxe2 26. Kxe2 Ra6 27. a4 Re6+ 28. Be3 Bxe3 29. fxe3 Rb6 30. Kd3 Rd6+ 31. Kc3 Rc6+ 32. Kd3 Rd6+ 33. Kc3 Rc6+ 34. Kd3 Rd6+ * 1/2-1/2

Lol.

This "expert" either :

1-hacked the (proprietary) source code, which is illegal, to make another better engine which ould justify a commercial exploitation ;

2-upgraded the calculation engine (better computer, aka silicium steroids), from which I still wouldn't expect much improvement ;

3-claimed to have done something, and actually did nothing.

Being given the ridiculous games you post, the most probable is 3- maybe with the addition of fiddling some setting from the interface (like "stop looking at the opening book past move 5"), effectively making it worse.

lol now this is funny i did not mention that your Houdini is not performing on high level what im telling you is that either its an engine versus engine match or its not Houdini you had won against. Can you give the computers evaluation log that is saved on your games with Houdini?? That would settle the issue

I suggest you look at post 104 where I played the 6.Bc4 variation of the Wilkes-Barre/Traxler against Houdini 3. If it is an engine versus engine match, how do you account for the fact that many of White's choices were criticized by Houdini? Houdini thought it was doing well when I had an advantage of +/-.

If it were an engine versus engine match, don't you think the engine would consistently choose moves the engine thought were good? Show me one engine that would choose my 15.b3, for example. Houdini criticized me for 17.c3, yet it is the best move.

I told post members I would look at the evaluation when playing the games and was told I was cheating so I stopped looking at the evaluation and drew something like 5 games in the past 36 hours.

I have already jumped through hoops to satisfy my critics. I was criticized legitimately for relying on a laptop with inadequate software to run Houdini 3 at a desired level, so I took the box with the desired systems requirement to a computer expert and asked him to design a computer to meet the systems' requirement on the box so he provided me with the i7 CPU Socket 1155 4GB memory 500GB harddrive.

Show me anywhere that this is an inadequate software arrangement to run Houdini 3 at the level suggested by the manufacturer.

You can demand anything you like, but your request is ridiculous.

A couple of things about my computer setup---it was designed to meet the systems requirements suggested by the manufacturer.

It will play book lines of the Sicilian out to move 15.

When left accidentally to crunch a winning position it took 1/2 hour to announce checkmate in 45 moves.