Which Engine Plays Like Human?

Yeah, he did a number of endgame studies.  I think the basic winning position and technique for R+B vs R is his too.

I used the Crafty 19.19 engine inside Junior 10 interface. I usually use Junior 10 engine or Fritz 5.32 engine, but mostly Junior 10. For analysis I think Junior 10 is more useful. For playing I also prefer Junior 10 but set up at at least 2200. This way the engine creates "small mistakes" (I think it is the best move within a short period of calculation) that is tolerable. And the point is how to beat the engine for making those small mistakes.

It seems like those small mistakes are actually the best moves that the engine can find within certain constraints such as calculation depth or number of iteration in the calculation algorithm. I think this is more tolerable than dropping pieces.

Mm. And the basic R vs R+P defence too.

I heard about it here, by marvellosity.

Of course that isn't realistic. But I don't believe any programmer, even if they don't know chess well (surely there is a chess player adviser beside them) will come up with that stupid idea to simulate different strengths. They can cap the depth of the calculation, combine it with other constraints for example.

From my experience playing with Chessmaster GME (above 2210 rating), he did drop a pawn, but with a positional compensation that can be understood tho inhuman.

What?! When I played with Chessmaster I had to choose a pawn up with better position but with opposite color bishops on the board. I didn't think I could win the position but I tried. In the end I had to accept that it was a draw so I offered a draw (it's just 10 minute game but shuffling your pieces like that seems forever). But Chessmaster never accepted! With a draw position and worse position and less time he didn't accept the draw offers?? So I decided to punish him to make him lost on time, so I pay attention to not repeating the same positions 3 times. But then after so many moves (it was the longest game for sure) he drew the game without saying there was a move repetition! I thought there might be some kind of 50-move rule or something but I felt like he cheated on me. Poor me, I didn't realize there was the 50-move rule handicap, so the next time Chessmaster refused my draw offers, I simply resigned the game.

Interesting, because I've never noticed that phenomenon. May be because I prefer to play in rated mode, and a relatively high strength. For example, when I played with Chessmaster2000 I played only with its full srength in tournament mode. When I played with Chessmaster GME I played against 2210+ players. When I played with Junior8/10 I set to a minimum of 2100. But I don't like to play with Fritz 5.32 because I found he was too strong in that level, even when I set to 1800 about half year ago he was still stronger imo (but I have never experienced Fritz 5.32 play weak moves such as dropping material).

But I did notice that the Chessmasters have poor opening and middlegame skill but very strong endgame. Junior just makes "weak moves" in the middlegame to endgame, sometime intolerable if set below 2200 but most of the time I can tolerate. But with Opening Book option, the Junior almost always pick opening lines that give higher winning percentage in its database.

May I ask the opposite question: which humans play like engines?

As a chessplayer and (once) a programmer, I know exactly that you can program a posititional understanding (and it has been done), it is just not easy to achieve positional understanding of master level. "Not perfect" is the key. Sometimes not possible at all (even to compare with my positional understanding). It is still one important weakness of engine as compared to human.

Current engines still rely very heavily on tactical calculation. And this is highly affected by processor speed. The number of possible moves are exponentially increasing so running an engine in normal processor in 10 days will not give better result than in 3 hours (my prediction).

Computers just cannot see the future.That is what you can utilize to beat an engine (tho I think it is unfair). A common phenomenon is an engine cannot see if a certain "hedgehog" position contains dangerous threat. It is crowded and passive but it is like a spring that can release high energy.

Bishop pair potential is also difficult to program because it depends on the position that cannot be modeled.

And my question: which players play like me?

Everyone is different. Even the same engine can be setup with different styles. Even I can change my playing style.

And everyone makes mistakes, big or small. So, actually it was not about finding a perfect engine, or engine that can play like human, but originally I wanted to know which software that can be used as a sparring partner to improve my chess. Of course all engine can, but there must be some characteristics that will work better to improve our skill.

What motivated me to ask the question was this:

When I first played with Lori (2210) I was beaten for more than a hundred consecutive games, then I started to win but lost on time, then I started to outplay her in 3 days only. Then I moved on to beat higher rated players. If only it is that fast and easy to improve and beat real human players, isn't that wonderful??? What if I can beat Kramnik (the one in Chessmaster) after a million games, should I join the World Championship?

 Players who play computers all the time learn only how to play against computers.

In a real game the players have to press their own clocks, record their own score and move their own pieces. The computer does this for them.

If you are beating a good player he will try to confuse you by playing trappy and double-edged moves. The concept of confusing a player is something a computer does not understand and it would never gamble on playing a move it knows is bad in the hope the human player misses it.

Any experienced tournament player reading this will know this happens in every  game in every round in every tournement. Human players turning lost games into won games by simply playing double-edged moves and complicating the game.

Eventually constantly playing a computer makes the human player unable to spot human errors and the art of 'swindling' is something they never gain because you cannot swindle a computer.

So to answer your question. "which humans play like engines."

Humans who play engines.

Easy to recognise at tournaments. They are the ones blowing won games.        

OMG, kidzebra, you said it well

So yusuf...could you please tell me which program you suggest? I am also very interested in this feature....and you have done a lot of testing....so which engine/program should I get?

thanks

That is the point of my question too, so I don't have the answer yet :)) And I'm not sure if I can be convinced easily on this. Especially because we have to combine many trainings and the methods. Anyhow I think the various playing styles of the Chessmaster's personalities is a very useful feature, especially for rapid games.

Didn't you figure the answer until now? You should not play against a computer! It will never play like a human. Never! You will be surprised to beat your computer at let's say 2000 ELO and, when playing in the real world, find out that your 1600 opponents put up a much tougher resistance. You would see them creating tactical traps, throwing everything at you in a worse position, while the computer you were used to was simply going down slowly.

A long time ago I had a friend (2100+ ELO) who drew 2-3 standard games with "Karpov" in Chessmaster. It was clear to him afterwards that he should delete this software from his computer.

Rybka, Robolitto, junk compared to the experimental engine I use. B.R.A.I.N. Broad Ranging All Incompassing Network. Then again if you bribe me I will let my M.I.T Professor know but the bidding starts at 1,000$ US.

This is where I cannot easily be convinced :) As long as we know its limitations, what we can achieve from them, what training method we should adopt, what other trainings to combine, I think computer is still a valid training tool.

Yes, I'm very curious to find out what level I would come out with if  play seriously with human. Because I believe we can still "handle" human the way we handle engines. We just need to find out the right method.

Most engines do not play "tactically" or sharply. That's why I value engines that can provide attacking personality, gambit playing style.

Oh I've seen that before, but I never knew that was another philidor position.

Costelus, sure the computer is not concerned with practical chances, but it's still a tactical beast, and compared to a human extremely resourceful and never gets frustrated. I don't know if the elo they give on the computers is realistic though, they could just be estimating.

I have Fritz 10 and like it best of the ones I've tried. I'm a beginner, so most of the good chess programs can beat me even at lower levels. However, none of the various software feels like a human being. Strange that playing online against a real person (that I can't shake hands with or see) still feels very different than playing a computer.