This would be a sparing opponent for my son 1700-1800Elo but improving rapidly. This question has come about because finding online opponents for anything 30 minutes plus is often difficult.
I have access to a strong player and coach (approx) 2250 Elo next weekend* - I will get him to play some of the recomendations and report back.
* His daughter and my son will be competing for places in the U11 England Squad so we will need something to distract us during the wait between rounds....
That said, I think the Homer engine claims to change its playing style to match his opponent's as you play more games, so it may approach its human playing style, or something. Probably worth a look.
So the feature is very inhuman-like.
Actually, Pro Deo is an engine that gets this constancy of strength right, with the problem that it's still way stronger tactically than humans so it's going to punish and beat you if you make a mistake, unlike a human that may overlook the mistake. I'm still going to shameless advertizing self-plug my version of Pro Deo for this, called the Drunken Master Suite with installation instructions at the top of this post (huh, just unzip and install RebelUCI.)
Thanks I will add it to the list...
P.S. The second link did not seem to go to the right place? However I think I am fine from the zip file!
> Thanks I will add it to the list...
> P.S. The second link did not seem to go to the right place?
It was meant to go to a post that said this:
Another approach instead of handicapping a strong engine is to use one of the lower rated UCI enginesHi,
There is a very stable engine rated at 1900elo in the UCI Engines League called Clueless 1.4 by Helmut Fett. I have used this a lot and in spite of it's name it plays very good consistent chess at that level:
Another one rated about 1880 in UCI Engines League and very stable is ROCE 0.0380 by Roman Hartmann:
Also Monarch version 1.7 has been out there for many years and I have used it a lot. Rated about 2000 .You can find it here:
Here is a great list showing UCI engines of all strengths with download links, although I have found the above 3 the most consistent and stable:
The fact is that the way a human, ELO under 2000, plays cannot be characterized in any trivial way. It would be really good if qualified people [IMs and Chess Engine designers, preferably both in one person] would turn their attention to this problem.
As an old person, I have deeply ingraned counterproductive mental habits which I cannot break. To avoid these, I find that I must make an almost super human effort while playing my games.
The best thing that a trainer can do for a young chess beginner is to help that beginner to avoid forming bad thinking habits.
The way young people play is different from the way older people play. Two players with the same rating, but one young and the other old, would play differently.
My post-mortem analyses of my games show me that I am not the only person who makes bad moves. Most of the games I "won" were actually lost by my opponent who simply made too many bad mistakes.
Today's "dumbed down" chess engines are like neanderthal cave men. They simply do not do the job!!!!!!!!! It is time for a modern approach to this problem.
> My post-mortem analyses of my games show me that I am not the only person who makes bad moves. Most of the games I "won" were actually lost by my opponent who simply made too many bad mistakes.
This is true regardless of rating and continues on up to very high elos.
All of the above is mainly for the middlegame and maybe endgame too.
Openings can be handled by using customizable opening books. One might make a copy of a high quality general-purpose opening book and then delete all lines which would not occur with the human's opening repertoire. This would be done seperately for White and Black. It is also possible to adjust probabilities so that openings of special interest come up more often. This process might help someone to develop and refine their opening repertoire.
You can't predict what a human player would play if you don't have data of the human in that position. Good luck having a program emulating Kramnik's blunder here:
Apparently, the pattern with the mate treat of the knight in last rank and queen as seen on the board is extremely rare so Kramnik hadn't learn it and several commentators missed it as well:
And, huh, apparently big blunders like that have happened before to famous chess players:
1.d4 Nf6 2.c4 e6 3.Nf3 b6 4.a3 Ba6 5.Qc2 Bb7 6.Nc3 c5 7.e4 cxd4 8.Nxd4 Nc6 9.Nxc6 Bxc6 10.Bf4 Nh5 11.Be3
I know of that Karpov blunder game since it was in one of Silman's books.
As I see it, the simulator would have to contain a chess engine or at least incorporate a lot of chess engine software.
The theoretical challenge of representing human chess thinking is daunting as well. Do we really know how humans think?
My intuition tells me that amateur chess is dominated by relatively horrible blunders, primarily tactical but sometimes positional or strategic. A GM would probably find it very hard to watch an amateur game.
The programmer would have a problem expressing the characterization of human mistakes into a form that chess engine software could process.
Perhaps this is simply asking too much! Who would want to take on such a challenging task? But beware! The engineers in Japan are creating robots and they, the robots, may play chess too! Watch out, Issac Assimov!
Of course, awhile back they said it would be too difficult to write computer programs that would play good chess. Hmmmm!
>My intuition tells me that amateur chess is dominated by relatively horrible blunders, primarily tactical but sometimes positional or strategic.
Can you define this three kind of blunders?
I don't know if there is a big difference between "strategic" and "positional", I think they are basically the same thing, but if not, someone will correct me.
> I don't know if there is a big difference between "strategic" and "positional", I think they are basically the same thing, but if not, someone will correct me.
I just think one is long term and the other is short term. If suddenly a player is winning and nobody knows when the blunder was made and it turns out it was a bad move 20-30 moves ago that looked natural at the time and to which the reasons it's bad aren't obvious until much later, it's a strategic blunder.
Maybe a move by the Bishop allows the opponent much later in the game to force it into a locked corner, so moving it into the corner isn't the blunder because it's the only move; the blunder was made when the variation was allowed, and it would have been a strategical blunder.
I haven't done this myself but my guess is that you would see subtle differences as you climbed the ELO ladder. A 1200-level book would have lots of really absurd openings, practically losing within the first six moves a lot of the time. A book at the 2000 level might play commendably until move 15 and then suddenly play a bad move. There would be a very nice variety to the play.
I said this was all very easy. Of course, conceptually it is very easy. Actually doing it is a lot of work! But I suppose someone could do a credible job of it and perform a fine service. Now, getting an engine to play like a human once the book has been exited--that's more of a challenge.
> Now, getting an engine to play like a human once the book has been exited--that's more of a challenge.
The biggest challenge would be to make the noticed strength difference between book and exit smooth. One moment you're playing a 7 years old that has just learned the game, the next you're playing Kasparov on his prime.
Actually, current approach of human emulation feels like that, you're playing a super strong opponent that now and then drops a piece for no reason, very inhuman like.
Obviously there are problems with this. The engine might play with no evident plan and make many unnatural moves a human would never play over the board. Calibration to simulate a given ELO would certainly need to be meticulous. Sometimes the engine might play far above its projected ELO by chance and other times it might play abysmally. You would never get the same sense you get playing against a live human, that's for sure. We are not yet at a stage where machines display human personality traits. However, I contend that programmers can simulate human OTB behavior pretty well with a little imagination.
I think that the approach I've outlined, when combined with a big opening book that plays just like human players of a given strength, would be a workable solution. This does not seem like a really difficult programming challenge to me. It's a lot of work, though.
I remember suggesting this at least a couple of years ago to one of my friends. You're absolutely right, of course. We need The Big Reset.
I checked, and actually I have proposed this at least three times before. First comment, November 12, 2008:
http://www.cnbc.com/id/27641538 if all this comes to pass, look out. The United States declaring bankruptcy, imagine that. Bankruptcy is such a loaded word, though. I prefer the happy word "jubilee". The United States declares a jubilee! All debts are cancelled! Hurray, there is a reset button after all! Then we can party like it's 1999 again! Obama would be a global hero if he simply declared a jubilee!
Seriously it seems like every story every day is more ominous than the previous day. I see no end in sight and things getting much worse than people dare contemplate. We're staring at the abyss. A jubilee may be the only answer.
> Including trash talking dialog during the game would be a nice added touch!
That's already within reach of the UCI protocol, the Gaviota Chess engine sends:
Emoticons on fail lows or fail highs.
It would be trivially easy for a programmer to make the engine send phrases like:
"Wow, nice move, where did that come from? :-O"
"That doesn't get the bull off the ice!! X^D"
"Pfft, what was that? OMG you suck! >:-("
Depending on score changes.
Would this really help to make the engine more human-like? Because it's something way easier to implement than what Nelson suggested.
you play like a girl
im gonna defeat you,patzer
sorry,can i go to the toilet?i need because you defeated me
try tomorrow,now you are wasting my time and im gonna drink a beer
>> Including trash talking dialog during the game would be a nice added touch!
> That's already within reach of the UCI protocol, the Gaviota Chess engine sends:
> <img src="http://i.imgur.com/Azbru.png" class="emi" />
> Emoticons on fail lows or fail highs.
> It would be trivially easy for a programmer to make the engine send phrases like:
> "Wow, nice move, where did that come from? :-O"
> "That doesn't get the bull off the ice!! X^D"
> "Pfft, what was that? OMG you suck! >:-("
> Depending on score changes.
> Would this really help to make the engine more human-like? Because it's something way easier to implement than what Nelson suggested.
I have actually a project that is bit stalled related to this. Fernando Villegas wrote several of this messages for different situations to be used for Gaviota, but I have not implemented this yet. The main problem are the GUIs, which are not really very friendly in communicating those messages to the user. Is there a GUI that would communicate this UCI strings in a suitable way? Sort of a chat window to dump this messages? winboard have pop ups, which is not convenient.
I would really like to get this going.
Also, please note you don't need a GUI for this, you can have your chess engine launching a third party chat-like window with the message of the engine spoken out, and it would appear in all GUIs where the engine is used and even if the engine is used via Command Prompt.
I don't know how to do it (I'm not a programmer) but I've seen it done with Rybka Visualizer:
There, Rybka Launches her own window with Internal Analysis Output and doesn't depend on a GUI to show it. Chat-like messages could be sent that way. Chat-like messages could also be received from the user and reply properly picked out ala jabberwacky.
Still, this could be used to communicate with the user in a friendly way. The engine could trashtalk now and then, but not be based on this.
He is wanting a digital computer sparing opponent for his 1700-1800 ELO son. He does not want a sparing partner which would discourage his son by whipping him every game.
What his son needs may be somewhat different for different phases of the game. For example, the development of a personalized opening repertoire is very important for someone just starting out in chess, especially if he plans to compete in rated events. Perhaps playing against someone who plays openings badly would be counter productive. Or, maybe not??
My feelings regarding the middlegame are that avoidance of bad thinking habits [snap moves, not looking to see what Opponent might do, etc. etc.] is very important for the beginner. Perhaps one of the worst bad thinking habits is not looking for opportunities which unexpectantly come up in the game. A machine which could serve as a trainer should offer opportunities of a type that someone at that level would have a chance of spotting if he/she looked for it.
I am not sure what to think about endgames.
Rybka 2.3.2a mp 32-bit seems not to be very Stratagically & Positionally Orientated.
I played a Remise game with her where she could have opened up the Position in the sake of Stratagy & Positional Play, but she did not because I would have gained an Open RookFile or she would have been 1 Pawn back... but that advantage would have finally Diminished for me I think.
one of/the Extern Weakness(es) of some, or all Chess Engines is that they think that you play 100% Perfect.
They think that you see the Tactics so well as they do.
So if they somewhere see that by opening the Position you get a small, Tactical or Strategicall ( which is Programmed I guess ) Advantage - they don't want to do.
However the Open Position by itself can already be losing for you as we know...
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill