Playing against strong chess engines.

I am wondering if anyone else on this site is in the habit of playing against strong engines and analysing the games afterwards as a means of improvement?

I've been trying to do this rather systematically lately and I've noticed dramatic results in my improvement. I'll set HIARCS on it's highest level (with an estimated elo of 3000), knowing that it's very likely that I'll lose (probably not make it past move 35) and afterwards analyse the game closely for both sides. I think that playing against an engine systematically in this way is possibly better than playing against stronger human players. The computer is blunderless and develops its "ideas" in a completely different way than humans do.

I'm just curious. Maybe it's a common thing for people to do this. I know that it's something I've avoided on my path to improvement because I'm fully aware that I'll be crushed by the computer. 

In a strange way, though, I'm actually beginning to marvel at the remarkable chess HIARCS (and other ultra-powerful engines) plays. It comes up with some pretty stunning moves at times.

Thoughts?

I can only attest to the fact that my game has been improving dramatically since I've started doing this.

Yes I do this on a regular basis. However, I haven't bothered to actually analyze the games too much.

One thing that certainly is different from playing humans or weak engines is, that my position always becomes extremely passive as I have the feeling that the engine will shut down any kind of play that I attempt.

What I felt to be quite fascinating though, is that different of these strong  engines, while relatively equal when playing each other, have distinguishable styles.

For example Stockfish tends to push its way through me with it's pawns, Critter mostly does overloading weaknesses with its pieces and Houdini is even fond to trade me all the way down to the endgame and finish me off there by exploiting all the small and not-so-small advantages it got earlier.

Of course none of them will ever miss any tactical shots but the circumstance they have different approaches when it comes to strategy is really nice to see.

This is why it's not as useful as playing people.  Not only will your human opponent not play this way, but you can't play this way either (unless you change your style to calculating millions of lines, which is where its "ideas" come from).

Also, you never get to practice winning a won position (very important and sometimes very difficult skill) or even realistically pursuing an advantage.  It teaches passive play which is a chess sin.

Good question, it pops up now and then on the forums.  What's common is the advice against it actually.  Conventional wisdom goes even further and says class players shouldn't even use engines to help analyze a completed game.

Still, I think playing full length games against an engine now and then potentially has two good points.  First is Dan Heisman's concept of hope chess being played more or less by all players under 1600.  Playing an engine helps abandon the notion of your opponent missing a winning tactic for itself... so you always always blunder check (or lose instantly to simple tactics you never see).  My guess is this is the aspect you're seeing most improvement.

Secondly, I've been told some players are uncomfortable facing humans (even online) and playing an engine takes the emotion out of it so they can relax and approach the position more rigorously and without hurt pride when they lose.

@wafflemaster

I actually find playing against a computer to be siginificantly more intimidating than playing against strong human opponents. The emotional state evoked is different but with a computer there is always a feeling of certain defeat awaiting you.

I can't understand why the people on this website have problems with the central role chess engines have in modern chess. It seems to me that engines are the biggest reason that you are seeing more and more incredibly strong human players appearing. I also can't understand why playing like a computer is bad! Certainly the psychological process will be utterly different than what the computer "experiences" but playing in a way that is blunder-free and positionally sound seems worth imitating.

Also, have you seen the way the latest engines play?! Sacrifices abound, material and positional, amazing defensive abilities, perfect attacks, etc.. They are beginning to play more and more like ultra strong human players should play.

I have long thought and been interested in an experiment. Imagine that some chess star was on the rise and crushing everyone in his path. Some years pass, and this person essentially goes undefeated in human play and becomes a legend. His games are analysed and marveled at for their solidity and perfection.... etc. Years later, it's discovered that he had been getting told the best moves all along through a small earpiece where a person, far away, was playing his games with Houdini! People are shocked!

I guarantee you that 99.9% of people and chess players would consider this person one of the greatest players ever. I truly believe that our "human interface" blinds us so much to the beauty and usefulness that our chess engines offer us.

How to define perfection? Right now, 1. e4 and 1. a3 is a draw (assuming PERFECT PLAY). How can 1. e4 then be better than 1. a3? Yes, because it has better chances for getting an advantage against NON-PERFECT PLAY.

Lets say we have a computer which can find all the perfect moves. Lets call it PC (Perfect Computer) for rest of this message. If it finds all perfect moves, It can't find the diffrence between 1. e4 and 1. a3, since they are both draws. So, for winning the games, it will have to choose between moves together with another engine. However, what engine is best for finding moves which generates winning chances?

Of course, PC could say: If I play 32. Ne4, black has 11 losing moves and 10 drawing moves, while 32. Kf2 black has 12 losing moves and 9 drawing moves, therefor 32. Kf2 is best. However, if you ask Houdini, it will say that 32. Ne4 is best, since it wins a pawn. Black has compensation, but not enough for equality. But with PC, it will be a draw.

So, as you see, we will need to combine PC and another chess computer. But which one? In a complicated, positional position, computers might have different evaluations on the moves, even with deep analysis, since they evaluate things such as pawn structure, initiative and bishop pair diffrent. I don't know anything about computers style, but when Houdini says white have a slight advantage in one position, Rybka could prove equality in the same position, because of the pawn structure.

So, chess (luckily) has no single answer. I am sure that Carlsen and Anand can discuss if white has advantage in a complicated position, since they evaluate the position diffrently. That is good, since else else chess would be all about calculating.

How do you learn winning if you only play against an engine? We win because our opponent makes a mistake, we notice it and expoit the created weakness. How do you train this against an engine if you play a full game???

Playing like a computer would be good, if it were possible for a human to do so (with one notable exception...).

@ViktorHNielsen

Thank you for your input.

I'm not using the word "perfect" here in the absolute sense. There are other discussions on this site about what "perfect play" in chess means, but that's not what this thread is about. I'm more interested in why or why not computers make excellent training tools and whether their style should be imitated and/or their games regarded as highly as great masters' games.

I'm not interested in speculations about solving chess on this forum.

Between you and me I'll go even farther and say I'm afraid to face even the computer much of the time even though it has no awareness.  I'm afraid of finding out just how good I'm not.  But knowing where we stand tells us what we need to study in order to improve so it's a real psychological catch-22.

I don't understand how one can be afraid of playing the computer.
Since it is expected to lose, there is no shame in it at all. Of course it still makes a difference how you lose, but that's the same against humans.

For me the biggest psychological difference is if I lose due to a blunder or due to slowly being outplayed.

At my level blatant blunders still decide 90% of all games.
And I think playing against an opponenent that is guaranteed to exploit those will certainly help in identifying and avoiding blunders. 

Well, like I said this pops up often enough.  It's not the people on this site that have a problem with it.  It's the coaches, authors, and general advice I've heard over the last 10 years.  As recently as this year in the Candidates tournament Gelfand giving mini speeches about how training with a computer stunts your development.

We're seeing more incredibly strong players because information is widely available.  In the 50s Fischer taught himself Russian so he could read the best chess books and spend all day at chess club reviewing old games.  Today online resources (including human opponents) means earlier and easily available exposure to information leading to stronger players.

This is actually the problem though.  They often don't have the slightest positional idea at all.  In the middle game they can strike upon the correct plan if you give them a lot of time on a good computer.  Mostly their great play comes from flawless tactics.  This is why opening books and tablebases are used to help them through the other phases of the game.

In any case, you'll never learn good positional chess by sparing with a computer.  If you did adopt its style, you'd have terrible results because you can't back up counter-intuitive moves with flawless calculation to prove it works in the end.

And in other cases, their move is simply wrong.  Although like I said on a good computer with a strong player helping it, you can avoid the more obvious errors (important to ICCF players for sure, where computer assistance is allowed in games).

They find more and more "human" moves, you're right.  But they've never played like a strong human "should play."  Their play is backed up with a mountain of calculation so they play better and at the same time impractical chess compared to humans.  i.e. it works for them, but only for them.

This already happened, or I should say is currently happening.  The guy's name is Borisav Ivanov and it was obvious he was using Houdini right away.  It was obvious because there is a stark contrast between human and computer play.  Humans play with a plan and overall logic.  Computers make moves based only on calculation.  So when he won with counter-intuitive and downright bad looking moves that under analysis had no flaw provided you search as deep and wide as a computer, it was obvious to all professional and strong amateur players.  (Also his use of an opening book, and other huge clues).

An anecdote, but i've played against a few players that played mostly computer opponents. It's as Wafflemaster said, their games were passive. They either sat back building a defense against an imaginary attack, or fell to a quick strike.

Yuck, really?

I mean, the process is really neat, from a technological standpoint, but to me the games are not.

@waffllemaster

I really enjoyed thinking about most of what you said.

There is one point that's a bit grey for me, though. Being able to distinguish top-level human play from top-level engine play is a gap that is closing. 

Top level human play has largely become about squeezing minute advantages from your opponent where the innacurate moves are subtly exploited usually for long term strategic gains. I find it fascinating that engines appear to do largely the same thing (although much of the time making moves that seem unintelligible in positionally equal circumstances) but humans go about the process completely differently. I'm awed to see that great players can fathom the most ambiguous positions and sense the right direction to steer the game. This faculty to me is very mysterious and so different the computational way.

I guess I believe that chess should be played without taking risks that are based on one's opponent making inaccurate moves either immediately or down the line. Surely errors, however tiny, will occur, and the goal is to pounce when the moment is right. The chess engine seems to be nothing more than a mirror that points out one's flaws.

I also don't quite agree with the fact that too much computer play will make a player excessively passive. Surely, it's important play a wide variety of players, to get a taste of style in chess, and know in what mode to play and when. Also, with these arguments, I just can't reconcile them with the fact that computers beat excellent players all the time! Just as the games of great masters are worth imitating because they win so the same should apply to the world of the chess engine. 

As Fishcher said, "I don't believe in psychology. I believe in good moves."

Thanks for the input, everyone!

Keep on playing against the computer. Analyze only your mistakes. Don't waste time analyzing Hiarcs move as it's almost perfect. The beauty of sparring with engines is you can see how weak , bad and very bad moves are punished. When your tactical skills improves, you can move on studying endgames, openings and positional plays from books.

(Yes, i'm quoting myself.)

I should add that it applies more to lower rated players. As a player develops and understands chess better, the less playing a chess engine too much would have a negative impact.

I think one possible reason might be because they learn early on that attacking a computer opponent too soon leads to a quicker loss.

If one wants to make good use of chess engines as sparring partner in preparation against human opponents, play your usual openings and defense  like you play against humans. Adopting anti-computer strategy would only ruin you.

Pretty sure I get 30-40% match with an engine... at least against weak players...  and I sure haven't been catapulted into a national / professional level.

But maybe I'm overestimatnig my match up.

That's why in a later post, i added that the better the player, the less it would negatively affect that player. Better players are more developed, leading to less of a chance of learning improper style, and have more knowledge, allowing them to take advantage of a better analysis of the game.