Engines play chess in a way reminiscent of the 1s and 0s that make up binary code. But 1 is boundless and zero is infinite. 1s and 0s are all you need.
How good do chess engines play chess?

Oneday human will defeat computer, if engines are not aided with book moves.
Most often engines do misevaluation in fortress endgame. Google it, you will find the truth.
KR vs K'N' endgame has result after too many moves but not in 50.
The fact says chess is much more deeper than engines think. Even a very slightest mistake will end in a result , if we play accurately like endgame from early middle game.
Was it possible playing accurately ?
The answer is "Not today but tommorrow", " not by ordinary but by extra-ordinary", "not daily but suddenly".

Somebody wanted a metaphor. Here's one:
Playing chess is a bit like doing your taxes. The rules are, if not exactly simple, at least completely spelled out. Most people can learn to be basically competent, but most never learn (or want) to get the absolute best performance especially in the more complicated scenarios that sometimes turn up. (Multiple deductions, trusts, non-profits, inheritance,...) Rare, freakish humans are really good at it, but computers are even better - they don't get bored or tired or forgetful or emotional.
Now imagine that the method for calculating certain deductions is so cumbersome that even a computer can't complete it with perfect accuracy within a human lifetime, so if the computer encounters such a situation, it has to just calculate a rough approximation and move on. So computers still have room to improve. There's even a small number of situations where the best humans can find an exact answer even though the computer doesn't know how. But those situations are too rare to matter. If you want the best results available, use a computer.

If you want to think like machine , you can never be perfect even in a simple position. Deep tactic will be appeared as overwhelming to you.
Give me a certain position, i will provide you an open raw scientific analysis.
Then you will understand, before that you do not use your brain.

Oneday human will defeat computer, if engines are not aided with book moves.
Stockfish (for example) does not need opening book to defeat the best chess-players. This is a fact. The days when humans won agains computers are well into the past, not into the future. Let's get real.

@Smyslovfan: What I meant to say is that there will be still a lot of improvement for chess engines. Every level of chess has its own theoretical moves. A lot of theoretical lines are actually too complex to handle for a chess player of my strength. The same applies to Magnus Carlsen - otherwise would he not lose the majority of the games from a chess engine. And the same applies to the current chess engines in comparison to the future chess engines. I have read in a thread of ponz111 about the question if chess is a draw that the maximum rating in chess would be 5000. That implies that the current chess engines can improve another 1700 points - which is more or less equal to the difference between the current level of chess engines and my strength of playing chess. :-)

Oneday human will defeat computer, if engines are not aided with book moves.
Stockfish (for example) does not need opening book to defeat the best chess-players. This is a fact. The days when humans won agains computers are well into the past, not into the future. Let's get real.
Play against me with your best engine. I bet you can not win. you may think that i will take help of another engine. To ensure that i am not using engine I will explain my each plan detailed for a single move.

Earth, I'll play you a 12 game bullet match. That should be enough games that if the winner uses an engine, they will get booted from the site. Why bullet? I don't really have time for a slower time control.
I'll see you in live now if you're interested.

Earth, you lose more games than you win against humans in blitz!
I think you are not intellectually disabled.How can you think that i will play blitz and will explain my plan simultaneously.
I will play correspondence.

My assurance that you won't cheat is that a) the time control is very fast and b) if we play 12 games, the site will shut down the person who uses an engine. That's good enough for me.

I'm amazed at these people who make empty challenges, but when faced with a real one, run away.
EDIT: ADDED:
I challenged you to a bullet match in part because I saw that you do play bullet chess. You've played more than 700 bullet games.
This is an interesting topic.
I haven't yet heard a good metaphor for the way engines play chess. It would be nice to have something to compare it to, to explain it to non-chessplayers. Suggestions?
We can say that it adheres to a sophisticated set of rules that provide a very close approximation but not perfection, and then explain that the engine is still unbeatable because execution is always perfect within these rules and the engines limited but sufficient field of vision. It's boring though, people lose interest. I'd like to be able to say something about how engines play chess with an elegant metaphor.