# Humans against computers

Are human ratings also progressing, could humans catch up with computers? (maybe playing against strong computers makes us strong)

Are team of humans able to defeat computers? Is any competition like this going on?

• a human alone can't possibly win against a computer anymore (even with 2 pawn advantage and white) is never going to happen, not anymore... But collectively maybe we stand a chance if the time control is really small (like no more than 2 minutes per move) Oct 18, 2013 at 0:37
• I thought small time control was good for computers, not the other way round. Oct 18, 2013 at 1:23
• if you let a super computer finish brute force every line chances are we stand zero chances, but if we collectively look into many lines maybe we can outsource computers, but again, in small time control Oct 18, 2013 at 2:43
• @ajax333221, I don't believe you are right at this one. Computers can't analyze any ramification of a position. Chess has not been completely solved. But, that's the topic of another question. Oct 18, 2013 at 22:12

## 4 Answers

The best computers are much better than the best humans, and if I had to make up numbers I'd guess that they're improving at least ten times as fast as humans are (it's a wild approximation but, more or less, humans are improving by around 2 Elo points a year and computers are improving by around 20 points a year). There's no way humans will catch up.

• and I believe as in any sport, there is a curve that we will get stuck eventually, and records will not be improved unless by very small difference Oct 18, 2013 at 23:58

Humans will never again be able to defeat computers at chess. We're finding the limits of humans while computers and algorithms are only getting better and better.

Generally speaking, humans have already lost the battle against computers. Tablebases have "guaranteed" optimal play by computers in the endgame. Opening books have removed much of the computational burden in the opening. (It's MUCH easier to generate move or position lists for the first ten moves and index them than it is to compute them on the fly.) Computer hardware continues to get better.

Any situation that involves straight memorization is one that humans are going to lose to computers. With the opening and endgame "thoroughly optimized", the middlegame is the only place that humans really have a chance. Computers are deterministic. Compare these two situations with White to move:

``````[FEN "rn4rk/pp4pp/2bN4/q7/8/8/Q5PP/2BR3K w - - 5 10"]
``````

And

``````[FEN "6rk/6pp/3N4/8/8/8/6PP/7K w - - 5 10"]
``````

Ignore for a moment how contrived they are. In both cases, the correct move is the same, `Nf7#`. However, even though a computer will pick up on the correct move, these positions are totally different from a computer's perspective. A human will look at this and recognize intuitively that the left half of the board is totally superfluous. A computer will have to compute that the positions are functionally equivalent.

However, as storage capacity increases, the number of these positions that aren't indexed slowly gets smaller. With the pasing of time, more and more of the middlegame approaches determinicity.

Consider also that it is much easier for computers to "learn" than for humans. It is trival (for given values of "trivial") to plug chess libraries in and give a computer access to all the accumulated knowledge therein. Humans cannot do that. We have to study for years to impress that upon ourselves.

TL;DR: As chess moves more toward exhaustive electronic storage and indexing of positions, humans will continue falling behind computers. Unless there is a breakthrough in our understanding of how memory and the brain work, there will be no reversing this trend.

• allowing computers to use any stored information while forbidding it to humans in a match is unfair, but well, since humans have some sort of small stored openings it would be unfair to computers too, so I think fairness is a bit hard to get, unless we cut part of the brains of a chess player so he need to recalculate the opening lol Oct 19, 2013 at 0:04
• @ajax333221: I'm not sure what you mean. The main reason computers are as dominant as they are is their ability to index positions. Without that, it's just a matter of throwing processing power at it, and that isn't efficient. Oct 20, 2013 at 14:20
• Storage hasn't been the driving factor in ratings improvement of modern chess engines... in fact is has been effective aggressive pruning techniques and search ordering which has driven improvement. If you had computers play against the best grandmasters in Chess 960 (no opening books, random starting positions) I bet computers would wipe the floor with the GMs, much worse than in traditional chess Dec 17, 2013 at 23:43

I think that the whole theme of human against machine is unfair. After all, the computers use excellent algorithms and are trained on games played by human grandmasters and hooked up to high quality opening databases and endgame tables. An engine has a lot of human knowledge at the tip of its fingers and can access everything in a matter of seconds. From this point of view, chess engines are a combined success of human chess know-how and progress in computer science.

Today, the question of whether a human can beat a chess engine seems equivalent to asking if a boxer can beat a Terminator machine. I seem to recall that Kasparov said that humans can beat a chess engine, if the humans are given unlimited thinking time. It could be true but it has not been tested as far as I know. Considering that home computer driven engines reach Elo 3100, I don't see a human versus machine tournament any time soon.