Is there any chance that a 1300 rated player can beat a 2700 rated player?

I once read some place that perfect play was estimated at 3600. Meaning that a top 2800 should get a draw once in a 50 games, a win once in 100, or any reasonable combination thereof...

You think about someone like Nakamura, Anand, Magnus, the other guys up there on top - maybe also Capablanca, Alekhine, Fischer, Kasparov, Karpov, up there along the best of them... you say - hell, why not?

It is my impression that chess improvement goes like an inverted pyramid - you need to learn very little to get from 500 to 600. More considerable effort and difference in understanding goes on the road from 1200 to 1300. Every 1800 knows that it's hell to get to 1900, and for an 2300 to go up to 2400 - well, it can be the known impossibility of a lifetime (actually this applies for different people in different levels). 

Now, what is perfect play? There can actually be levels within perfect play which won't translate to elo - like, these two 3600's playing hundreds and hundreds of very interesting and highly exciting games, full of multi-layer manouvers, impossible to fully explain to anybody below 3100 or so... of course they draw every time, as they are so perfect (they do manage a 64% score against the less fortunate 3500's, of course), and of course some kibitzer here would call them 'drawnik' (except when playing against 3400 patzers or so) and 'void of fighting spirit' etc., but - 

the amazing thing is that chess is so intricate and with so much of a margin for an error, that it would turn out that 3600 player A always gets 'the better side of a draw' then 3600 player B - he's really more skillful... he's always the one stalemating when a pawn up, or ending up the exchange for a pawn in an impossible-to-win endgame where only he has winning chances, or gets to create that winning endgame up a pawn whereby his opponent will need to sacrifice a piece to get to K+B+wrong bishop pawn against bare K... 

many different possible levels which will not translate into Elo, the Elo system not being sensitive enough to such fine differential of skill between two perfect, incapable-of-losing players. Perfect and 'more perfect' if you want. 

"Or to give a practical example : what would be the chance of a strong GM to at least draw a match against a 'near perfect player', if we assume every draw counts as a win for the human ?"

Thanks for complimenting my train of thought :-) I'll try to answer your question - so of course, 'near perfect' - depends HOW near to perfection :-) Near is subjective, and I had my share of people staring at me in awe, believing I represent personally chess perfection (happened to you too, I'm sure...). "This guy just sees everything!!" Yeah, right... 

But if perfect chess is 3600, then a 3400 is expected to draw the match under the conditions you stipulating (supposed to get 25% against someone 200 points up).

"what would be the chance of a strong GM to at least draw a match against a 'near perfect player', if we assume every draw counts as a win for the human ?"

If we could answer that question, hicetnunc, we would know how to play perfectly.

This is actually a mathematics question, but not for 10th graders...

The statement that it "won't ever happen" will probably not be contradicted; nonetheless, I think it's more accurate to say that it is extremely unlikely to happen, and that it would be foolish to expect it to over the course of your life (in other words, any sort of optimism about this would be unwise).

But if we are talking in practical terms, then of course we can think about questions such as "how valid is my 1300 rating(if you haven't played in a tournament in a while, who knows how much higher it has become)," "is my opponent feeling well today," etc, although this most likely is a very long way from making up for 1400 points, even if it's just on paper.

Chess and probability are joinded at the hip?

What planet (or alternate universe) do you hail from?  Lots of people in this thread are from the "What If" universe, even more from the "So What" universe.

But we need more players from the "WTF" universe, just to balance things out.   

It is impossible- it will NEVER happen.

Coming Soon to a Theater near you.

What the hell, this thing is into 400 posts?  20 pages of peple repeating themselves??

"No, it will never happen"

"lol, maybe"

"there's a chance, but not likely"

"Zero chance, at least practically"

"Infinite universes something something"

The trivial nature of the topic is what makes it so enjoyable.

It's so fun to annoy people in the practical world with infinity arguments . Makes them go crazy.

Anyway, your post is spot on of where the debate has been, and it has basically been a never-ending cycle of those things. Haha, infinite universes...

Ouch hicetnunc! I'm merely doing a masters degree in physics. Would you care to present a solution to your question?

"I once read some place that perfect play was estimated at 3600" (solskytz)

I'd be interested how you would define "perfect play". It certainly can't be "winning by force" because you can't do that in chess, chess is a draw. Would it be "playing any moves which never allow the opponent the opportunity to reach a winning position"? Possibly, but there would be quite a large range of different moves within that definition, and you'd have to find some way of defining non-losing move A as better than non-losing move B.

I remember discussing the same thing in a thread about computer chess and I suggested a computer with "solved" chess would be hit with this problem. At first I thought it could pick the move which gives the opponent the best chance of going wrong, i.e. the highest fraction of the opponent's possible replies are losing moves. But then I realised this isn't much good because it would make the computer swap pieces off aimlessly, because that gives the opponent a wide selection of losing moves (everything but the obvious recapture!). So it wouldn't be too hard to draw with probably.

Just curious how you would define "perfect play" because I've never got my head around it.

To the probabilist team:

A random move generator of 1300 level playing a "perfect" game against a 2700 GM... is the argument.

The more move, the less likelly (probable) that the random move generator would continue to play a perfect game.

Hence the GM just need to "lengthen" the game (repeat non commiting move) and ultimatelly he wins, because the random move generator re-start to play at 1300 level. cqfd

So even with the probability theory, 1300 can not beat 2700 GM.

What do you say to that? :)

<Madhacker> yes - it looks like you've got the concept right :-) the guy won't ever be capable of losing games.

They will need to create a game inside the game to keep on playing - like the guy who remains with most material, or the guy who makes HIS OPPONENT force a perpetual, or the guy who stalemates his opponent... otherwise these guys will just lose all interest in the game (unless it is to impress the poor masses rated below 3600)

<jclheriteau> indeed, I've seen strong (2400ish IMs) players do that in blitz all the time - play retreating moves, do stuff that is kind of meaningless, but very quickly - and now I finally understand what it is they were doing :-) thanks for clarifying :-) as it was always kind of a mystery to me - they are giving their 'weak' (like, 2100ish) opponent more opportunities go to wrong, to get stressed out by their speed and perceived depth of play and understanding... great technique. They were simply not being serious and letting the other guy defeat himself, with no effort on their part...

1300s can expect to win 0.000003467892% of their games against 2700s because that is how often heart attacks kill one player during competition.

Ah, but Ziryab, there are other ways of a player dying during the game, and they are not all neccessarily accidental!

solskytz, If you are interested the best solution I've ever thought of is to use a present-day Fritz-style engine, but combine it with a "lost-position-check" against a database of all lost positions. That way you would have a computer which was "perfect" in the sense that it would never lose, but also played good moves consistently.