How many rating points is one "level of performance"?

I'm sure it varies a lot, but roughly speaking what difference of ratings points means an opponent should be able to beat me quite easily?

I've had a few easy victories over people 2-400 points above me recently - like really easy, I felt bad - but also I've lost really hard-fought games to people the same level as me.

Any rough guide? Or at my level, are blunders so common that anyone 1800 or under can be considered a fair challenge?

Very interesting, thankyou.

At what sort of rating do people stop making stupid errors even when having a bad game - you know not noticing they left their rook/queen undefended, that kind of major blunder not just failing to spot the best move?

@MisterBoy My current online rating is 1500+, so if we played (online) I would expect to beat you... but my expectation might be disappointed!

And I definitely wouldn't expect it to be easy! 

I play most of my chess at a different site, I've never beaten anybody rated much more than 400 points higher (and more than 300+ only once).

I have lost to people 500ish points lower though... I assume that the better a person becomes the more consistently they play.

I don't think that I've ever lost to anyone 600 points lower though - but most of my opponents tend to be within 200 points in either direction.

Aww, you're not accepting challenges :)

Kramnik overlooked the threat of a one-move mate when playing Deep Fritz.

Like everything else, it's a numbers game. GMs drop pieces too, but very, very rarely.

OK I'll rephrase, at what rating do players, even on a bad day, become very unlikely to make these kind of mega-blunders? 

I would really like to see a large sample of class player games analyzed for  blunders and correlated with chess rating.

From my experience bb_gum's numbers are too generous. I think 1600 players miss simple 1-2 move tactics much more often than in 1 in 100 games.

But the idea is sound that the higher one's rating the less likely to miss simple tactics. It would be great to fill in some numbers based on real data.

within 100 is a pretty even, fair game. 200 points should start to win more often than not, but is still a fair game.  300+ is rarely winnable by the lower player.   There might be exceptions to this at the top and bottom levels of play (below 1000, for example, is basically a free for all and either party can win no matter who is rated higher) and not so sure that above 2000 follows the same generalizaton or not. 

At 1300-1500 you have an interesting mix of abilities.  Some players have favorite openings that they know cold, but are weak in the end or middle game.  Some are good at positional play but miss tactics, some vice versa.    

A 1300 player is already above average on this site.  One-move blunders (hanging piece, simple fork/pin) are unusual but still happen.  Losing a piece (not a sacrifice, I mean a blunder) doesn't necessarily decide the game, as the other player may still commit an equal or worse blunder if you create some complications.

As you go up your opponents will become less forgiving of your mistakes.  By 1500 you'd better never drop pieces, you'd better avoid well-known opening traps, you'd better know some endgames, and you'd better have a grasp of positional play.  I'll let you know what 1600 takes when I get there.

All the numbers are distored by cheats.

As a result, it's very easy to have your various on-site ratings shift (rapidly) by @200 points in either direction.  That's about a 400+ point framework to cycle through, which is much too large to be accounted for by "normal rated play."

Unless of course you're effectively "drunk" for many of your games.

Then All Bets Are Off.

I think you way over-estimate here.   Ive had 16-1700s drop a piece.  I lost a game in the opening this month, one moment of carelessness is all it takes.   I beat a guy in a "start from here" opening game in like 5 moves ... both of us close to 1800.   Keep in mind that is only 2 examples but they were days/move play and 2 of less than 10 games played to completion over the past month.  In 2 more the guy played like a 1400 or lower but was rated mid 17s.   There is no real generalization on blunders vs rating that I have seen yet apart from higher = less, but it sure never falls to zero (at least not for the under 2ks).

In the USCF, ratings classes constited 200 rating points.

2400 and up was a Senior Master

2200-2399 was a Master

2000-2199 was an Expert

1800-1999 was Class A

1600-1799 was Class B ... and so on.