3 piece endgame puzzle rated 2600

I just ran across a puzzle that seems to have an unusually high rating. It's K vs. KR endgame puzzle rated at 2598 with a solve rate of 19.9%. Puzzle number 1335815.

Black to move.

kc3

ez

?

Direct link: https://www.chess.com/puzzles/problem/1335815/practice

After ka2, people probably rush and not think about checkmating using the a file as the back rank.

It is mate in 3 moves, and a rather trivial one.

So how in the world is this rated 2600? I've never seen anything like it. Isn't this a 1400 puzzle?

Yeah lol happens. I have no idea how does the site assign a rating to puzzles

probably ,cause its a new puzzle , they have usually a high rating ,when get throwed in the puzzlepool

I believe it's assigned by the ability (rating) of those attempting to solve the puzzle. So at 2600 presumably only fairly skilled people will be attempting it. The more people who solve it, the lower the rating goes. The fewer who solve it, the higher the rating goes. At a solve rate of 20%, not many people are getting it right. 

That's the part that I don't get. How could 1400s get this wrong, let alone 2600s. And it's not like it's a totally new puzzle. I understand if only 50 people have tried it. But over 400? That's starting to get a decent sample size. 

Ah makes sense . I guess most of those 400 people that attempted the puzzle are probably below 1000. The rating of that particular puzzle will eventually come down I guess as more people attempt it . 

Also I was told by someone that a puzzle at a particular rating is actually taken from an actual game of chess corresponding to the same rating . Eg a 2200 puzzle , is taken from a match between two 2200s. That doesn't make much sense I guess.

Exactly. Two 2200s (or 2600s) wouldn't have much trouble with this puzzle. But a 1000 player might. 

So I dont see how a 1000 would even see this puzzle. As far as I can tell, it's the 2400s and 2600s that are getting it wrong. Which also doesn't make sense. 

Dude 😅

Yep

As you probably know, puzzle ratings tend to be much higher than live ratings. For example, my rapid rating's been in the 1200-1300 range recently, while my puzzle rating has been in the 1900-2100 range.

I suspect that people are getting it wrong because they're playing what they would play in a game rather than trying to mate as efficiently as possible.

If you tell me "find the shortest mate" then I stare at the puzzle for a minute or two and come up with 1...Kc3, after which white has 3 possible moves: 2.Kc1, 2.Ka1 or 2.Ka2.

If 2.Kc1, then 2...Re1#.

If 2.Ka1, then 2...Kc2, and white is forced to play 3.Ka2, after which 3...Ra4#.

If 2.Ka2, then 2...Re1, and white is forced to play 3.Ka3, after which 3...Ra1#.

That's easy enough.

However, if you asked me what I'd actually play in a game, I'd go with the simple 1...Re2, confining the white K to the first rank. Then I'd use the usual K+R vs. K mating procedure, making a waiting move with my R along the second rank when necessary. This would look something like: 1...Re2 2.Kc1 Rh2 3.Kb1 Kc3 4.Ka1 Kb3 5.Kb1 Rh1#.

It's a mate in 5 instead of in 3, but it seems much more natural to me based on how I think about K+R vs. K mates.

I understand that. And agree. But 2600s shouldn't be making that mistake. 2600s should be thinking not only what's the fasted mate, (and see it) but also what is this simple puzzle asking me to do. 

I've never seen  a 3 piece puzzle rated 2600. 

So a lot of people don't understand this, but the way the Chess.com algorithm is set up is that puzzles do not allow alternative solutions.

Puzzles are designed so that either:

1) Only one move wins, everything else either draws or loses. You must find the winning move. Or:

2) Only one move draws, everything else loses. You must find the move that draws.

So these kind of endgames (king and rook vs king) aren't supposed to exist, since many moves supposedly win, right?

Well, this is where the algorithm screwed up: it took this puzzle from a game where 47 moves had just been played with no captures, meaning the 50 move rule was about to happen (the game would end in a draw after 3 more moves). So the computer thinks that anything other than a mate in 3 or less will result in a draw here.

The computer isn't trying to ask you to find the fastest win: it's trying to ask you to find the ONLY win, since your 50 moves have been almost used up.

The problem is that the person solving the puzzle doesn't know this. They don't know that they're on move 47 of no captures and are about to reach the 50-move rule. So we're trying to solve this puzzle without full knowledge of what the puzzle actually is, which is bad, because chess is supposed to be a game of perfect information, where we're told everything about a position and the rules before it's presented to us.

So that's where things went wrong here.

^See the analysis board above. The computer believes that there's only one winning move for black here: 114... Kc3. Everything else is a draw.

Deranged, that makes sense. But even if we didn't know that, wouldn't it make sense to find the fastest mate anyway?

I understand many moves lead to mate. But in only a second or two it becomes obvious that "any" move that leads to mate isn't going to be the solution. 

That's where I'm confused. I get that in this particular case one and only one move avoids a draw, but it just so happens that one move is the fastest mate. I'm only a 1800 player. But isn't that fastest mate something a 2600 would see anyway? My point is, how do 2600s get this wrong even if they don't know there is a 50 move draw coming up?

Hahahahahahahahahahahahahahahahahahahahaha!