Try To Solve These Annoying Puzzle Rush Puzzles (mostly easy)

So this is a whiny topic, but at least you get to solve some puzzles while I whine 

The length of the solution and difficulty of puzzles in puzzle rush are not consistent. Sometimes this makes the puzzle less instructive.

Let's start with length.

I understand how this happens... the engine that mines these tactics assumes the best move by the opponent. After this, if you have only 1 move to keep the advantage, then that move will be part of the puzzle.

So you end up with puzzles like this that go on and on.

-

-

But when is this not consistent? And how does this cause some of them to not be instructive?

Lets look at another puzzle:

In the previous puzzle forking the king and undefended rook could have been the end, but we're forced to capture two undefended rooks.

In this puzzle the knight is not a free piece. In fact you're sacrificing the exchange, so it should require a followup. However the engine knows that black's best move is to castle and then lots of white's moves maintain the advantage, so that's not part of the puzzle.
(If you're curious, if pawn takes rook then Qxf7 is mate)

-

-

Now for difficulty. The next two puzzles are rated 700

Ok, fork two undefended pieces. A very reasonable 700 rated puzzle.

Next

-

-

An elegant minor piece mating net constructed on the side of the board, and it's a mate in 5. Plus there are many sidelines to notice, starting with Kg1 in the beginning fails to Rf1 mate.

So why is this puzzle rated 700? How is it anything like the previous rook forks two bishops puzzle? Because the rating is set by who solves it. When beginners try to solve this puzzle they have no idea what to do, and just play check. It goes check, check, check, oops, the puzzle is over.

When a better player gets a puzzle like this in puzzle rush, they start to calculate and they can't see the end, so they make a guess. If you don't guess like a beginner then you might fail the puzzle.

---

These examples are not the best I've come across. These 4 were collected from only 2 puzzle rush attempts. IIRC I solved these, no problem.

But sometimes I'm short on time, and I'm forced to capture a bunch of free pieces for no reason.

Sometimes it's a low rated puzzle, but because I calculate I lose time, or even fail the puzzle.

Sure there are strategies around this (play mindless checks on low rated puzzles without calculating and 9/10 times you'll get it right), but strategies like this have nothing to do with being good at chess, and even when solving puzzle in an untimed mode, not including the logical continuation (pawn takes rook, queen takes f7) is not instructive for players who are trying to learn.

---

One solution is to allow people to flag puzzles for having extraneous moves, short continuations, and low rating due to beginners guessing. Then chess.com could, at their leisure, select a certain percentage of the most reported puzzles to fix.

Is there a good cost benefit ratio? Probably not. Probably it's better to come out with the next hot app. Popularity trumps quality, at least in the short term. But I hope it's something they consider.

Thx

I agree with your diagnosis on the 4th diagram but not on the 2nd one. The puzzle engine does not always play the "best" defense since it knows the position is lost anyway. It is quite possible that it will recapture the rook and permit you to checkmate on f7. Sometimes it makes you prove your point, and sometimes it does not.

The best way to fight the discrepancies in puzzle rating is to only play puzzle rush survival. With infinite time it's up to you to allocate the required amount of time to each puzzle whatever its rating.

The point is there's  a large difference between the 1st and 2nd puzzle.

Sure, the 2nd puzzle by itself is fine.

It doesn't "know" anything at all. It juts mindlessly calculates. If you think sometimes it asks you to "prove your point" you're deluding yourself. Chess.com is a business. It's not going to waste time making quality puzzles. Sure sometimes the engine line happens to also be the human "prove your point" line. But that's just a coincidence.

Exactly. No sane author would include a garbage puzzle like#4 in the OP.

#4 took me ~20 seconds.

Ok, I also solved #4 during a puzzle rush.

It's still a garbage puzzle.

I also occasionally get low rated puzzles that require significant amounts of calculation.

It is hard to discuss with someone who contradicts himself 180 degrees. You wrote in your post:

However the engine knows that black's best move is to castle and then lots of white's moves maintain the advantage, so that's not part of the puzzle.

You gotta make up your mind. Does the engine know or does it not know?

i think the point is that some puzzles go on with what are obvious moves and others don't. the first two puzzles are fine examples (i think) of the disparity of when it is solved.

Any point multiplied by a contradiction equals zero (Plato's 3rd law).

ok then. i don't claim to think like hobbes or bacon, so these are my words-

op makes a good point that the determination of when a puzzle is solved seems odd. 

Engines can know moves because they calculate. I say the engine knows ...0-0

Engines can't know concepts like "black is lost, so I won't play the best move and ask the human to prove his point"

In any case I was using the word loosely. You can attack my semantics but not the logic. The person I corrected was wrong regardless of the word he used.

Don't be too proud of yourself. Semantic quibbling is easy to do and adds nothing to the discussion.

The first two were about inconsistent length, yeah.

The last two were about inconsistent rating / quality.

Forking two undefended pieces is a normal low rated puzzle. Some nebulous check check check with no pattern is not only not instructive, it's only rated so low because beginners can consistently guess the correct moves.

I'm glad you have been assigned the task to decide what an engine can or cannot know. It is senseless to state that the engine knows 0-0, because  you might as well have stated that the engine knows h7-h6. The whole point of knowing 0-0 is it knows it is a (relatively) good move but not good enough to save the game. Had it believed that 0-0 would save the game, it would have rejected the whole puzzle because then white wouldn't be winning.

Like all the knowledge we discussed, it comes from the program designers and not from any independent discovery by the engine. The designer team has decided that a certain + score relates to a win and it has decided that it will sometimes ask you to prove your point and sometimes not. It's rather random. In this example, the randomizer could just as easily have captured the knight and let you checkmate on f7. Since all other moves lose as well - including 0-0 - it then rather tests your alertness to see the checkmate than take refuge in another losing continuation.

Concretely, sacrificing an exchange does not require a follow-up for the simple reason that no exchange is sacrificed. Only a knight is captured. If the engine had decided to recapture the rook then indeed an exchange had been sacrificed and a justification would have to be provided by handing out the checkmate. Both roads are open and are occasionally travelled by the engine.

I'm of course talking about it from a student's perspective. A new player will be reluctant to play RxN because it "loses an exchange." Of course no strong player would call that an exchange sac, they'd call it a free knight.

You say the programmers use a randomizer, but I've never encountered a puzzle that doesn't follow the rather obvious rules I mentioned earlier. If you have only one move to maintain the advantage (lets say only 1 move is +2), then it's included in the puzzle. If multiple moves are +2 then the solution ends.

You claim sometimes a puzzle, which was mined out of a database by an engine, will have a solution with a move that is not considered best? Show me such a puzzle.

I have done a lot of puzzle rushes recently and these are my experiences (besides my logic). The engine does not play a move below the appropriate result category (win lose draw) ,but it might play varying moves within that category, i.e. one that mates or another one with a +5 score. It might even have an algorithm (I don't know) to select a response that is difficult for the solver, even though it is objectively not the most stubborn defense. I'll be back with you with an example(s). If not, I have failed and #hidingathome.

You solve more puzzles than me, so you might know better than me.

But I don't trust people's interpretation of their experience (I don't trust mine either). I think sometimes we think a move is not best because it looks dumb, but the engine sees some absurd series of moves that make alternatives worse.

That's why I ask for an example.

You are spot on with this llama44, but how would you fix it? You could allow users to flag things like "needs more moves", "needs fewer moves" or "needs alternative move" . The problem is that then chess.com would have to spend time acting on this info and modifying the problems - and as you pointed out, they're running a business and their time costs money. Perhaps a group of volunteer chess.com members could do this work.