A new kind of puzzles: no tactics, only avoiding blunders - what do you think?

NOTE: If you just want to solve the puzzles, scroll down a bit, there is a link to the study at the bottom.

Methodology

It is fairly simple. I take a position, I run it through all versions of Maia, look at the moves Maia makes, assume those are the moves that a human at that level would make, and filter out all cases without a blunder.

The estimated elo of the puzzle is the lower bound of elo of the lowest Maia that doesn't blunder (so if the blunder is made by Maia1, Maia2 and Maia3, but Maia4 makes a way better move, then the estimated elo is 1400, and your job is to find a move at least as good as Maia4 makes). (Note that all elos are Lichess-based).

Since I'm using Maia engine and there is nothing below Maia1 and nothing above Maia9, currently all the puzzles are roughly between 1000-2000. I do have some ideas how to go lower and higher, but for now, those are the ones I've got.

The why

I use both Chess.com and Lichess, and I noticed how chess.com has those Brilliant moves while Lichess only shows mistakes. Eventually that made me realize that assuming we play against a machine or Magnus, all of our moves are either the correct ones, or blunders. Yes, of course it's too simplistic of an approach. First of all, Wayward Queen Attack works against many beginner chess players. Humans are not machines, they have blind spots, you can trick them. However... can you *really* trick Magnus?

And especially at the lower levels, players very often lose because they blundered - their queens, their rooks, their smaller pieces. At higher levels big blunders are less common but when they do happen, they are ruthlessly exploited. Obviously, decreasing one's rate or severity of blunders would allow their strategic genius to shine brighter, and therefore increase the elo.

That's my line of thought, at least.

Maia engine's website actually mentions "detecting blunders" and even has a web app to compare different Maias' responses to the same position, however it's very limited, it only shows a handful of positions: https://csslab.github.io/Maia-Agreement-Visualizer/ .

I wanted to put it into puzzles because I think it would be useful for me to learn from, and I love puzzles - but I did notice that whenever I don't know how to solve something and have to guess, they conditioned me to try to put my queen in the most dangerous position on the board (not... always, but... often enough), and I don't think it's the best strategy for a full game. Those puzzles are supposed to be different: to teach me to be more careful, to look into the positions deeper, to not rush, because the whole point of is don't do the stupid thing I know you want to do.

Current state

I've only made a few puzzles for now. There is no application that shows the puzzle correctly (for example Lichess only allows one-solution studies, which is understandable, but I want it to work a bit differently).

I am writing a small application to be able to play them (Windows, Linux and Android, nowadays it's so easy to make programs for multiple platforms). If you are curious, or like the puzzles, or want to see how it all turns out, you can follow my account. I'll be writing more blog posts about it in the future.

Example puzzles in a study (1100-1900)

Please do read the instruction on the first chapter, or here.

As I said, there are no application that I know of that would show this kind of puzzle correctly (but please point me to it if you know of any, or if you know how to set up Lichess Study to do it), so you gotta do a bit more work than you'd usually do with a Lichess Study.

EDIT: Good news, I made the first version of the app!

It's located here: https://adventure-dev.itch.io/blunder-annihilator-chess-puzzles-demo and you can play in your browser, as long as it's Chrome or Chromium-based and fairly up to date. Just select the "Special puzzles" from the main menu. It will be 7 puzzles, not 9, but should be good enough to check out. The Lichess Study still exists, linked below, of course, but it's more comfortable with a program that is actually made for that. Though, it still lacks other features...

There are multiple good answers, but only two-three are marked:

• the one marked "solution" is the one expected from a player at 2000 elo,
• the Stockfish move, marked as "Great" and with a comment that it's a Stockfish move (often the same as the "solution"),
• the incorrect move expected from a player just below the elo for the puzzle.

Make your move. Then turn on the Analysis board.

• If the move you made is as good or better than the move marked "solution", you solved the puzzle.
• If the move you made is worse than the "solution", you didn't solve it.

Note: in one of the chapters Stockfish at first shows the lower elo Maia as making the best move, but after I made it go deeper, it changed it's mind and agreed with the higher elo Maia.

Please, please, if you do the Study, tell me your thoughts, and your results. I really want to know how good (or bad) those puzzles are. Are they boring? Do they look useful? Is the elo accurate? (I have my doubts about the 1900 one).