What are some unconventional chess problems that ordinary chess engines can never solve? I'm attaching two examples below so that you can get an idea. My students seem to love these puzzles. I'm researching and learning about it now.

Example 1


  • Thanks @ETD for pointing out that it's possible to create programs to solve, at least by brute force solutions to any chess problems. What I mean though is normal chess computers.

  • EDIT 2: I realize that the question is a bit vague. What I really like to know is creative puzzles, which if given to a beginner player should not be solvable by putting it on his computer and running an analysis.

  • 8
    In what sense do you mean that a computer can't solve these? I agree that standard chess engines cannot, but that's only because these aren't typical chess problems asking the solver to find the best move in some position, which is what chess engines have been built to do. But one could certainly create a program that solves these two examples, even just by brute force. E.g. for your first example, one need only check all combinations of first 4 moves for Black (and there's not too many of those) to see if one yields a legal sequence ending in mate when paired with those White moves.
    – ETD
    Commented Jan 10, 2016 at 18:08

6 Answers 6


I will do the obligatory retro and proof game plug:

Proof games

A type of problem where you are to reach the diagram position within a given number of moves.

A good introductory PDF by Alain Brobecker, featuring many "classic" problems that are well-known among the proof game community.

Jeff Coakley has a lot of shorter (but still tricky) 4-move proof games on chesscafe (although the site put them behind a paywall since 2014, the resourceful reader can probably figure something out...)

Another simple example (original by me): Reach the position after black's 4th move.

rnbqkbnr/1ppppppp/8/p7/3P4/8/PPP1PPPP/RNB1KBNR w - - 0 1

Shorter proof games of about 4 to 7 moves are pretty good for regular chess players, and you literally only need to know how the pieces move. Longer ones can go up to 30 moves (for a "normal" long proof game) or even over 180 moves (the longest I know of).

Retrograde analysis

Given only the diagram position and the fact that it was reached in a legal chess game, determine some feature of the position like side to move, castling rights, en passant rights, whether a piece is promoted...

An excellent collection of simpler problems that slowly rise in difficulty.

A magnificent introductory text by Nikolai Beluhov, an excellent composer of such problems. (Link is to a post containing PDF attachment.)

Another easy example by me again (although the idea is not new): What was the last move played? (I don't even need to give you the side to move for you to solve it.)

8/8/8/8/8/4PP2/1PPPRKPk/4nQ2 w - - 0 1
  • 1
    Thanks for this Remellion. Jeff Coakley is continuing to make new elementary and not so elementary problems which he posts on his fun website: coakleychess.com. He is also progressively uploading his old material from ChessCafe, presumably by agreement with them
    – Laska
    Commented Feb 20, 2019 at 14:06
  • Spoilers (rot13): 1.q3 n6 2.Dq2 n5 3.Du6 Au6 4.q4 At8 /// -1...Xu1kOu2
    – Evargalo
    Commented Jun 15, 2023 at 11:18

This is an excellent source for problems from different range of difficulty, where you can find these types of problems:

  • Selfmate
  • Helpmate
  • Reflexmate
  • Retrograde Analysis

Which regular chess engines can't solve because solutions are not the best moves. (Actually, worst moves in selfmate problems for example)


Here is a fun puzzle:

Construct a legal chess game where

  • white starts with 1. a3,
  • white delivers checkmate on move 5, and
  • the last move is made with the rook that started on a1.

Hint: Black takes the pawn on a3.

  • This is fun! Thanks for the hint! I'm still working on it, but found other interesting lines. 1. a3 e5 2. Nc3 Bxa3 3. Rxa3 Ke7 4. Ne4 Nh6 5. Ra5 Rf8 6. Rxe5# 1. a4 c6 2. Ra3 Qa5 3. Rc3 Qa4 4. Rc5 Qa1 5. Rc8
    – Arun J
    Commented Jan 13, 2016 at 18:17
  • 5
    Tada! Found! 1. a3 e5 2. Nc3 Bxa3 3. Ne4 Bf8 4. Ra5 Ke7 5. Rxe5# *
    – Arun J
    Commented Jan 13, 2016 at 18:18
  • Another version of that puzzle : puzzling.stackexchange.com/questions/73474/…
    – Evargalo
    Commented Jun 15, 2023 at 11:10

There is a branch of Problem Chess called Fairy Chess Problems with all kinds of stipulations.

However, there are specialised programs (e.g. Alybadix or Chloe/WinChloe) to solve fairy chess problems and nowadays problem composers routinely use them to verify the integrity of their creations. For more information, look at https://en.wikipedia.org/wiki/Software_for_handling_chess_problems


Try also Burt Hochberg's Chess Braintwisters, which often require "thinking outside the box" to a move that is logical but illegal (like promoting to a King) and thus would never be found by an ordinary chess-playing or problem-solving computer program. A couple of examples from that book have already been posted in this forum in reply to another question: Is there any position where not promoting(illegally) would be the best move?


Apart from non-standard problems (fairy chess, retrograde analysis, etc), there also exist particularly difficult regular chess problems which are used for testing chess engines, like Nightmare II. These might not be solvable by some engines/computers. Take a look at this list

  • The site at wikispaces.com has been shut down. Can the link be replaced by something meaningful? Commented Feb 19, 2019 at 12:45

Not the answer you're looking for? Browse other questions tagged or ask your own question.