I have no idea what the rules are. There is no explanation at all. At one point, the label "Tutorial levels 1-3" appeared on the screen for about 0.5 seconds and disappeared before I could click on it.
I feel it would be better to start cycling 2 colors, so that you naturally "catch" the reasoning you have to do. Then keep adding colors.
It reminds me of the dual n-back game, where you had to remember N steps before of 2 things. You start with 1-back and progress once you "get it".
As it is now, I just couldn't "get" what was the required reasoning behind this puzzle, before I got frustrated and left. And the hints didn't give me nothing personally, because once you get it, it basically solves it, without actually helping you understand the reasoning process.
Also similar to Sudoku... if you start with a difficult one, you just get lost. You have to learn the reasoning tricks.
This is a variant of the "Lights Out" puzzle which has interesting mathematical properties related to linear algebra over GF(2) and can be solved systematically using Gaussian elimination.
I like it! I'm curious how the target move count (the one used to calculate the efficiency score) is determined. I thought it was the minimum number of moves required, but then I solved a puzzle in 5/7 moves for 140% efficiency.
edit: I then got a level 8 puzzle that could be solved in just 3 moves! I wonder if this is a deliberate possibility, or an issue with the puzzle generator/classifier?
the math is hard. bascically the level generator starts with a solved grid and then does "reversed clicks", so yeah, there might always be cases where there are faster solutions, but I try to minimize the propability of these cases.
you aren't really trying to minimize anything. I just wrote a simple gaussian elimination solver that generates all N solutions (the win condition is that all the cells are the same color, not a specific one) and then chooses the one with the least moves. It is optimal. It pretty much always gets over 200% efficiency.
I can't really say that your calculation is off, just that you're not doing one at all.
Found it: the puzzle generator had a "parity check" that only allowed clicking each tile once. On a 3x3 grid (9 tiles), this capped all puzzles at 9 moves maximum, even though levels 10-20 were supposed to require 10-20 moves.
The root cause: We were using binary logic (clicked/not-clicked) in a game with 3-color cycling. With 3 colors, each tile can be usefully clicked up to 2 times before returning to its original state.
fix v1.87.0: Changed from binary parity to tracking actual click counts, allowing each tile to be clicked (colors - 1) times. This enables up to 18 moves on a 3x3 grid, exactly as pyt correctly calculated!
should not be, what browser / environment? there is some math after each move to check if the level is still solvable. not all states are solvable (in all grid sizes) but it should be pretty fast.
That should be it then. You shouldn't need to do that after each move. You should do that once at the beginning of a game, and if it's solvable on the first move, it's solvable for any configuration that can be reached by applying moves.
Ex. if there are 4 colors, and I clicked once, I can click 3 times to get to my original state. Therefore if you prove one configuration solvable, then, given that I can always reverse my own moves to get back to that state, it means all configurations are (that can be reached in the current game).
tldr: any move on a solvable grid, will result in another solvable grid.
RGR / GBG / RGR would be the final level if the game progressed beyond 10, it requires 18 clicks to solve. All other states can be solved in fewer than 18 clicks.
I have no idea what the rules are. There is no explanation at all. At one point, the label "Tutorial levels 1-3" appeared on the screen for about 0.5 seconds and disappeared before I could click on it.
I have no idea how this is supposed to work.
I feel it would be better to start cycling 2 colors, so that you naturally "catch" the reasoning you have to do. Then keep adding colors.
It reminds me of the dual n-back game, where you had to remember N steps before of 2 things. You start with 1-back and progress once you "get it".
As it is now, I just couldn't "get" what was the required reasoning behind this puzzle, before I got frustrated and left. And the hints didn't give me nothing personally, because once you get it, it basically solves it, without actually helping you understand the reasoning process.
Also similar to Sudoku... if you start with a difficult one, you just get lost. You have to learn the reasoning tricks.
will do the 2 colors for starters after my business meeting right now.
done
It appears to be "lights out" with more states. https://en.m.wikipedia.org/wiki/Lights_Out_(game)
i improved the tutorial and also added a proper flip effect so that it is now more clear what happens!!! working on a how to play tutorial also
added a "how to play" section plus more clear tutorial - the "flip-over" effect of the tiles helps to I think
Remind me never to play Mao with you lol
This is a variant of the "Lights Out" puzzle which has interesting mathematical properties related to linear algebra over GF(2) and can be solved systematically using Gaussian elimination.
A long time ago, I solved a 5 by 5 version of this with a genetic algorithm. Wasn’t the most efficient way.
As you say, Gaussian elimination can be used for a more systematic approach.
Aren't the rules basically the same as Conway's Life as well?
Not at all: https://en.wikipedia.org/wiki/Conway%27s_Game_of_Life#Rules, https://en.wikipedia.org/wiki/Lights_Out_(game)
https://en.m.wikipedia.org/wiki/Lights_Out_(game)
Just curious, did you see my post before starting this project?
https://news.ycombinator.com/item?id=44091036
https://tincture-574568295911.us-west1.run.app/
Good implementation, but this puzzle had been invented many times before.
I like it! I'm curious how the target move count (the one used to calculate the efficiency score) is determined. I thought it was the minimum number of moves required, but then I solved a puzzle in 5/7 moves for 140% efficiency.
edit: I then got a level 8 puzzle that could be solved in just 3 moves! I wonder if this is a deliberate possibility, or an issue with the puzzle generator/classifier?
the math is hard. bascically the level generator starts with a solved grid and then does "reversed clicks", so yeah, there might always be cases where there are faster solutions, but I try to minimize the propability of these cases.
you aren't really trying to minimize anything. I just wrote a simple gaussian elimination solver that generates all N solutions (the win condition is that all the cells are the same color, not a specific one) and then chooses the one with the least moves. It is optimal. It pretty much always gets over 200% efficiency.
I can't really say that your calculation is off, just that you're not doing one at all.
Maybe try not asking chatgpt to do it for you...
Too hard to play, felt like it told me what to do for 3 levels then got impossible
I like the look and feel.
I did not really get how the mechanic works. Once I got how the mechanic works I did not get how I can use it to solve it.
That said: I would have said the same about a rubik's cube with a 10s attention span.
> A new kind of logic game
Until you find the pattern.
Also all level after Level 10 becomes the same.
Fixed it. The level 10 bug, not the finding the pattern.
Neat game. Needs levels and milestones in order to keep me motivated. How about a theme as well? Maybe adjustable difficulty. Overall, good start.
Other post: https://news.ycombinator.com/item?id=44712829
After level 10 the levels are all the same?
Thanks for the bug report!
Found it: the puzzle generator had a "parity check" that only allowed clicking each tile once. On a 3x3 grid (9 tiles), this capped all puzzles at 9 moves maximum, even though levels 10-20 were supposed to require 10-20 moves.
The root cause: We were using binary logic (clicked/not-clicked) in a game with 3-color cycling. With 3 colors, each tile can be usefully clicked up to 2 times before returning to its original state.
fix v1.87.0: Changed from binary parity to tracking actual click counts, allowing each tile to be clicked (colors - 1) times. This enables up to 18 moves on a 3x3 grid, exactly as pyt correctly calculated!
Now levels properly progress: - Levels 1-18: 3x3 grid (max 18 moves) - Levels 19-32: 4x4 grid (max 32 moves) - Levels 33+: 5x5 grid (max 50 moves)
Thanks for playing!
Why does it take a whole second after each click for it to update?
should not be, what browser / environment? there is some math after each move to check if the level is still solvable. not all states are solvable (in all grid sizes) but it should be pretty fast.
That should be it then. You shouldn't need to do that after each move. You should do that once at the beginning of a game, and if it's solvable on the first move, it's solvable for any configuration that can be reached by applying moves.
Ex. if there are 4 colors, and I clicked once, I can click 3 times to get to my original state. Therefore if you prove one configuration solvable, then, given that I can always reverse my own moves to get back to that state, it means all configurations are (that can be reached in the current game).
tldr: any move on a solvable grid, will result in another solvable grid.
RGR / GBG / RGR would be the final level if the game progressed beyond 10, it requires 18 clicks to solve. All other states can be solved in fewer than 18 clicks.
grid gets bigger over time
This isn't new. This is exactly the same as "Flip" from Simon Tatham's Portable Puzzle Collection.
Play here: https://www.chiark.greenend.org.uk/~sgtatham/puzzles/js/flip...