Thinking aloud about remixing the sqoosh's Blinded project

Though my game could also have a straight-line solution, it just wouldn't be practically probable.

Yeah, I am not smart enough to re-create the complexity of the maze that your pen approach can provide. I don't know how to change my code so that it would produce more complex mazes. Any ideas?

I feel like it would be beneficial to re-increase the size of the grid, maybe even do half-steps? Increase the distance that each iteration step takes, but stamp halfway in-between so that it creates gaps.

Like this:
image

image

Already more complex.

EDIT: Though I have to admit.. this brings errors at times, so maybe some extra detection to prevent the following from happening?

image

Great idea, you're smarter than me. Yes, that would do the trick. Do you mind if I change the code and give credit to you in a comment inside the code?

Of course. - I don't mind I mean.

Yeah, extra detection has been added, but still from time to time disconnected islands appear, like this upside-down British and Irish island : ~)

disconnected_islands

Oh boy, that is troublesome.

I didn't know what else to do - so decided to add another slider to enable jumping over multiple (minimum 1 and maximum 3) tiles lengths at once, when islands appear.

Alright.

I have never been fan of games, maybe because I've not been good at games, so I prefer doing the interactive simulations, instead.

I am now wondering if the game could be somehow turned into a simulation. Maybe something similar to the self-driving car simulation made a year ago.

I think it would be cool to make some kind of pseudo-neural network that evolves over time to solve a maze.

Maybe using lists for moves that change randomly and scores are taken to see what to keep and what to change or rollback.

As for the interactive bit, you could change how score is given, to make it strive for different things, or you could affect something inherent about the simulation to see how it adapts.

I think it would be funny to see it try to win as fast as possible but then have its goal change to losing as soon as possible, and seeing how it reacts.

I wonder if it is possible to turn the orange vector (starting at the blue tile and ending at the violet tile) algorithmically into the series of blue vectors.

orange_to_blue_vector

another_image

Maybe.. It'd be difficult though.

One way of thinking about it would be to try and make a 3-tiles jump in either X or Y direction towards the goal; if that would end up in black, then try 2-tiles jump; if it doesn't work, then try 1-tile move.

I thought of that, but it could take a while.

Yeah, I am, too, just thinking aloud. It's nothing wrong if you share your thoughts only, instead of an actual code or a project; I guess.

I'm doing some more thinking aloud :~)
After each jump, a new orange vector can be drawn towards the goal; and the new vector's longer component (red on the picture below, either X_dif or Y_dif) can determine a new direction for the next jump...

new_after_each_jump

Using a rise over run kind of thing would be smart.

Yeah, the 'rise' ('run') = differences in Y (X) coordinates of the blue and the violet tiles.

The differences need to be calculated again at each new position of the bot, using the differences between the violet tile's X (and Y), and the bot's X (and Y) value.