"Toby Walsh at the University of NSW Australia has, as reported in New Scientist, studied a generalized version of the popular game Candy Crush Saga and found it be an NP-hard problem, indeed he suggests 'Part of its addictiveness may be that Candy Crush is a computationally hard puzzle to solve.'
His paper shows that early rounds in the game can be modeled as a collection of 'wires' transmitting information across the board, with candies forming inputs and outputs, which can be seen as equivalent to logical statements, this reduces the game to an example of a Boolean satisfiability problem which is known to be NP-complete. A similar technique has been used to show that Super Mario Brothers and Zelda are also NP-hard.
Given that people have spent millions of hours playing the game he notes 'It would be interesting to see if we can profit from the time humans spend solving Candy Crush problems, perhaps we can put this to even better use by hiding some practical NP-hard problems within these puzzles?'"
I have never played the game, but I have played something similar where you try to line up 3 or more like-objects. Could you explain your chocolate comment, as in, what properties does it do or not do that changes the gameplay?
Basically it spreads out each turn unless you make a match next to it or otherwise destroy it. You can't swap it and it can block incoming candies so if it grows too much you can lose by default.