I might be dense, but all I see is a square with a triangle in it and a circle, and can't seem to make it do anything. It is throwing a 408 timeout though, so maybe its just under load?
I like the game and that it teaches you the rules as you play without needing any explicit instructions. I thought some of the comments between levels were annoying. Also the little flag in the corner indicating you're over the correct target could be a little easier to see at a glance.
Just curious, how do you design these levels? Work backwards from the solution?
edit: I'm a little further in, and I'm starting to think the annoying comment thing is on purpose...
Are you from the US? I'm from the UK and I really liked the comments too. I really find that people from the US are more likely to dislike humour in this kind of game. Especially Californians.
Ever notice how people from the UK tend to generalize in a condescending way that only really serves to highlight their backward, provincial attitudes? Sort of like this?
The comments kinda ruined it for me (I did enjoy the "I'm starting to feel like putting jokes between levels was a bad idea" one).
They were alright at first but I started developing a sense of urgency, like I just had to complete a few training examples before some kind of main introduction to the "real game" (which I expected to develop into some form of a puzzle game world, with labeled levels and sections, maybe themes, etc). After more and more one liners (that don't add any substance and aren't particularly funny) I became increasingly annoyed and was trying to pass levels just to get past that phase of the game. After a while I realized that that was the whole game, and my expectations had set me off on the wrong foot, so to speak, and I lost the will to continue.
I do like the concept and the puzzles were enjoyable, but I guess my recommendation would be to leave out the comments that make it seem like they are leading up to some prize or event but never actually go anywhere. The complete lack of indication of the level I'm on and where that is in the puzzle world (Am I still in training, or...?) also contributed to me losing interest.
I really enjoyed this game. I went through every level and what amazed me most was the creativity it took to come up with the different levels of difficulty. It's not the repeated stuff you get in Candy Crush. Finding a solution is one thing but designing the levels is a different skill set altogether. We'll done.
A very promising start and I'm very much in favor of the "randomly click to learn" approach to teaching the game rules. Can do without the comments between levels though.
Great game. Was it inspired by the telescope game that was on the Dyson (vacuum cleaner) website a few years ago? Seems to be gone from the dyson site now, but I've linked to a copy in a separate submission [1].
And go! Tomorrow this game in full will appear on the Google Play store with the exact same colors. In 3 days this game will be rewritten in Swift, Erlang, and have a community based variation. In 4 days someone will find a way to merge this game with 2048 and Flappy Bird. In a week there will be 20 variations of this game including one called Dodge Squares on the iTunes App Store. I'm not psychic, I've just seen this script before.
These are all cool things that I would be happy to check out, play with and read about. It's either this or another Show HN filled with angry detractors (kinda like this one), something about the NSA reading my email, or the latest boycott because Facebook, Amazon, Dropbox, Google, Apple are doing something mean to someone.
I don't think so. 2048 and Flappy Bird are based on a mechanic including random gameplay. This game is full of content (design of levels and text between them). Sure, someone can clone it, but it will take them slightly longer and it will be much tougher to make variations (possibly coming up with new levels, but that's hard). It's not the same kind of game.
It should be possible to write a solver for this with a "decision tree". There are decisions where the blocks run out of the screen - you ignore these. You only take into account "sensible" decisions (this has to be formalized). When there are multiple possible "sensible" decisions you branch. One decision consists of an action "click x times on block X and y times on block Y …"
Any solution is on the form 12232322... where the number expresses what square to click. It's pretty easy to brute force the game from there. Sure, you can use tree search and prune the search a bit, but it's still brute force and may take exponential time for tough levels.
You can also look into some of the research that has been done on Sokoban AI's.
There's no need to search all move sequences, because so many of them end up in the same game state (a game state is a combination of positions and orientations of arrows, with a cut off on positions because far away ones because don't affect whether there's a solution or not).
Of the levels I saw (up to 31), the relevant board area was never more than 20x20 and there were never more than 4 arrows. So the state space had size 16(2020 choose 4) ~= 17 billion. Of course most of those states won't be reachable from the starting state... so the actual number of states you need to explore is probably more like a million or even a thousand.
I think a program could solve all of the game's puzzles in under a second.
Shouldn't that 16 be a 256? 4 squares with 4 orientations each = 4^4 = 256.
But yeah, definitely pretty doable with a brute-force memoized search. I imagine a lot of state space could be pruned by doing some analysis of when squares can only move further away from their home dots (and not be turned around or pushed back or anything).
Should be far easier to create a heuristic search solution. Very naively you could probably get good results by exploring nodes in order by sum of distances from blocks to colour points if blocks could move in any direction. You'll eliminate most of the move away and out of screen behaviours naturally and by preventing the exploration of a previously explored state. You could explore better heuristics but given that the branching factor is only n where n is the number of blocks, most puzzles are solved in under 50 moves and most solutions get pruned quickly both position duplication or bad solutions quickly mangling the heuristic. The hard problems for this program are ones where you have to make a whole sequences of moves away from the objective in order to get a key directional move that accomplishes it.
I've gotten to 31 without the need for pen & paper - but yes, your right, the difficulty balancing is great. It's had me scratching my head a couple of times, but not so much to cause me to quit.
For up to four squares (npi) I thought a simple mapping A S D F would do. Pressing A pushes Square A in its current direction. Squares would have the letter tagged on them subtly.
207 comments
[ 3.0 ms ] story [ 258 ms ] threadI got stuck on 21, and that's enough for me for now. Good puzzle game!
Just curious, how do you design these levels? Work backwards from the solution?
edit: I'm a little further in, and I'm starting to think the annoying comment thing is on purpose...
They were alright at first but I started developing a sense of urgency, like I just had to complete a few training examples before some kind of main introduction to the "real game" (which I expected to develop into some form of a puzzle game world, with labeled levels and sections, maybe themes, etc). After more and more one liners (that don't add any substance and aren't particularly funny) I became increasingly annoyed and was trying to pass levels just to get past that phase of the game. After a while I realized that that was the whole game, and my expectations had set me off on the wrong foot, so to speak, and I lost the will to continue.
I do like the concept and the puzzles were enjoyable, but I guess my recommendation would be to leave out the comments that make it seem like they are leading up to some prize or event but never actually go anywhere. The complete lack of indication of the level I'm on and where that is in the puzzle world (Am I still in training, or...?) also contributed to me losing interest.
Where are the mobile versions? I'd like to play it in my iPad.
Kudos.
[1] : https://itunes.apple.com/us/app/swap-boxes/id753811885?mt=8
(FF 31 on OSX, why do you ask?)
For me, it was because I had dom.storage.enabled=false. LocalStorage obeys cookie settings, so I guess disabling cookies would do that too.
[1] https://news.ycombinator.com/item?id=8092648
https://play.google.com/store/apps/details?id=com.chuger.squ...
I am the 1%.
Edit: got it. http://www.flotcharts.org/
You can also look into some of the research that has been done on Sokoban AI's.
Of the levels I saw (up to 31), the relevant board area was never more than 20x20 and there were never more than 4 arrows. So the state space had size 16(2020 choose 4) ~= 17 billion. Of course most of those states won't be reachable from the starting state... so the actual number of states you need to explore is probably more like a million or even a thousand.
I think a program could solve all of the game's puzzles in under a second.
But yeah, definitely pretty doable with a brute-force memoized search. I imagine a lot of state space could be pruned by doing some analysis of when squares can only move further away from their home dots (and not be turned around or pushed back or anything).
https://gist.github.com/Zimmux/c0fdb3ddad654b6b0ae2
Sign of a good game.
Such silliness really drives me away from hn.
<meta name="viewport" content="minimal-ui">