Post-mortem: Roach Escape

Posted by
December 18th, 2016 9:58 am

A week has passed and here’s the post-mortem, before this coming week and all the holiday preparations. I want to thank everyone who has played the game so far and given me feedback. I’ve started working on a post-compo version that will hopefully fix some of the issues that most people have reported, and will expand the content with new levels and elements.


Play it and rate it here!

Roach Escape is a puzzle game inspired by games like Lemmings, Chu Chu Rocket and the Santa Fe ant trail problem, where you need to take your character (in this case, a cockroach) to the goal. By itself, Roachie can only go straight or turn right when it bumps agains a wall, so it can’t get to the goal on its own. By placing items on the grid you can help it find the way.

And that’s it, in a nutshell.

How it all came to be:

Well, you know how it goes. You wake up on Saturday morning, take a look a the theme, possibly start cursing at the choice (or not) and then you spend some time coming up with ideas, pick one and start working. And that’s what I did, essentially. Except that I had two, and both ran in parallel as I was designing for a loong time.

Roach Escape, early mockup - Back then it was "Kitty escape"

Roach Escape, early mockup – Back then it was “Kitty escape”

In case you’re curious, the other game would have been an odd hybrid between a card game and a decorator sim. I had been thinking of Reigns for the cards part as a first approach to the cards part and then moving to a different mechanic that could involve multiplayer, so there would be a balance between randomness and rewards. I really liked the idea, but after describing the concept to friends and family I decided that it would be too content heavy before it became a fun prototype, as opposed to the more proven “build a path to escape” concept. Another reason why I was reluctant at first to what would become Roach Escape was that I knew for sure (although I couldn’t remember the name at the beginning -_-), that there was a clear reference that I’d played years ago, so it might not feel too original. The game, as I would finally recall hours later, was Chu Chu Rocket.


  • Relatively solid: The submitted version was relatively bug free. Granted that gameplay was quite simple and straightforward, but I still think it turned out okay.
  • Decent emergency art style: I had in mind to rehash most, if not all of the game assets to make it more fitting with the theme, but due to the lack of time I and had to compromise with quick improvements on the original placeholders. Luckily, they didn’t look bad IMHO.
  • Lots of potential: As I’ve kept iterating during the jam and outside it I’ve come up with lots of ideas, from world elements (switches, teleporters,…) to objectives to provide variety to the gameplay and make the player tackle a level in different ways. Also, some friends of mine IRL have made several suggestions about potential new features that can also be interesting.


  • Indecision. I think one of the worst mistakes I made this time was to take too long before I started coding. As I said before, I had two ideas and I admit that I kind of preferred the decorator one. However, as the concept sounded riskier I went with this. I don’t regret the choice, but still I lost several hours trying to decide, and even after that I had a slow start because I wasn’t completely focused.
  • Flawed first-time user experience. The game has different states, but it basically comes down to two: the edition mode and the playback mode. From this one you could either pause the game or lose/win when you finish, and you can return to the edition mode pretty much anytime you want. This setup sounds easy enough and it is, but it wasn’t clearly communicated at all on the game. As a result, some people including myself at times would click on a tools and get no visible response because they weren’t aware that they were outside the edition mode. This is one of the first things I’m going to fix.
  • Reducing the walls width: Originally the walls were thicker than on their current incarnation. This implied that either the player was drawn above the walls (and that didn’t only make no sens, but also looked ugly), or that it was below and appear partially covered. Also, because of the way I’d packed and read the sprites originally, and due to the way I calculating collisions, reducing the player sprite size wasn’t an option, so I chose to make walls thinner. The result? Many people have trouble seeing them, and they’re right.
  • Usual Haxeflixel amnesia. The last few weeks before the Ludum Dare up until a couple of days before the theme announcement I had been coding on Unity for a personal project, and on C++ with Cocos2dx at work, which means I was already juggling between both of them. Then I chose to throw a third one into the pot and use Haxeflixel for the game because I’ve become mostly comfortable with it for LD, and it makes super easy deploying builds for a variety of targets. However, I tend to go for months without using it, which means that every time I pick it up again I need context switch, not only in terms of language, but also in the way all those engines and libraries handle coordinate systems or scene entity hierarchies.Because of this, my first two-three hours of coding work I’m generally like this until my brain starts digging for past knowledge and I enter the proper “Haxe is cool! Everything is easy again!” mode

Thanks to this, the last hours are generally a breeze because it provides a lot of functionality out of the box(or there are lots of addons available) and because it makes painless things such as integrating audio or hacking a quick and dirty but functional UI screen.

  • No music, superbasic sound. Yup, lack of time :/
  • Post-submission caffeine high I can get anxious for any stupid reason easily enough by myself, so I tend to stay away from caffeine except for tea drowned in milk and sugar. That means I’m quite sensitive to sharp increases on my intake, as I haven’t built any tolerance. During the last weekend, though, I had about a couple of cups of coffee (not too black, but still), a couple of energy drinks and a half a litre bottle of cola. The result? Pretty much I went to bed on Sunday night and started tripping balls.

    Graphic depiction (Now seriously, feeling my mind race like a crazy horse even though it was exhausted was kind of unpleasant)

And of course, going to work the next day was a nightmare.


  • Loose connection to the theme. As I mentioned, the emergency art style did the trick of not looking too awful, but that came at the expense of having a game that doesn’t really feel related to One Room. My intention was that each of the levels represented a room (the easiest, laziest kind of connection), but also reinforce that through the graphics. However, that wasn’t possible in the end.
  • Too short. At the moment there are only 5 levels. People haven’t complained about the amount, or the difficulty progression too much, but I still think that the design itself wasn’t particularly good either. I felt like I needed a lot more levels, and that they should show more effort.

Here are some screenshots of some mockups I’ve been sketching for the post-compo version. I’ve just started implementing them today and fixing some other stuff, which means that I’ve broken the game. However, I expect that soon I’ll be able to demonstrate the improvements.

02_mockup_post_compo_landscape_editv2 03_mockup_post_compo_landscape_play 04_mockup_post_compo_landscape_paused 05_mockup_post_compo_landscape_won

Leave a Reply

You must be logged in to post a comment.

[cache: storing page]