Drill18: The Post-Mortem

Posted by
May 1st, 2014 7:26 am

[Play Drill18: The Mars Project]


This was my 8th Ludum Dare but only my second time doing a strategy/simulation game — which is weird, because those are the kinds of games I live for. I think the barrier is typically that coming up with game mechanics and balance is so much trickier for strategy/simulation games than a more arcade-y one. Additionally, most strategy/simulation games take a while to learn, master, and fully experience — and for Ludum Dare I always aim for something that is playable in 5-10 minutes.

What Went Right:

  1. Streaming. Nothing else keeps me more motivated, interested, on-track, and just plain-old entertained as much as streaming the creation process does.
  2. Low-FPS Pixel Art. I’m actually pretty quick at making simple 3d models so I’m not sure that it saved me any man-hours, but given the amount of stuff that would be visible, pixel art made more sense from a visual and technical point of view. The overhead on the CPU/GPU would be reduced, but also most of the animations in the game are set to run at 2 FPS…and they look good doing so. They only have 2 or 3 frames of animation, and that turned out to be ideal to show “work” happening without being too busy.
  3. My schedule. I always do the same thing, and it always works well. Friday is 1 hour thinking about the game I want to make, then 2 hours prototyping. Saturday morning I’m allowed to throw everything out and start over — but otherwise is all feature development. Sunday is meant to be all polish (though a few last minute features usually get developed here too). I always plan on submitting the game an hour before the deadline (that way I can deal with any last minute disasters). I get lots of sleep, and I try to go for a walk around the block every hour or two.
  4. Knowing my tools. I’ve worked with Unity and C# for three years now, including six previous Ludum Dares. While the 2d toolkit is still relatively novel for me, most of the fundamentals are solidly rooted in my brain now, which means less time checking documentation or figuring out the best way to implement various mechanics.
  5. Working within a genre I know well. While almost none of the mechanics work in any way like SimTower’s (there’s no elevators or day-night cycle), the fact that I was able to visualize the look and feel of the game before I started made it easier to stay focused.


What Went Wrong:

  1. OMG WTF IS WRONG WITH YOU. Every competition you say “I’ll just make something simple like a themed Pong,” but nooooooo…you have to go and make a game that requires a ton of animations, relatively complex AI, interactions between different types of units and different types of buildings, resource management, etc… The number and complexity of the bugs you had to squash was ludicrous. You are not a good example of what people should attempt during Ludum Dare. At least you didn’t do multiplayer again (LD 23 & 26).
  2. It’s an established fact that Unity really doesn’t perform well when you have many, many active GameObjects in a scene. They are just too “heavy”. Despite this, I went with a game where each tile is implemented via its own GameObject…with several components. This made development much quicker and easier, but I was taking a massive risk that the game wouldn’t be performant enough. I had to adjust the scope of the game (camera zoom level, victory condition) to help keep the number of tiles modest. Anyone who continues the play past the victory condition will start to experience poor framerates.
  3. God damn freaking gaps between god damn freaking tiles. Lots of people have experienced this issue with Unity 2d’s system, and no amount of Point filter modes or camera orthographic nonsense could get true pixel-perfect placement in Unity without using custom shaders. I will NOT be using the built-in 2d sprite system for background tiles ever again. It’s slow and doesn’t work right. I wasted far too much time trying to resolve this, and the issue can still sneak up in the final build at certain resolutions.
  4. No ability to do spritesheet/palette swaps using Unity’s built-in 2d animation system. Unity’s 2d sprite splicing and animation system is excellent for a lot of things, but because it does all the heavy lifting in the editor there’s no way to do a spritesheet substitution at runtime (to make it easy to have the different worker careers have different outfits). This is something that would have been possible if I’d written my own sprite handler (which wouldn’t have been difficult).
  5. Improper nutrition. I always prepare good, high-quality food before Ludum Dare…but this time I more or less forgot to eat it. I spent Sunday afternoon completely burned out until I had a proper meal and my energy levels shot up. I should have forced myself to eat more consistently.

What’s Going to Happen:

I was already working on a Unity Tile Engine. I’ve now added support for multi-tile, animated “rooms” and run tests to see if the performance problem was resolvable.

The Ludum Dare version needs about 30 tiles wide and maybe 30 tiles deep (about 900 tiles). I want to support an area that is at least 100×100 (10,000 tiles). So…I tested a map that was 1,000 x 1,000 (One million tiles). To make things even more difficult, I tested on a 3-year-old MacBook Air, the weakest computer I could get my hands on.

Visual FPS – Minimum Required: 60. I got 350 with ~100 tiles visible and 150 with ~500 tiles visible (which is so far zoomed out that you won’t be able to make out tile details).

Simulation Thread FPS – Minimum Required: 2. I got 30. On a 100×100 map, I get 3,800 fps. If I want to tempt ever more complex multithreading issues, I should be able to improve the performance even further for multi-core systems since I can easily chop up the map into chunks for simulation.

So I’ve definitely got an engine that can support an extended version of Drill18. It’s also immune to any weird “gap” issues in the background.  People also seem to like the game. Will this finally be my time to release a polished version of a Ludum Dare entry?

Leave a Reply

You must be logged in to post a comment.

[cache: storing page]