Adventure Kit post mortem

Sickhead Games: Adventure Kit post-mortem



It's been a couple of weeks since our release of the Adventure Kit. We've had a great response from the community, as well as some valued criticism. In all honesty the scope and capabilities of the project had grown quite a bit from its beginning to its release, and we felt a number of growing pains along the way. So let's begin a little post-mortem on the project!

First of all... to those unfamiliar with the Adventure Kit, basically it's a starter kit for Torque Game Builder which includes tiles, sprites, code, and sounds used to create a top-down action and RPG game, but also included was a demonstration "game" that's comprised of two diverse levels just to show what can be done with the kit. By now we should have a new video on the product page to show some of that off.

And while we're discussing the rights and wrongs of the development of TGB games, I must first point everyone to Andrew Douglas' helpful summary of his own experiences creating art for TGB. Some valuable tips in there.

What went wrong.

Prototype, prototype prototype.
That's the mantra we hear often and brush aside to get into the fun stuff, right? Well, brushing it aside cost us at least 3 weeks of lost development time on a project that ended up being about 3.5 months. A somewhat considerable chunk of time spent creating art that wasn't necessary or that had to be reformatted. And here's why...

Sprites or Tiles? Underestimating the abilities of sprites
In the beginning, we were really unsure what kind of performance we would get if we used a lot of sprites. For this reason, we began by creating most of the environment art as tiles. Eventually we realized two things: 1) that we could put 800+ sprites into a map and still get 95+ fps, and 2) that tiles were much less flexible than sprites. Check out the following example. The image on the left illustrates the assets required to create a stone wall with tiles, and the image on the right illustrates the assets required to create a stone wall with sprites.



Not to mention that sorting individual tiles is virtually impossible. Underestimating what we could do with sprites cost us a lot of time.

Packing: get to know it.
Of course efficiently arranging your art into tile sheets and sprite sheets only helps keep down the size of your game on disk. How well those images can then be "packed" into usable chunks for a game, resulting in less RAM usage for your game, is another matter.

Now one problem we ran into was the packing of tiles. We figured that 128x128 pixel tile made sense. Good general size for us, power-of-2 dimensions, etc. What we failed to take into consideration was how the additional 2 pixels per dimension added by the filter pad would affect packing. This caused an inherent problem in the efficiency of the packing system where the tiles were always slightly larger than what would fit into a power-of-2 packed map. So a 4x4 packed image would need 260x260 pixels of space, just a hair over 256x256, and as such had to be packed into 512x512 maps, with quite a bit of unused space. If we had created the tiles at 126x126, they would fit perfectly in power-of-2 packed maps. Absolutely no wasted space.

We got around this a number of ways and still kept the packed maps fairly efficient. But 126x126 or 252x252 would have been perfectly efficient.

Polish, documentation, marketing, and the last 10%
They say that the last 10% of the project is the most time-consuming. We spent at least 2 weeks on the verge of shipping. Polish items, like tweaking the AI, adding little gameplay bits, final sounds, and Gigantork were not completed until the last week.

The amount of effort required to construct documentation and the product page were both severely underestimated. While we pride ourselves as being renaissance men of sorts, writers we are not. While the TDN documentation stretched our skills as technical writers, the product page stretched our skills as marketing-type people. Indies sure do wear a lot of hats! Even though the publishing tools on the GG site are excellent, the product page took a while longer to build than we had anticipated. GG offered to help in this area, but unfortunately we didn't take them up on it. Lesson learned... take the help when it is offered.

And don't underestimate that last 10%!

What went right!

Letting your tools do the work for you...
We decided early on to use the Torque Orc as the 2D character for this starter kit. He's sort of the face for Torque, and the model and animations were already made, so we could concentrate on making it work in TGB. Tom wrote a clever little app we called SpriteMachine that allowed me to set up the camera angles and number of frames for any given DSQ and batch export the 2D frames out to a folder. Then I used automation in Photoshop (Actions and batch processing) to process all the frames. This allowed me to resize, crop, adjust color, and sharpen over 235 individual frames in a matter of minutes. What could have been a horribly long process was very easy.



Genetica Pro. I love this program. It's a node-based tilable texture creation program. Node-based, meaning you create textures by setting up various programmatic nodes (kind of like a flowchart) based on noise, repeating patterns, and "filters" to create the textures you want, and you can modify nodes at any point in your chain to adjust your texture. It also exports normal maps and some height maps, so if you've been catching the latest on normal and parallax maps in TGB, there's just that much more for you to love about it. I've really just cracked the surface. I haven't found too many people using this. If you are, please let me know.



Love LevelBuilder...
When we started the process LevelBuilder was still rather new and config datablocks were partially working. Still from the beginning we saw the potential of the new tools and decided to deeply ingrain the workflow with them. LevelBuilder is a breeze to use. The simple power of drag and drop sprite placement and the new tile layer editor makes building levels fun and shockingly addictive. With a few drags and some cut and pastes you've populated your level. LevelBuilder even held up well when our levels got up to the 1000s of objects in the scene.

Good timing...
The release of the AdKit happened to fall within weeks of the big TGB 1.1 release and the Torque X announcement. So a lot of heads were turned towards Torque Game Builder at that time, which happens to only have one expansion pack/kit. Wish I could say that was all planned, but things really just happened that way. There's a lot to be said for good timing.

Working with the Garage...
We decided early on to partner with GarageGames to develop and distribute this product. We made a good decision there. The publishing tools are great and they offered help and encouragement all along the way. Tim was always available for any questions we had on the publishing and legal side of things, and even sounded like he knew what he was talking about! That was comforting, even though somehow we still think the lunatics are running the asylum. ;)

We have to point out the great effort that Justin and the rest of the TGB team put into the 1.1.1 TGB release. If you haven't checked it out, you're missing out on some phenomenal changes. Also, they listened to our feedback and added features to TGB that made it possible for us to ship a far better product. In addition, they offered great advice and criticism on early versions of art and functionality.

Conclusion
It's hard to tell how successful TGB Adventure Kit will be in the long run, but so far is doing really well. We have plans for a 1.2 release with a little more content for current owners, a few more features, and support for non-Pro users (even trial users could buy it). Keep an eye out for that!

And for those AdKit owners that made it this far into my exhilerating post mortem... some Easter eggs! In the console, try either "miniMe" or "supersizeMe".

Thanks for listening.
Russell



Development Data:
Developer: Sickhead Games
Distributor: GarageGames
Platforms: Windows/OSX
Number of full time developers: 1 x 2.5 months, 2 x 1 month
Length of development: 3.5 months
Release date: August 18, 2006