TOM FRANCIS
REGRETS THIS ALREADY

Adventure Time Game Post Mortem

On the weekend of the 15th, Fantastic Arcade held an Adventure Time-themed Game Making Frenzy. It meant anyone could make a game with Adventure Time characters for the purposes of that compo, which is rare, so I did.

I finished mine at 2am that Monday, left the next day for a work trip, then spent all of last weekend working on Gunpoint, so I haven’t had time to talk about how it went. Here’s how it went!

Firstly, here’s where you can grab the actual game and try it. It’s for Windows. If you’re not using Windows or don’t care enough, here’s a video of me playing it:

I wanted to make an Adventure Time game that:

• Wasn’t just an existing type of game re-skinned to look like Adventure Time
• Focused on something unique about the show and central to it
• Featured both Jake and Finn and in some way reflected their friendship

So I made one where:

• You play as Jake, and can stretch your legs and your arms with the cursor keys
• Finn is AI controlled, but you can pick him up
• You need Finn to fight the enemies
• Finn needs you to reach the enemies

But I didn’t have time to:

• Generate randomised enemies from component modules, to keep encounters fresh
• Have boss enemies with modules like ‘Forcefield’, which would require you to drop Finn behind them to take it out
• Have flying enemies you had to grab and hold down for Finn to bash
• Have a Jake Suit mode where Finn would control movement and you control punching
• Let you grab onto enemy modules and pull them out to help Finn defeat them
• Have a tiny crown of invulnerability on some enemies that you can bonk off by nudging it, like in the Adventure Time pilot

An enemy visual redesign I immediately undid.

What went wrong

For one, that’s an awful lot unique mechanics and ideas. It’s not a mistake to have more ideas than you’ll be able to implement, since sometimes as you build the foundation, it becomes clear that some ideas are more viable and have more potential than others. But the problem with my laundry list is that there was no one thing on it that I was sure would be fun.

Once I had the stretch mechanic working, I re-prioritised. It felt nice, so it was an easy call to scrap Jake Suit mode, which’d be a lot of work and take you away from the bit that already worked well. I reckoned the biggest gains would be in something that added challenge, since it didn’t have any yet. So the Forcefield thing became my top priority.

This turned out to be a son of a goddamn bitch. It’s easy to make an enemy invulnerable from one side, but really hard to script out exactly what Finn should do in that situation. If he stops attacking, you feel like he’s not trying, and might not even know why. If he carries on attacking, he’s effectively paralysed, and the logic of whether and when you can pick him up and when he learns to stop was super messy and complex.

I still couldn’t get it behaving well after an hour or so, so I tried a different idea for the same effect: fan modules that would blow Finn back. Similar problem – it was hard to fully model Finn’s behaviour in all the possible states this could push him into without him seeming stupid or annoying, both of which were unacceptable.

Ultimately the basic design, where the guy who does all the combat is AI controlled, didn’t lend itself well to having a challenge bolted onto it. If I made bullets really hurt either of you, it was just frustrating, since you weren’t directly in charge of making sure you didn’t get hit.

I decided it was better to have no challenge than a challenge that spoiled what did work: stretching. So I spent what time I had left – half a day – making a tutorial, dialogue system, character progression, random level generator, redesigning the look of the enemies, fixing bugs, and adding the snail that the compo rules dictate must be included.

What went right

Stretching your legs
Is a strange and enjoyable feeling, which was the main thing I wanted to achieve.

Art
Using screencaps for reference, I was surprised how easy it was to make decent-looking sprites of the characters. Some of them I traced directly using layers, others I hand-drew, most were a hybrid, where I had a good base but needed to redraw limbs.

That said, I still knew I wouldn’t have time to do any proper animation, and I wasn’t sure I could make Finn attack stuff convincingly without it. Turns out a ‘charging towards enemy’ pose and a ‘just struck enemy’ pose looked fine – the crudeness of it is funny but not out of character.

In general, too, Finn kinda feels like Finn: super enthusiastic, utterly fearless, always takes the direct approach.

The random level generator
It has almost nothing to work with – there’s literally one platform type it can place, and two types of near-identical enemy. But it gets way more value out of that than I ever could have hand-designing levels, even if I had more than a weekend. And it took less than half an hour to code. I just made a mental list of all the things that should be random about the levels, how they should scale, and wrote out this:

And suddenly my game was infinite, as varied as it could be with one platform type, escalated in ‘difficulty’, and scaled itself dynamically to the player’s current stretch limit – something I couldn’t know as a level designer, since it’s up to you how many ‘Jake bits’ you collect to upgrade yourself.
In Game Maker’s internal structure, the game is actually just one room, and every time you get to the end of it, it reconfigures itself before putting you back at the start.

Originally it was going to end when you collected enough Jake bits to restore Jake’s normal stretch capacity, but since I found myself playing way beyond that, I decided not to limit for the player either. So you can keep boosting Jake’s maximum stretch upgrades, keep playing new levels, and you just get a little exchange between the two of them to acknowledge that they’d succeeded, to let you know it’ll just go on forever if you want. (You probably won’t.)

Conclusions

• I think my design should have included one element of challenge that I was 100% sure I knew how to implement. Maybe something puzzly, maybe objects for Jake to pick up or switches for him to push. Hinging it all on all this complex enemy interaction left too many unknowns for Finn’s AI – I should have kept thinking until I had at least one idea with less uncertainty.
• Random level generators are amazing, and easy to make when the ingredients are simple.
• Any kind of persistent character improvement makes the interaction that leads to it much more meaningful, and again, it’s super easy to implement when the ingredients are simple.

More Adventure Time