Harold whipped together some sweet prototypes of SLOA this weekend. He agreed to let me post them as an educational example of how prototyping works. Think of this exercise as speed dating mixed with the post mortems that you find in the back of Game Developer magazine.
Update: I added the latest prototype. You can download it here
The first version of SLOA sported the core game mechanic, a simple conversation system. There were two token, the drab and asexual Suburbanites and the dapper green Spies. There really isn’t any game here, just a technical demonstration of a basic gameplay system.
What went right:
- Primitive graphics: I’m very proud of Harold’s graphics composed entirely of cones and spheres. It took him substantially less than an hour to put these together and that is as it should be. Time spent on creating pretty graphics during the prototyping stage is time misspent.
- No UFO: Prototyping is as much about what to leave out as it is about what you build. In the design, the UFO, the main character of the entire fricking game, is really just a mouse pointer when you come right down to it. The system mouse pointer is good enough.
- Smallest possible step to feedback: Even though there is no game here, there was still enough to make some comments that dramatically improved the next version.
What went wrong:
- Getting hung up on graphics: There was in fact a prototype-0 of SLOA. Harold spent a good hour trying to get the antennae to smoothly animate out of the head of the spy. But there was a bug someplace and a few minutes turned into an hour of frustration. He was focusing on making it pretty instead of making it work.
- Colors: The green of the spies was too light and they blended into background. Even at this early stage it was possible to get feedback on the game’s information design.
- Complexity: The conversation behavior was too simple. With so few characters there wasn’t much interesting happening.
Version 2 fixed a couple of problems with version 1. The colors were brighter on the spies and there were a lot more spies on the screen at once. Everything was peachy keen. Except for one little detail…
What went wrong:
- Conflict: There was a desperate need for some sort of conflict and challenge. It turns out that conversation alone is boring. This is a wonderful discovery since it shows us that we can’t use the conversation mechanic as our core game mechanic. It is best to discover these things early on in a game design. 🙂
Version 3 sought to augment the basic mechanics with some classic conflict. Harold went all out and added in Agents and the detection score. The agents had sun glasses and black suits. MIB of the most classic sort, baby. They also had health meters so you couldn’t just rapidly click-massacre a dozen agents in a millisecond.
What went right:
- The game was born: There was actually a challenging little game evident in this prototype. The conflict escalated, agents wandered about, your score increased. After a decent 5 minute playing session, the frantic clicking all came crashing down and the game ended. People were comparing high scores and improving their scores with each play. When people play your dismal prototype more than once, you know you are onto something.
- A recognizable setting: People grokked the alien-themed setting. The agents added the necessary burst of context that helped initial players feel like they were doing more than just clicking on random icons.
What went wrong:
- Weak connection between action and reward: The score in the game would go up and you’d kill agents, but there wasn’t a clear connection between what you are doing and the rewards. A good game should be like an instrument. You swing the stick, hit the drum and it makes an enjoyable sound. Swing, boom. Simple enough. Version 3 didn’t that rhythmic clearly delineated musical response. It was more like one of those little rain sticks that you turn upside down and you get a long tinkling noise. Pleasant, but not addictive.
- No focal point: Another problem was that little guys were wandering all over the place and you had literally 20 different things to worry about all at once. Successful games with lots of moving objects on the screen tend to have one or two focal points that the player can watch. In most shooters and action games, there is a central location that must be protected and then the player zones out and lets their peripheral vision take in all the various threats. Version 3 lacked this focus and the result was a game that felt more stressful than fun.
- Killing agents wasn’t fun: To continue with the music analogy, when you hit a drum, there is a satisfying physicality to the action. Most decent interfaces, be they simple buttons or complex game mechanics, have this tactile element. The interface for killing agents involved a slow and steady draining of their life force. It made sense when you described it on paper, but holding down a mouse button didn’t feel exciting. Combine this with bad collision detection and killing agents was another frustrating experience.
In version 4 we finally had a robust enough system to start playing with the game balance. We added several mechanics that improved the feeling of direct control in the hopes of reducing the risk / reward cycle time.
What went right:
- Control over individual spies: We had talked about being able to drag a spy in a direction in order to reorient them. This would let the player be more tactical in their position of the spies and hopefully result in their feeling more control over the board. The implementation was a qualified success. The game immediately became more interesting and a wide variety of strategies came into existence. The ‘wild spy’ was born. This was a spy that ran about the screen like a lunatic at high speed, talking to every suburbanite in existence. He was hard to keep from whacking into agents, but he gathered a lot of points. Also, the ‘slow spy’ was born as well. This was a spy that basically was stopped in his tracks. He didn’t get a lot of points, but he was much easier to defend.
- Collecting Agent bodies: When you kill lots of agents, their bodies pile up. We thought we could include a ‘coin collecting’ activity by letting the player click on bodies and gain points. This worked slightly, but wasn’t a great addition.
- One spy at a time: This isn’t either good or bad, but when players started learning to beat the game they began adopting the tactic of only having one active spy at a time. The game was too confusing to play well, so players reduced the complexity of the game until it was manageable. This demonstrated clearly the important of building focal points into the game.
What went wrong:
- Physics for the sake of physics: In the initial implementation of dragging spies in a direction, your motion took into account the spy’s current momentum. On paper this was a great idea that sought to add that ‘tactile’ element to the verb. In practice, the dragging felt sluggish. Spies were harder to control than necessary. One possible solution was to remove the momentum factor and see if it improved the feel of the dragging action.
- No short term penalty for failure: If a spy was caught, the abstract evidence meter went up but that didn’t really affect things much. Also, a spy could blast through several agents in a second making agent encounters rather hectic.
- Coin collecting had no risk element: You could leave the agent bodies on the ground forever and collect them when you got a spare moment. A better solution would be to have them fade out after a period of time so that there is some risk of not getting the points.
With version 5, we finally are at a game where players can demonstrate mastery. It is horribly unbalanced and still need some major additional systems. However, I can sit down and happily play for 5 minutes without feeling frustrated.
What went right:
- Removing unnecessary physics improved the tactile feel of dragging: One variable tweak later, the control of the spies became much more responsive.
- New strategies that improved the focus of the game appeared: The surprising side effect of improved dragging was that the ‘slow spy’ technique became the dominant way to own the game. By clicking on a spy twice, you could stop it dead in its tracks. Gather enough little spies up in the corner of the screen and you had a highly defensible position against the wandering agents. Also, you ended up being able to farm major info points by clustering so many spies and suburbanites in close proximity. Finally, we were starting to get away from a series of random events towards a player created environment that had interesting strategic implications.
- Agents convert spies: When an agent runs into a spy, the spy is converted back into a suburbanite. This makes the player feel that they ‘lost’ a spy. We can augment this in the future by given the player a limited number of spies. The first benefit here is that there is now a more immediate result associated with the player letting a spy and an agent interact. This tightens the risk / reward cycle. The second benefit is that focus is improved. A spy that gets in trouble is lost and the player can move onto a new activity. This clearer definition of the beginning of an activity and the end of an activity helps clarify the game play.
What went wrong:
With vesion 5, the gameplay is starting to open up and we have easily a dozen different paths that we could go down to actually make the game fun. This is typical of a prototype. We add and add and add until the game becomes quite interesting. At a certain point, we need to ask ‘What elements are the most fun?’ and trim back the various gameplay experiments to focus on those items. A prototype should naturally grow and shrink in complexity as you try a bunch of things and then boil the gameplay down to the essentials.
- Long term level structure is non-existent: Levels never end. We need to add in some meta game mechanics that govern longer term risk / reward sequences. We can set up goals for each level and start sequencing challenges. This will give the player a feeling of progression.
- The evidence meter is bogus: Though it was in the original design, the evidence meter doesn’t seem to add to the game play. Each evidence level should effect the map in some way and the player should be able to manage the evidence level.
- Obvious player tactics need to be countered: If the clustering of slow spies is a fun technique, we should make it challenging. What if agents targeted your ‘base’ with more intelligence?
- Support for focused gameplay: The base is a fun concept that really focuses the gameplay. However, it is one that emerges from the core mechanics. Should we build this concept into the game so that it is supported and more obvious to new players?
- The tactile feeling of the game could be improved: Dragging still doesn’t feel perfect. This could be improved with some better dragging code. Also, we can’t forget to improve the Agent killing mechanic.
Iterative design in action
SLOA is by no means fun quite yet, but we are getting there. Every iteration (some of which take no more than 5 minutes to create) demonstrates the same basic methodology.
- We create a theoretically enjoyable prototype.
- We play the game
- We record what works and has potential
- We record what doesn’t work
- We propose some changes and implement a new prototype. Wash, rinse, repeat.
We are still only a few iterations into prototyping SLOA. We could put in a dozen more before the gameplay gels enough to start building out content. Even then, the game won’t be polished until close to release. Game development is a fundamentally iterative process where the game design is just the beginning.
I’d like thank Harold for all his hard work. I like to say that a game design is like a movie script. It only shines in the hand of a great director. For SLOA, Harold is acting as the director of the game development and the real genius of the final product will be the result of his inspired implementation decisions. It is a wonderful process to observe.