Virtual Villagers – The Lost Children Postmortem: Behind the Scenes on the Island of Isola

Americans are known around the world for taking way too little vacation time. If we are going to confine ourselves to a small room 51 weeks of the year, we might as well be working on a virtual life game that takes place on a tropical island paradise, right? For those who have not played it, Virtual Villagers:  The Lost Children is the second chapter of the Virtual Villagers series, a sim/virtual-life game based on the fictional South Pacific island of Isola.

Virtual Villagers is a game where the player manages, or perhaps micro-manages, the fate of a growing tribe of villagers. The player guides them through food problems, helps them to procreate, and leads them on a journey of discovery as they try to solve puzzles and unravel mysteries surrounding the island. The game is designed to be open-ended and continues even after all the ‘official’ goals and challenges have been met.

Some things that went right

1) Last Minute Features

Time and time again we try to stay on schedule, get to ‘first playable,’ then to alpha, then to beta phase. Once in beta, we are not supposed to add new features, but instead just tune and test the features we have in place. The problem is (in particular for our games which are notoriously hard to test all the way through) once we are in mid-beta, some of the game’s shortcomings emerge and reveal themselves for the first time. An example from VV: TLC is of our elderly villagers: as we tested, and restarted, and retested the game, we ended up with young villagers more often than if we had been playing in a natural way. During beta, when we finally ended up with more tribes of ‘oldies,’ it occurred to us that it was annoying to see ‘Elderly’ villagers moving slowly and behaving old, but looking too young. We added a new ‘older, grey-haired’ look for the elderly villagers during beta.

This is a pretty drastic addition to be throwing in, two-weeks before scheduled launch, but…well, we give up. We seem to add features that we love in each of our games’ beta phases, and it has always worked for us. The lesson? Don’t deny the fun just because you feel obligated to stick to the traditional rules of development.  

2) A Satisfying End Game

We had received many complaints about the ending of the first chapter of Virtual Villagers. We puzzled over this quite a bit. The complaints ranged from “Now what?” to “Is that all there is?” to “This ending sucks!” The ending came when the player solved the 16 mysterious puzzles, and then they were rewarded with a game-end illustration.

In VV: TLC we tried to telegraph more clearly to the player when the end was getting near. Instead of arbitrary linked puzzles, we have the player collecting 4 pieces of The Gong of Wonder. As they collect each of the 4 pieces, the end draws nearer in their mind so that when it is completed the ending does not blind-side the player.

As a final enhancement to the end/post-game phase, we tried to take a page from the World of Warcraft book and add some content that was only present for people who had ‘finished’ the game. In VV: TLC, it comes in the form of special, rare, heavy-duty Island Events (these random, periodical text events that are inflicted on your villagers), that only appear for people who have completed all 16 puzzles. I suppose the verdict is still undecided on the effectiveness of that feature. It takes a long time to get good feedback on such late-game elements.

3) A World Coming Alive

By the time we released VV: TLC, it was the 4^th original game set on the fictional island of Isola. Each new release forced us to flesh out the back story, understand this mysterious island better, and make the world more complete. With this second chapter of Virtual Villagers, we really were starting to feel like we knew the local, we knew the story and history of the island. This caused the story to just flow naturally, and helped us to stay very consistent. It can be helpful to develop the back story beyond what is expected to be exposed to the player. Inconsistencies tend to leak out in subtle ways and can really break the player’s immersion in the game.

4) An Evolving Casual Framework

After we released Fish Tycoon, we finally bit the bullet and invested some time in our game framework. We spent a great deal of time tuning it for low system requirements, adding support for specialized drawing/rendering (high-speed alpha blending, additive drawing), and adding really solid hardware abstraction to allow for a ‘free’ Macintosh build from our game source code. As we finish each product, we roll the framework over into the next, so that each game can benefit from the enhancements of the previous one. The basics of this include the rendering layers (for Mac and PC), as well as what we call ‘casual game best-practices’ such as save-slots, Windowed/Full-screen toggles, music track selection, etc.

With Virtual Villagers, we went an extra step and integrated higher-level services into our framework such as environmental sound management (we can place sounds on the game map, and they are automatically played only when they are in view or near the player’s view), highly flexible animation services (looping and non-looping animations and ‘sparkles’ that can float, rise, fade, fall, etc), and a simple scripting language for villager behaviors and AI. These features took a lot of time but have really enabled us to create a factory for these types of games and beginning with VV: TLC we are enjoying a very high-level work environment for creating these games. It makes it much more fun to work.

5) A Virtual Office (The Light Side)

Here at Last Day of Work, we use a virtual office. This means that there is no physical office, our entire team works from wherever they wish, and communication is done with IM, Skype, Webcams, and email. While this is not feasible beyond a certain size team, when it works it brings some really amazing benefits to the production. One thing it does is allow us to work from anywhere, and we do. We spearheaded much of the production of VV: TLC from Italy over the fall of 2006. This approach, of course, reduces overhead significantly. It increases the length of everyone’s work day by removing commuting time. It also adds flexibility to whom we can choose to work with by not requiring people to relocate and allowing our passionate and talented independent artists and engineers to remain truly independent. It is not without its pitfalls, however, and I have also included the virtual office in Things That Went Wrong.

Some things that went wrong

1) Critical Path Overloads

We are huge fans of our own talented artists. On VV: TLC we enthusiastically over-allocated art production to our lead artist Michael Grills, who hand-painted all the in-game illustrations, all the UI, the game play map (in all of its different states), and so much more. We did try to estimate the timeline and how long everything would take, but it was not a roomy schedule and everyone felt the pressure. In the end, Michael came through, as he always does, but we think it would have been smarter to divide the work among additional artists.  

2) Repetitive Stress Injuries

One of the new game mechanics that we introduced to Virtual Villagers:  The Lost Children involves seeing a rare collectible object appear on the map, then hurrying to find a child and dragging the child to that object before it vanishes. It is intended to be a quick thing, and kind of tense. What we did not foresee was the degree of compulsion experienced by many players (including myself) to never miss one of these objects when they appeared. This resulted in lots of map scrolling, which is done with a dragging motion with the mouse. Very early in beta we had testers buying new control devices and struggling with repetitive stress injury. We had to put in keypad/number navigation in the 11^th hour to give the more ‘enthusiastic’ collectors an alternative to dragging the map around.

3) Insufficient QA time

We have always tried to schedule not more than 4-5 weeks of beta for our games. In addition to being predisposed to feature creep (see “Last Minute Features”), these games are notoriously hard to test because many of the late-game puzzles only emerge after a week or more of playing. Attempting to simulate those conditions is helpful, but not perfectly reliable. In the case of VV: TLC, we were still finding significant bugs in the final week of the testing schedule, and we continued the beta through the initial 2-week soft launch of the game. Just for the sake of polish, and because of the nature of many of the emergent aspects of the game (like the random Island Events that popup roughly once a day), we could really have used several more weeks of testing.

Determining how we could do this and not tire out our enthusiastic beta group would be a separate challenge.

4) Excessive crunch period

We had scheduled this project, and then made some marketing and publishing commitments. Pretty routine, right? We ended up with an insane 6-week crunch, complete with Red Bull and panic. It also did not make the job any easier for QA. In retrospect, we would definitely embrace a few approaches that are becoming popular in larger studios. Agile Development is an approach to development (and other production disciplines) that is best suited to larger teams, but one thing we really should have done was approach the project as a series of vertical slices. Each slice would be a set of tasks, scheduled as a 2- or 3-week ‘sprint.’  Then, prior to the delivery of this chunk, it goes through its own QA and, if necessary, its own mini-crunch of maybe 1 or 2 days. This approach would essentially break up our 6 weeks of crunch into multiple mini-crunches, and we think it would improve the quality of work and the quality of life.

5) A Virtual Office (The Dark Side)

I mentioned the Virtual Office as something that went right, and we are firm believers in it. However, it has some very serious downsides that continue to challenge us. The first and most significant drawback is that more effort must be made to communicate. It is easy for the team to drift apart a bit between meetings and for people to get off track or go after different visions. While we save a lot of time by having no daily commute, maybe we end up spending more time having meetings and making the extra effort to communicate. The other real problem that comes with a virtual office is the inability to grow beyond a certain size. I have heard some people suggest that a team of 10 is about as big as you can get before you really need to be geographically in the same location. Our efforts to grow beyond that size (with a virtual office) have resulted in massive loss of efficiency, miscommunication (resulting in wasted work), and an overall sense of chaos.

Statistics and Info

In Virtual Villagers, the players can view their own game stats, which for the most part are fun, trivial things like “Food gathered,” “Number of Totems Made,” and “Number of Triplets Birthed.” In the spirit of that, here is some random information and statistics from the development of Virtual Villagers: The Lost Children

Time to develop VV: TLC, excluding legacy code and framework:        ~5 months

Time to develop VV: TLC, including all legacy code and framework: ~20 months

Number of people involved:  Full-time: 4, Part-time 3+

Number of beta testers: 80

Weeks of beta: 5

Number of beta posts/reports: 8641

Maximum number of open bugs: 251

Rendering layer (PC): Gapidraw

Rendering layer (Mac): PTK

Final compressed size of game: 29mb

Final size of game executable (no assets): 1.05mb

Number of possible unique-looking villagers: 1800

Main computers with catastrophic failures: 2

Main backup drives with catastrophic failures: 1

Wrists with catastrophic failures: 1

About the Author
Arthur Humphrey is founder and CEO of Last Day of Work, a San Francisco based developer of games for the casual audience. Last Day of Work has created an uninterrupted string of hits, including Fish Tycoon, the popular Virtual Villagers series, and the upcoming Plant Tycoon. Arthur started creating games on his VIC-20, including a lovely Pengo clone for ‘public domain’ release. In his spare time he plays even more games, both casual and core. Arthur holds a Bachelor in Business from UCLA and a Bachelor in Computer Science from Hayward State University, and has a level 70 Druid in World of Warcraft.