Postmortem: Frozenbyte's Trine
June 3, 2010 Page 4 of 6
What Went Wrong
1. Design, budget and schedule -- do they ever match?
The original budget for Trine was 30,000 euros. The plan was to buy a lot of ready-made assets to speed up the development -- graphics quality was not a concern at that point. We weren't sure if it would be released as a Frozenbyte game or as something else; a different label was discussed and it could have ended up as a true indie release as well.
When the project was elevated to "proper status" and the budget was increased to 300,000 euros, it was clear the original plan needed revising. While the broad scope of this change was successful and didn't have many problems, we kept making small design changes to the revised design down the line.
It is hard to say whether or not this was a bad thing as such because each change made the game better, but the journey became very stressful to everyone who was involved in the design of the game.
In many cases it was "design by committee" -- only to be overruled by the design director further along, and despite those final decisions being key in making Trine a successful game, they caused a lot of grief to the people involved.
Everyone was passionate, especially the original designer/producer/programmer who also had money invested in the project, and passionate people do not give up on their ideas easily. Each design change also had to get communicated to the team and sometimes the confusion stemming from design arguments would cause uncertainty as to what the final design was supposed to be.
When the game moved from PC-only to PC and PS3, another slew of changes was required. The budget was almost doubled once again, and the team's ambitions were set even higher. And even after this, the budget continued to increase in portions of 100,000 euros, and each time some of the previous work was rendered useless, and a new bar was set for quality.
The budget finally ended up at around 800,000 euros at gold master time. Support, patches and other post-release activities have added their own share. As our contract with the publisher was more akin to a distribution one than a full-blown publishing contract, our increased ambition was not in any way reflected in the milestone payments we had agreed upon, and ultimately we ended up funding two-thirds of the project on our own.
2. Unrealistic scope of design and production
Frozenbyte's previous games, Shadowgrounds and Shadowgrounds Survivor, both suffered from horrible delays in production. Shadowgrounds missed its original release target by two years (delay of 100 percent), and even Shadowgrounds Survivor -- which was supposed to be a quick but professional job -- by six months (delay of 50 percent). But we thought we had learned from those experiences, and everything was fixed -- as if by magic -- and we wouldn't need to improve anything or put more focus to scheduling.
The project's budget and resources were in constant motion. Instead of cutting down design and scope in the early stages, we pumped up production resources with a "yes we can" attitude and gradually added more people to the project. Luckily, our empty bank account kept us from going overboard, otherwise we probably would have made the fatal mistake of adding too many people to a delayed project.
Programmers had too many items on their to-do lists, and instead of rebelling and throwing scissors (like the artists might have), the programmers cut corners in their code and crammed in as much as they could in as little time as possible. This was known by the management -- and was essentially a continuation of practices from the early days of the cash-stripped company -- but the problems started to become evident with Trine, especially during the testing phase where weird and "unfixable" bugs occurred.
We neglected proper design documentation, production plans, programming plans and sometimes even art supervision. Everything happened at the last minute. In hindsight it's great to reminiscence how the team pulled together, but it certainly did not feel like the right way to do things at the time.
Because of indecisiveness and procrastination, the final story -- not just a revision of it, but the whole thing -- was written two days before recording was scheduled to begin in March 2009, a few months before the scheduled gold master date. The night before recording, a new, important revision was written.
Another day would probably have helped and made the story more sensible -- we doubt most gamers understand the story about the three artifacts (and in fact, there's even a notable inconsistency in the narrator's loading screens -- we lost the plot in the wee hours of the morning ourselves.)
The state of the overall project and its overambitious design was finally understood in May 2009, roughly a month before the scheduled gold master. Some partially-developed gameplay features were cut, such as the third upgrade slot for the characters' skills, along with a graphically unpolished icy mountain level and other art content, as were many small gameplay features.
This also caused headaches down the line, for example in localization, as much of the text had to be revised and combined even though the game had already been completely translated once and VO had been recorded.
Screenshot of the inventory screen from early summer 2008, showing an early draft of the inventory itself, old character models/faces, six skills (final game has three) and five upgrade slots (final game has two).
Testing of the game started a couple of months before the scheduled gold master date and it was imposed on inexperienced, part-time trainees, who did a good job but could not fight against the reality of the situation. Despite this, our in-house QA was much more useful than the external QA the publisher had set up -- in the future we plan to use in-house QA and general testing sessions to a much greater effect, and start them earlier.
Nevertheless, development was more or less completed by mid-June and the PC version was released shortly after, on July 2, 2009. The PS3 version, however, fell into a QA mess.
Page 4 of 6