Postmortem: Saber Interactive's TimeShift
April 4, 2008 Page 1 of 8
[Saber Interactive's PC/console shooter TimeShift was released late last year after a "strange and convoluted" - but fascinating - development cycle, and in this exclusive Gamasutra postmortem, the creators detail what went right and wrong in its creation.]
Few games have gone through a development cycle as strange and convoluted as TimeShift's. After four years of development, three extensions, two publishers, three re-recordings of voice acting, three iterations of the story, two sets of discrete FMVs and multiple iterations of Saber's technology base, the game finally made it to the shelves for Holiday 2007.
The game started off as a small proof-of-concept demo in 2003, which we completed a few months after we released our first title, Will Rock. Atari picked it up a year later with plans to publish the game on the PC and Xbox.
The title showed a lot of promise early on -- and visually, it was ahead of the curve -- so Atari made the decision to move over from the Xbox to the Xbox 360. Two years into development, Sierra acquired TimeShift from Atari. Atari was having some cash flow issues and needed to sell off some of their projects.
The original plan after the extension was to give the product another six months to enable Sierra to improve on some of the production elements of the game. Dennis Quaid, Michael Ironside and Nick Chinlund were hired to add their voice talents to the game and extra effort was spent in improving visual quality and gameplay elements. Literally on the eve of GMC, Sierra made the rather unorthodox decision to give TimeShift another year of development.
While extensions are common, they are less so when a game is essentially ready to ship. This was not only a vote of confidence in the title, but also one in Saber's ability to turn the game into a AAA franchise.
The thought behind the extension was that TimeShift was built on a very solid concept, but one that could use additional refinement. It was also decided to add another SKU to the game -- the PS3. This meant that Saber had about 10 months to make vast improvements to the game on all levels -- visuals, gameplay and polish -- and also to write a PS3 engine and port the game to that console. While this was a tall task, we ultimately proved up to the challenge.
In the course of a year all FMVs, voiceover acting, story, the beginning third and ending levels of the game, characters, animations, vehicles and weapons were completely thrown away and redone.
Finally, in the fall of 2007, the new and improved TimeShift was released to the public. While the competition has been stiff, the game has really resonated with the gaming public and user feedback has been exceptionally good.
What Went Right:
1. Dealing With the Scope of the Project. Back in 2003, Saber had a small team of 25 talented people. It was enough to put together a working prototype which demonstrated our innovative time-control mechanics, but to handle a AAA project coming out on three SKUs, we needed to expand.
By the time we shipped the game, we expanded our studio to over 90 people, all working on the title. Undergoing such a major expansion was not easy. Talented people are hard to find anywhere, but even more so in Russia, where there are very few teams working on larger titles, and therefore there are very few people with the requisite experience to handle big projects.
Fortunately, our core team was comprised of the most talented game developers in Russia -- people with 10+ years of experience in the industry, including console work. This was the core team that worked together on our first game, Will Rock -- a game that was put out in less than a year from start to finish. As we kept hiring, most of the core team members were promoted to manager or senior positions, and headed up newly created departments within the company.
What was once simply the "art team", with a flat structure run by Dmitry Kholodov, transformed into a multi-discipline department that includes Concepting, Texturing, Lighting, Art Outsourcing, Asset Modeling, Level Modeling, UI Design and Special Effects -- all led by their own managers. Establishing these departments allowed us to build "centers of growth", streamlining production and communication pipelines.
When a new person was added to the team there was always a manager available who kept track of the tasks in a particular discipline and who was able to train a new hire.
While we knew it would be hard to find the right people, some other issues related to this rapid expansion were rather unexpected. Finding space for all of these new people was one of them. Because the project was ongoing, we decided to stay in the same office building rather than move to a larger office and disrupt development.
Eventually we ran out of space. For a while, we couldn't hire more people because our office was at full capacity. Fortunately, we were able to work with our landlord to relocate other tenants and take additional space on our floor.
As part of our move into next-gen territory we needed to improve our production pipelines and supporting infrastructure. During normal development, we used to make one build a month that was sent to our publisher.
That was enough to demonstrate the progress we were making. By the end of the project we were pushing 15 builds a day, every day (this includes two or three regional builds on each of three SKUs, for the full and demo versions of the game).
We had to establish a special Build Department which worked closely with our QA team to ensure that builds were made and tested daily so that the problems were immediately communicated back to the team.
We built a number of server racks with "hardware farms" to help us churn all this data on a daily basis, compiling code for all three SKUs, re-exporting art assets and assembling builds, compiling the shader database, and doing the uploads of the final assets back to the publisher. The process was working around the clock, and sometimes we were already working and testing the new builds while the ones from the day before were still uploading.
Page 1 of 8