[This in-depth postmortem from the creators of the latest Tomb Raider title, originally printed in Game Developer magazine earlier this year, explains exactly what went right and wrong during the making of the critically acclaimed franchise update.]

The concept phase of Tomb Raider: Underworld began while its predecessor, Tomb Raider: Legend, was in final QA and nearing submission. Previews for Tomb Raider: Legend were very encouraging, and we felt that there was still plenty of unrealized potential to tap in the existing feature set.

Enough so, the reasoning went, that we could focus on content and leveraging existing functionality to develop a bigger and better Lara Croft adventure in less time. In many ways this is what the team accomplished, but as is always the case in game development, reality was more complex than we anticipated.

It is particularly interesting to note that much of what went wrong in development involved pitfalls that we anticipated but still fell into despite our efforts to avoid them. It's important to note that most postmortems talk about "What Went Wrong" and not just "What We Did Wrong" because sometimes you make mistakes, but other times you suffer from acts of god and do your best to cope.

A Game Developer article last year ("What Went Wrong?" December 2008) specifically questioned why game development seems to make the same mistakes over and over. In light of that, some of the "wrongs" will be discussed in terms of how our methods to avoid known issues fell short.

What Went Right

1. Long Alpha

We scheduled an unusually long Alpha to deal with unresolved pre-production issues and to set ourselves up for a shorter Beta to make room for more polish time at the end. This paid off handsomely with respect to art production, and it's one of the reasons the game looks so good. We also recovered from a fair amount of design deficit that had carried over from pre-production, and on the code side we managed to get most of our core functionality up to scratch.

Another thing we did right during this long Alpha was to have multiple scope reductions. This was our first "next-gen" title, and we repeatedly underestimated issues of complexity. We would assess and determine that the game was too big, and then cut enough content to bring it under with margin to spare. Then two months later we would see that we were again coming in too big, necessitating further scope reductions.

The game was designed to be able to handle this degree of reduction, as seen by the fact that almost all the features and areas originally planned for the game made it into the final version, only smaller, and connected to each other in fewer ways. We managed to reduce the scope by trimming branches everywhere without having to uproot any of the trees entirely.

2. Focus

When making a sequel, producing a game just like the last one with slightly different content and a few more features is an easy mistake to make. Despite the fact that we had some significant new goals, like making the traversal less linear and bringing back more free exploration, we almost fell into this trap. But many on the team saw that we needed an additional focal point to both rally the team around and to use as a unique selling proposition for the game. The result of this was the concept of epic exploration puzzles.

Making large-scale in-game devices and areas with multiple layers of connected puzzles gave the game an exceptional expression even compared to previous Tomb Raider games, and it also gave us a litmus test for spending production effort across the game. This led to the creation of a sub team devoted to these puzzles, which proved to be complicated constructs.

While we would have been better off had we foreseen this need and planned for it properly from the start, the fact that we saw the growing concern, created this special sub team, devoted more staff and resources to it, and assigned a dedicated producer was an example of how well we solved unforeseen problems in development.

3. Production Flexibility

Even though it was hard to significantly redirect the juggernaut that was Tomb Raider: Underworld, we made a lot of successful changes when confronted with breakdowns of production processes. We started out thinking that we knew the best way to design the ruined jungle gyms that form the bulk of a Tomb Raider playground, based on lessons we had learned from the previous game, but reality confounded us again, and we made changes accordingly.

One clear example is related to level design and art. From the beginning we all agreed that it was critical to have both disciplines work in tandem from day one to make environments that were fun, beautiful, and credible. Prior outings had either started with designgenerated block mesh, leading to geometry that was extremely difficult to turn into plausible ruins, or with beautifully architected tombs that did not provide fun climbing opportunities or properly authored gameplay.

Our solution was to pair level designers and environment artists together by location in the world; such as Mexico or Thailand, and have them work together, iterating on environment architecture to make it both fun and aesthetically appropriate.

While this proved to be the right direction, many of our earlier processes didn't work out. But our willingness to accept the reality that these initial attempts were flawed allowed us take the big step of changing production workflows in progress, often multiple times.

It sometimes felt chaotic and frustrating to change process so frequently, but had we stuck with broken paradigms the situation would have been far worse. In the end, our production methods weren't perfect, but they were superior to what we thought would see us through from the outset.