Gamasutra: The Art & Business of Making Gamesspacer
Postmortem: Sony Bend Studio's Uncharted: Golden Abyss
View All     RSS
June 25, 2017
arrowPress Releases
June 25, 2017
Games Press
View All     RSS

If you enjoy reading this site, you might also want to check out these UBM Tech sites:

Postmortem: Sony Bend Studio's Uncharted: Golden Abyss

November 8, 2012 Article Start Previous Page 4 of 4

3. Documentation

Something we continually struggled with during development was how to document features and how to keep team members up to date on changes. It was not a new problem, and one we'd been trying to solve for a few projects.

For years we had used the "Master Design Document," a massive Word document that became the game's development bible, often running to hundreds of pages with cross-linked documents (when Word still had that functionality.)

We moved away from the bible for the same reasons that many teams have: no one ever read the thing, and it was hard to keep up-to-date. For the last two projects, we've moved more towards wiki and email, with mixed success.

Wiki can be a powerful communication tool because it seems more accessible than Word docs -- wiki pages tend to be less dense, with more emphasis on bullet points, images and diagrams; information is broken up into smaller chunks, with easily accessible links to related information. Team members seem more willing to surf our web to find what they need.

Early in development we used the wiki for all of our designs: worlds, levels, features, AI, weapons, etc.

An early wiki page shows how we tried to make information more accessible by using lots of bullet points, links and images. This page is taken from 2009, when the game was still a hub. (Click for larger version.)

Later, the wiki was used almost exclusively to track lists of things that needed to be fixed: Each level had its own wiki and team members were asked to check and updated these lists daily.

As we neared important milestones, the wiki was used to track "to do" lists. (Click for larger version.)

We also used the wiki to track art, design and engineering global tasks, production milestones, and more.

Production for the entire game was tracked via wiki. From this page we could see a snapshot of every important feature for every level in the game. (Click for larger version.)

But the wiki has its problems. For one thing, wiki pages are hard to edit -- wiki markup may be useful for a lot of things, but writing creative design docs isn't one of them. Late in the project we got the ability to edit pages using Word, but it was still hard to make sure formatting was correct, and at the time, adding illustrations wasn't easy, nor could you link documents via Perforce to ensure that everything was current. And though wiki pages were easier to access than Word docs (which had to be updated via Perforce), there was still no way to force team members to actually read them.

Equally troubling, though wiki pages were less dense than Word docs, they still tended to get out of date at an alarming rate. By the end of development, the Uncharted: Golden Abyss main wiki page had 171 child pages, each of which had dozens of children of their own, which had additional child pages, ad infinitum. Almost all of these pages were out of date well before alpha. Keeping information up-to-date became a full time task that no one had time for.

This is one of the reasons we came to rely on email. Emails can be effective because not only are they as succinct as wiki pages, they are almost guaranteed to be timely: if I read an email sent today, its information is almost guaranteed to be current. We used emails to generate to-do lists, create action items, track assignments, discuss issues, etc. Our Principal Environment artist used email to send out his "FixPix"; these were compilations of graphic issues, bugs and polish requests that Francois made for each level of the game. They proved to be an effective form of communication because they were almost entirely visual, making miscommunications between lead and artist very rare.

FixPix like this were used to visually illustrate graphic bugs and to request polish. (Click for larger version.)

When issues were more complicated, emails often turned into protracted chains that devolved into buck-passing and finger-pointing. Here's an example from 2011: the original email is a list of issues found while playing through the Shanty Upper Village level. The text in red is a team member's response, and you can see how the thread got out of hand from there. The obvious problem with email is if a problem requires team interaction, asynchronous communication probably isn't the most effective solution. It's too easy to punt the issue back and forth, or play the "blame game."

The perils of email: it's too easy to take comments out of context to play the "blame game" rather than work together to solve problems. (Click for larger version.)

For our current project, we're still using wiki (until we find a better solution) but we've made a conscious effort to move away from email. We spend a lot more time in one-on-one meetings discussing issues face-to-face, using email as a follow-up -- and there are far more ad-hoc meetings in common areas.

We also got some great advice from Brian Fleming at Sucker Punch Productions, and now have weekly meetings with the entire team. We've broken the team into four groups of roughly 10-15 people each; Chris and I, along with our production manager, meet with each group for a half hour every Monday morning, without exception. These meetings are strictly for information gathering: Chris and I give them information about the state of the studio and project, our production manager gives an update on what the rest of the team is doing, and team members give us information on their current tasks. All of this is designed to ensure that folks are reading the wikis, are working on the correct tasks, and that we're all moving lockstep toward the next milestone.

4. Scope

Bend Studio has always developed third-person action games, whether the emphasis was on Syphon Filter-style stealth action or Resistance: Retribution-style sci-fi action, but we completely underestimated the amount of additional content required by an Uncharted game. As we finished prototyping and began actual production, we realized this, and started making big cuts in the design and story. Here's some examples of things we cut before production began:

The Chindi

Much has been made of the fact that there are no supernatural enemies at the end of Uncharted: Golden Abyss. Originally, there were going to be. In an early draft of the story, Drake reached the underground city of gold and discovered that all of Guerro's men had been killed -- victims of what the local superstitions called "Chindi" (who would turn out to be ancestors of Quivirans, mutated by generations of exposure to the radioactive gold).

Partly we cut the Chindi because we knew we had our work cut out for us just nailing Uncharted-style combat with regular humans, without the headache of a completely different set of AI, animations (which we didn't have), combat mechanics, character assets, etc.

This slide is from an Uncharted NGP presentation dated January 2010.

Cutting the Chindi also helped differentiate our game from Uncharted: Drake's Fortune. Reading an early draft of the story, Amy had expressed some misgivings that the Chindi were pretty similar to the Descendants. Later, when I asked if the Uncharted formula required a supernatural ending, she said absolutely not. So the Chindi were cut. Some players missed them and said we weren't Uncharted without the supernatural ending; others thought its absence made the game better. We weren't making a judgment call either way -- we were merely trying to trim the schedule.

The cut didn't hurt gameplay too much, we felt, because we were able to rewrite the story (we were still in pre-production, remember) and have Dante show up toward the end of the game with an army of mercenaries. The mercs provided combat scenarios that contrasted nicely with Guerro's thugs. The mercs had better and more powerful weapons, flak jackets, and they looked different: more military, more organized. While the Mercs didn't offer unique gameplay mechanics the way Uncharted's supernatural enemies do, they helped our schedule immensely, because we could share behaviors and animations across all enemy types.

Isla Providencia

In the original story, Chase and Drake find Esteban's Crypt in the Sete Cidades Retreat only to discover that the Esteban's Sword (another important MacGuffin) had been carted off centuries earlier by none other than Henry Morgan, who hid the sword and other plunder in his secret hideaway on an island in the Caribbean. Drake teamed up with another old friend, Harry Flynn, to help retrieve it.

This page from a June 2010 draft of the story shows how Drake would have interacted with Harry Flynn on a deserted Caribbean island.

We loved the idea of showing how Drake might have screwed Flynn over early in their careers, perhaps explaining Flynn's behavior in Uncharted 2: Among Thieves. But we realized that cutting this plot element would allow us to cut an entire world's worth of characters, mocap, sets, movies, and gameplay. At the time, I hated to see Flynn go, but as it turned out, we didn't cut nearly enough. And thankfully, Uncharted fans don't love Flynn nearly as much as they love Sully.

The Highlands

This world would have taken place between Chapter 27, the last rainforest level, and Chapter 28, the first Quiviran level. Drake and Sully were to traverse dangerous cliffs, narrow trails and more rickety bridges as they fought their way up to a Machu Picchu-type city where they would have discovered the entrance to the lost city of gold.

High-level design and concept art for the Highlands, another world that would have starred Sully. (Click for larger version.)

This world didn't get cut until just before final production began, so we had concept art, maquettes, layouts, as well as playable gray box levels. But as production began we were still worried about being able to pull off the Naughty Dog-style open spaces, cliffs, and vistas.

As it turned out, our environment and tech team proved that we could do vistas just fine, and I can only imagine how spectacular the traversal in this world would have been. At the time we were sorry to see the Highlands go because we worried the player wasn't going to get enough Sully time, but that turned out to be a non-issue, mostly because the rainforest world almost doubled in size during production (which was, no pun intended, a much bigger issue).

Shanty Avalanche World

The last major cut came after the script was locked and full production had begun -- but the cut was not the result of our having realized our game had scope issues. It was cut because of a tsunami.

Originally the shanty avalanche world took place between Chapter 9 (the last shantytown level) and Chapter 10 (the "side scrolling" level where weaponless Drake and Chase run for their lives). In the avalanche levels Drake and Chase would have been trapped in the middle of a massive avalanche that was sweeping the shantytown down the mountainside (the bridge was originally a dam). They were to backtrack through the shantytown, but would now face a torrent of mud, struggling to stay ahead of the destruction all around them. It would have been a spectacular action set piece. Like the Highlands world, we had already completed maquettes, layout and gray-box levels -- the world was ready for production.

In this gray-box playable layout, Drake is contemplating a traversal path that will keep him out of the torrent of raging mud below.

Then in March of 2011, a disastrous earthquake and tsunami hit Japan and we, along with the rest of the world, spent several days watching images and videos of the awful destruction as entire towns were swept away by torrents of muddy water. In the aftermath of this, we worried that maybe the avalanche world would feel a little on the exploitive side, and that at some point we might be asked to remove it.

Something similar had happened to our studio more than once: Our opening world in Syphon Filter: The Omega Strain was originally set in Toronto, Canada -- until the Canadian press found out that "Sony was creating virtual terrorist attacks on their soil." Mere weeks before alpha, we had to scramble to change every reference from Toronto to the fictional "Carthage, Michigan." We also had our original cover and marketing pulled from Syphon Filter 3 because our story involved a terrorist bombing in Washington DC... unfortunate given our September 25, 2001 release date (just weeks after 9/11). We weren't going to take any chances with our avalanche, and the world was cut.

But even with all those cuts, Uncharted: Golden Abyss was still a huge game. Partly this was because some levels had to be divided in half in order to work with our streaming engine. In any other franchise this might not have been a big deal: divide a large level and you end up with two smaller levels, right?

But with Uncharted, that's not how it worked out. Because our streaming required a cinematic to mask asset loading, each level required a transition cinematic -- at one point we had trimmed over 20 minutes from the script, but by the time we reached gold master, we had added all those minutes back in, and then some. And it turned out that the amount of polish wasn't cut in half, but doubled; it was hard to tell artists, "look, this level gets only half your time and attention because it's not really a full level." Development just doesn't work that way.

Misunderstanding the correlation between level size and asset needs was a mistake we made from the start. Because we were worried about memory and streaming, we had made the decision early on to create more, smaller levels, thinking they would be easier to build, load and polish. But Uncharted is a dense experience in terms of assets per square foot of environment. Even small levels tended to get filled up with AI nodes, VO triggers, traversal edges, unique action moments, special animations, collectible hiding spots... all of which needed art, animation, audio, scripting and sometimes code support. And as the layouts got fleshed out, each experience grew in the telling until we were looking at a massive amount of assets that needed to be polished.

The "world" board in my office. Each of the 34 small index cards represents a level, organized chronologically into the game's five worlds. Each of the colored post-it notes represents some aspect of asset polish that is complete for that level.

Resistance: Retribution had only 23 levels -- at a PSP level of quality. Uncharted: Golden Abyss had 34 next-gen quality levels, each of which had five times the content of any PSP game we had ever made. What were we thinking? We were trying to make something akin to a next-gen quality console game with a core team that was a third the size of a Naughty Dog or Santa Monica Studio. It was insanity. I shudder to think what would have happened if we had not cut those three additional worlds: the game would have been almost a third again as large, maybe another 15 levels. We'd probably still be working on the game today.

In the end, we needed to cut even more, but that's where the reality of the locked script really hurt us. The only way to have fixed the scope issue, in hindsight, would have been to tell a more modest story to begin with. Maybe there should have only been one antagonist instead of two; maybe the whole Perez backstory was too expensive and unnecessary; maybe we didn't really need such a long trek through the rainforest (thankfully we didn't cut that -- yay Sully!) And maybe a handheld game didn't need two hours of cinematics (twice that of Resistance: Retribution.)

Given the chance to do it again, we would have spent much more time up front vetting every aspect of the experience, negotiating a target game length. It was an expensive lesson, given the amount of overtime we had to work to finish the game.

5. Crunch

From January to December of 2011, the team worked pretty much nonstop, many of us putting in evenings and weekends for months at a time. Every evening we tried to offer a catered dinner so the team could maximize their time and get home as early as possible. Our office manager tells us that we ordered over 158 overtime meals during that crunch. There's been at least some crunch on every project I've ever worked on, but this was harsh by any standard.

As studio directors, Chris and I were responsible for the schedule and hence the crunch, and we took that responsibility very seriously. Overtime on past projects could be blamed on poor planning, mismanaged design, lack of proper prototyping, unclear milestones, uncontrolled iteration... the usual suspects. But not this time... at least not entirely.

Our crunch in Uncharted: Golden Abyss was almost completely attributable to one thing: ambition. As noted above, we didn't do a very good job of managing the scope of the game, which for some players ended up being a 10 to 12 hour experience -- compare this to God of War: Chains of Olympus, one of the highest rated games for the PSP, which clocked in at six or seven hours.

During the project postmortems, "Managing a healthy work / quality of life balance" came up as a top concern for almost everyone in the studio, Chris and myself included. Our families didn't see us a lot in 2011, and we vowed to do everything in our power to keep that from happening again. Some of the steps we've taken include:

  • Negotiating as reasonable a schedule as possible, given our studio's burn rate and budgets
  • Holding weekly meetings with every individual on the team, where issues of scheduling, scope, progress and productivity can be raised directly
  • Vetting the scope of our next project up front in an attempt to find that balance between the optimal player experience and the time we need to create it


A year before we shipped, we gave the team a pep talk, pointing out that for many of us, Uncharted: Golden Abyss might be our only chance, ever, to work on a premiere launch title for new hardware, a title that could define that hardware for years to come. I believed it then and I believe it now: chances like this only come along once in a career.

Even though development was stressful and meant many months of hard work, I'm glad our studio took on the challenge and rose to it. Every project is a learning experience and developing Uncharted: Golden Abyss taught us a lot. We learned important things about planning and scoping. We learned about the immeasurable value that early and constant focus testing can add. We learned how to begin real development even while key parts of the project are in flux, by defining important, story-independent features up front. We learned a lot about next-gen graphics and non-traditional controls. We learned a lot about story, character, presentation and pacing. But most important, we learned a lot about how we work together as a team, how we react to stress, how we communicate with each other, how we get things done. The development of Uncharted: Golden Abyss was far from a perfect process, and the game isn't perfect either, but as a studio, we're proud of what we accomplished, and are excited to put these lessons to work on our next game... which will be even better.


After reading this postmortem, one of my senior designers pointed out that it is three times as long as other recent postmortems. I guess we still have more work to do on our scoping issues.

Article Start Previous Page 4 of 4

Related Jobs

UBM Tech
UBM Tech — San Francisco, California, United States

General Manager, Game Developers Conference
Infinity Ward / Activision
Infinity Ward / Activision — Woodland Hills, California, United States

Producer (Temporary)
Square Enix Co., Ltd.
Square Enix Co., Ltd. — Tokyo, Japan

Experienced Game Developer
Visual Concepts
Visual Concepts — Novato, California, United States


Loading Comments

loader image