Postmortem: Epic Games' Unreal Tournament

What Went Wrong

1. Bad timing

Many aspects of the game's timing worked against us. While the Quake 3 vs. UT hype increased our exposure, it also set a very hard deadline for completion. It was critical that we complete the game before Quake 3 was released. The media advantage belonged to id and we believed that if Unreal Tournament launched after Quake 3, we would be forgotten in the storm. At the same time, however, we were caught up in grueling contract renegotiations with GT Interactive. We did not want to deliver the completed game until we knew the contract would work in our favor. Many times during the development of the game we were promised that a resolution to the contract issue was close at hand. The team would race to reach a point where the game could be shipped, only to have negotiations drag on.

The gold master was delivered to GT days after a final contract was agreed upon. Unfortunately, the game hit shelves in November, pushing us very close to Quake 3's release date. While Unreal Tournament often performed better than Quake 3 in reviews, we believe that sales would have been much higher still had we released in October. Word-of-mouth is a powerful force and the extra month would have given us time to build a larger community before Christmas.

2. No central design document

While I am a big supporter of open, cabal-style design, I have to stop and wonder how Unreal Tournament would have turned out had we a strong initial design. It's quite possible that the game's weaker elements would have been much stronger if we had put together some concept art and focus material. In reviews, we have been criticized for not having enough variation in characters. If Unreal Tournament had had a library of concept art to draw from, we might have had more interesting alien warriors. The story is more or less nonexistent in Unreal Tournament, but at times we considered having in-game cutscenes as rewards for a player's progress. The idea was dumped, but a design document might have made it easier to visualize those scenes.

The Unreal Tournament development team felt that several of Unreal's weapons were a lot of fun. Here is a bot carrying the Shock Rifle, an updated version of Unreal's ASMD.

I suppose this isn't really a "what went wrong." It's simply more of a "what we should have done." I think it's important to think about the game in that light. Unreal Tournament is a very fun game with a lot of features packed into a short amount of development time. Those features were largely added through spur-of-the-moment decisions. A more unified approach to design would have allowed us to construct features that play on features, or even think of ideas we didn't have the perspective to realize. Epic will always be a very open, liberal company when it comes to the design process. If we develop a design document, we'll use it with the understanding that it can be modified at any time. That having been said, I think there is a definite positive argument for having some sort of central guide to everyone's ideas. Having the ability to sit down and look over the big picture is very valuable.

3. Co-development across two countries

Epic Games and Digital Extremes co-developed Unreal Tournament. The Digital Extremes team was located in Canada and Epic was located in the U.S. Epic supplied the programming team and a large group of content designers. Digital Extremes provided level designers, a sound guy, and texture artists. James Schmalz, the high-up man at Digital Extremes, contributed two of the game's player models and a lot of art. This co-development worked well for the most part, but near the end of the project it became very difficult.

During Unreal, Epic team members flew to Canada to work at Digital Extremes' offices. With Unreal Tournament, it became Digital Extremes's turn to do the traveling. Unfortunately, flying and driving back and forth every couple of weeks is a very draining experience. Many of the Digital Extremes team members spent several weeks away from their wives and girlfriends. Near the end of the project, they grew increasingly frustrated with the situation. To compound this problem further, Digital Extremes and Epic were attempting an expensive merger. As Unreal Tournament came to a close, it became clear that the merger would not happen. It was prohibitively expensive for a small company to move across the border. Many Digital Extremes team members already had apartments and plans for living in Raleigh, and the news of the terminated merger process was devastating.

In the Assault game type, players have to enter a heavily defended base and complete map-specific objectives to win. Assault was the most difficult Unreal Tournament game type to design, balance, and play-test.

Much to Digital Extremes's credit, the company quickly recovered and moved to its backup plan of developing its own game with the Unreal Tournament engine. Nonetheless, the process of co-developing the game had taken its toll on everyone. The ups and downs of the merger process had a negative effect on team morale. Had the co-development happened between companies more closely situated, it would not have been a problem.

4. Not enough artists

On the content side, Unreal Tournament was held back by the number of available artists. Epic's artist, Shane Caudle, is a supreme Jack-of-all-trades, creating skins, models, and levels of the highest quality. He spent most of his time working on new player models and skins for those models. Digital Extremes brought a few texture artists to the table, but not enough to create the huge libraries of new textures needed for the game. In order to supplement the skin and texture production, Epic turned to contract artist Steve Garofalo.

Steve Polge, our AI and game play programmer, made the bots understand the unique advantages and disadvantages of each weapon. Here a bot is moving in very close to use the powerful Flak Cannon.

Even with the additional help from external sources, the team was unable to produce enough new textures. Level designers who wanted custom textures for their maps had to make do with their own texturing ability. While the final texture and level count in Unreal Tournament is quite high, the levels would have been much more impressive had the team been able to act with full freedom. Since the completion of Unreal Tournament, Epic has hired both Steve Garofalo and John Mueller to strengthen the art team for future projects.

5. Visual Basic editor interface

The Unreal Tournament engine uses UnrealEd as its level design and content management tool. For several years, UnrealEd has used a windowing interface written in Visual Basic. The VB code is fragile and very old. Add to this the fact that nobody at Epic except Tim Sweeney knows or cares about VB, and you have a level design team that is stuck with a tool that's not easily updated.

Several interface bugs have plagued UnrealEd for some time and nobody on the team had the time or inclination to fix them. If we had a more easily extensible tool, the team would have been able to add extra features to the editor for level designers to use. As it stood, the editor was considered "off limits" for new features.

For our next few projects we will most likely use a new C++ editor that Legend Entertainment developed. For Unreal Tournament, however, we simply didn't have the time to work on a new editor. Fortunately, our time spent using UnrealEd taught us the dos and don'ts of tool design.