Choosing The Game Engine That Can

Selecting a game engine used to be a snap when there was only a handful from which to choose.

But today, developers have a plethora of possibilities -- from rolling their own to licensing an engine originally built for a specific game (like the Unreal Engine) to licensing a generic engine (like Emergent's Gamebryo).

To add to the confusion, says San Francisco-based independent games consultant Mark DeLoura, during the past few months the number of game engines available has suddenly increased from a relative few to something that needs a spreadsheet to keep track of.

Which is why, in February, he sent out a survey to industry executives asking for their feedback on the use of game engines. Last month, he shared some of his results on his Gamasutra blog.

Faced with such a grand buffet of choices, how do successful developers make their selections?

Alex Seropian considers himself "engine agnostic," having had all three experiences -- first, creating the Halo Engine at Bungie, then -- at Wideload -- using the Unreal Engine for a very un-"Unreal"-like Hail To The Chimp, and, finally, employing the Gamebryo Engine for the soon-to-be released Texas Cheat 'Em. Seropian is president and founder of Chicago-based Wideload Games.

But what was there about each project that influenced him to make the decisions that he did?

Choosing to create an engine -- the Halo Engine -- for Halo back in 1991 was a no-brainer, recalls Seropian. "Eighteen years ago you really didn't have many choices," he says. "Besides, the team at Bungie wanted to build Halo around some very particular technology features and core mechanics, and that necessitated our constructing our own game engine."

The project -- building both the engine and the first Halo game -- took almost nine years until the Xbox platform was released in 2001.

"It's a long, long process," says Seropian, "and things have changed since then. Teams are bigger, projects are more expensive, and anyone who, today, wonders 'Hey, should I start up a game project from absolutely nothing and write every piece of code from scratch?' needs to know what a very big, daunting, multi-year investment it is."

Indeed, exactly a year ago, Wideload released Cyclomite, one of its "Shorts" (or casual) games using the Torque Engine which it licensed. But first Wideload prototyped the game by writing its own 2D sprite engine.

"Even that took us four months to build," notes Seropian. "But if we had wanted to write a 3D engine that was going to be competitive on a next-gen console, that would have taken many, many years. To start making an engine today that's going to compete with Halo 3 or Gears of War, that's a four-year investment."

"If you've got a large enough team that's really funded well or if you have a specific feature in mind that you're looking to exploit, then that's practical. For an independent developer like me, that's definitely not practical."

Gamecock/Wideload's Hail to the Chimp

And so, for Wideload's Hail To The Chimp -- a party game comprised of a collection of mini-games -- Seropian looked at five different engines, but turned to the Unreal Engine 3 for several reasons.

"First, we license engines in general because we are a small, 25-person company and we want to stay that way," explains Seropian. "Second, we were building Chimp for both the Xbox 360 and the PlayStation 3, and the Unreal Engine targets both those platforms."

"And the third and most important reason was that the Unreal 3 offered all the systems that we needed -- one for save games, one for networking, one for the interface we were building, and so on. Basically, for us, it was a soup-to-nuts solution."

Unfortunately, no solution is perfect, admits Seropian. "Unreal is a fantastic piece of technology, but when you use an engine that's based on a game -- in this case, Unreal  -- it's usually optimized for that game. Chimp is a very, very different kind of a game. Without getting into specifics, we had hoped to be able to use some of Unreal's online features without too much reworking to support network play. It didn't quite turn out that way."

On the positive side, Seropian calls the Unreal Engine's tool chain "phenomenal" and excellent for a company like Wideload that relies on outside contractors.

"It's so easy to find artists who are familiar with it and have shipped games using the Unreal Engine," he says. "When we previously worked with the Halo Engine, it was impossible to locate anybody who had used it before other than chumps like us. Let's just say it was less contractor-friendly, to say the least."

But Seropian turned elsewhere when building Texas Cheat 'Em -- a Live Arcade title -- mainly because the Unreal Engine was "too large and complex for what we are trying to do. Any change we would make in one place in the engine would have ramifications in a dozen places elsewhere."

"For us, the Gamebryo engine proved to be much more adaptable for the small game we were building. Sure, we had to write more of our own systems, but each of those systems was very simple to implement because there wasn't any large-scale baggage tied to it."

Next up at Wideload is an unnamed Wii game for which the developers will be using Terminal Reality's Infernal Engine -- mainly because it targets the Wii as well as the Xbox 360, PS3, and PC platforms.

"While some developers prefer to stick to one engine for all their projects," Seropian explains, "we're not afraid to use the right tool for the right job."

But, at GRIN AB in Stockholm, Sweden, there is only one right tool for the right job, according to studio co-founder and director Ulf Andersson, and that's the homegrown game engine Andersson has been using for GRIN's games since 1997.

"We've never even considered using somebody else's engine, never," he proclaims, "and that's because we have a very flexible platform that we can create anything with and not be one step behind because someone else decided what's going to be in the engine or not."

Capcom/GRIN's Bionic Commando

Despite the advantages, making your own engine is extremely costly and very risky, stresses Andersson. "Even if you've got the money, you need to be prepared to go through a lot of years of trying and testing," he says, "and even then the risk of failure is high."

Nevertheless, when Andersson and his brother -- now GRIN's CEO -- started the company in the basement of their family home, they did it with zero funding.

"Not only couldn't we afford to license someone else's engine," he recalls, "but we didn't think any of them would give us what we needed to make our game project work. So we chose to create an engine that was so general that we could create basically anything from it."

"We've altered it by 20-30% each year since, but it has enabled us to keep very stable and, even though we've grown from two guys to 270, our teams are always working with the same basic tools."

Those teams have utilized the GRIN's so-called Diesel Engine these past 12 years to create games as diverse as Tom Clancy's Ghost Recon Advanced Warfighter, Bandits: Phoenix Rising, and Wanted: Weapons of Fate -- and it is being used to build forthcoming titles like Terminator Salvation and Bionic Commando.

While Andersson is adamant about being mainly a one-engine studio, his team has dabbled with other engines when necessary; they did some physics maps for Unreal Tournament using the Unreal Engine.

"Regardless what it does -- good or bad -- it changes your way of working and how flexible you are in certain areas," he explains. "You have to adapt to the constraints of the limitations of what the engine has. You have to think in a different way and that's not the kind of restrictions we want here."

His best recommendation to other developers just starting out is to weigh your budget, decide whether you want to take the risk of building your own engine, and determine whether your staffing is capable of doing that.

"I get the feeling that most developers who want to build their own are cocky bastards who just want to prove themselves," he observes, "and that's a good attitude to have, a very healthy one. But, for most start-ups, it's far less risky to license somebody else's engine. Today's climate is very different from when we started in '97."

"Projects are 10 times larger than they used to be and so the risks when creating your own engine are far greater. At the same time, licensing fees are much less compared to budgets than they used to be, which may make it more economical to use someone else's engine."

But which one? Andersson recommends identifying the game you want to make, find an engine that has created something similar, and license it.  "On the other hand," he says, "if you're building a Gears of War-type shooter and don't want it to look like Gears of War, Epic Games' Unreal Engine may not be what you want. It depends on how high you're aiming. If you're going to try to beat Epic at their own game, you probably shouldn't use their engine."

Apparently Andersson's advice to license rather than to build is taken to heart by the majority of developers, according to Mark DeLoura, an independent games consultant. Of the 100 developers he recently surveyed, 55% are licensing someone else's engine although they aren't all happy about it.

"About 46% of those surveyed said that if they had their druthers and an infinite amount of time and money, they would prefer to create all their own tech," says DeLoura. "About 37% said they'd rather purchase a middleware library suite. And only 9% said they'd rather buy a game engine."

The primary motivations, according to the poll, are money, time, and whether the strategy is going to work for the game. Developers report they are paying approximately $1 million to license a game-specific engine but only $250,000 per title per platform for a generic engine.

"Let's say that building your own engine will take a team of 10 engineers approximately 18 months," says DeLoura, "so figuring $10,000 per man-month, that's about $1.8 million. But what you get for that money is an engine tailored to the game you're building which minimizes the customization you'd have to do on a licensed engine."

In fact, notes DeLoura, developers tell him that licensing an engine doesn't necessarily shorten the development cycle at all.

"What it does do is allow them to focus on the game-specific technology," he explains. "They don't have to write the low-level code that accesses the memory pack, the networking, or any of that. They can focus on how they're going to tune their NPC behaviors, game event timings, and so on."

One advantage of licensing a popular engine is that it simplifies outsourcing. "For example, 39% of the respondents who are licensing engines say they use the Unreal Engine," says DeLoura. "If you need to send work out to contractors -- or even if you need to staff up your team internally -- chances are you'll find people who have worked with that engine before. Not needing to have them familiarize themselves with your unique engine is a huge benefit."