Splinter Cell Chaos Theory

There is usually a moment on a press trip, typically just after you first enter your hotel room after a transatlantic flight, when you wonder whether the game you've come to see is worth the countless miles travelled. This time, however, there's no such instant of contemplation because the answer is to be found lying on the room's desk. Ubisoft has kindly arranged for a Splinter Cell 'goodie bag' to be delivered but, unexpectedly, it's the hotel directory next to it that has boosted our jetlagged confidence regarding the quality of Sam Fisher's latest mission. In it, the hotel's general manager has personally signed the introductory note - twice, in fact, given that the guide is in both French and English. And that's for each of the hotel's 258 rooms.

This level of attention to detail extends beyond Montreal's hotels and, more relevantly, has been particularly characteristic of recent projects to emerge from Ubisoft's Canadian operation. Established in 1997, Ubisoft Montreal now houses 900 employees in 150,000 sq ft arranged over three vast labyrinthine floors of an old textiles factory. It is, it proudly boasts, the world's second largest development studio. It has its own gym, several recreational areas, and each of its kitchen units comprises nine microwave ovens and three US-sized fridges to cater for its ostensibly committed staff.

One group of employees that has certainly been making regular use of the culinary facilities since the autumn of 2002 is the original Splinter Cell team. (If you're wondering, Pandora Tomorrow was co-developed in Shanghai and Montreal, with the Canadian side of the equation providing a dedicated team to only look after that game's multiplayer aspect - a structure it has again employed. Currently with a headcount of around 150, it's Ubisoft's largest and a convincing indication of the publisher's commitment to the third instalment in its popular stealth franchise. For Chaos Theory, the company is addressing issues both press and public had regarding the previous two iterations, while attempting to push the concept of stealth further and evolving the Splinter Cell universe by enhancing the player experience.

"One of our ambitions was to make this game a little more massmarket, to get a larger audience by fixing all the frustrations we had in the [first] game," says producer Mathieu Ferland, "but also to support the hardcore gamer with the non-linear structure, with all the secondary objectives and opportunity objectives and bonus objectives and all the little details they'll discover when playing the game."

They may sound unconnected, but these elements are related to one another, working together to remove the linear nature so criticised of Splinter Cell (which, it emerges during our trip, originally included non-linear aspects that were removed for reasons unknown).

Clint Hocking, creative director, scriptwriter and lead level designer, explains Chaos Theory's approach: "We really wanted to have what I call controlled non-linearity or semi-open topology. The world is open and you can go in any direction at any time but what we have is 'the 80 per cent path'. This concept means that the level designer builds the map in accordance with the script and with the design objectives of the game, and he plans out these ten or 12 areas and how they're topologically interconnected. He plans the path he wants the player to take, and this is the path he focuses most on. If there are other paths available, that's fine, but we need to make sure that the path that the designer designs and spends the lion's share of his effort on is the path that 80 per cent of the players take. The non-linear paths are still there, but they are really intended more for the hardcore players, for the repeat players to add replay value."

One particular example, a Tokyo bath house brimming with guards, allows two main routes through (with the occasional sub-path thrown in) and while both encourage Fisher's trademark considered stealth infiltration method, it's promising to be able to test the developer's claims that the game can now also be played in more conspicuous fashion. "Just because we're encouraging and rewarding the player to play stealthily doesn't mean the game doesn't support action," Ferland offers. "Sam now carries a shotgun for those specific sequences where there's lots of firefights, and we've been developing the AI so the fight sequences are much more attractive and relevant."

More on the AI in a moment. Levels are now more structurally complex and offer a multiple-path system. Players can make decisions about game flow, but by implementing a series of objectives in addition to the multi-path approach the developer can, to a certain extent, still control the player's actions. While primary and secondary objectives are obviously mission-specific, opportunity objectives take the form of, say, planting bugs on six telephones - they're not heavily story focused but serve as a balancing and guidance tool for the developer to invite gamers into most of the map's areas. Similarly, bonus objectives reward the meticulous player who is prepared to explore the environment in minute detail. One interesting addition is the inclusion of 'fallback' objectives, which only come into effect once the mission has gone wrong and serve as a way of reducing the number of what previously would have resulted in game-over situations.

Your opponents, on the other hand, will be looking to increase such circumstances. The AI has been completely rewritten and now uses navigation meshes rather than navigation points of SC1. In addition, NPCs benefit from increased environmental awareness including mirror and shadow detection, contextual search animations (such as looking over a railing) and memory, rather than going back to an 'idle' mode after a set time period. They can perform complex tasks such as teaming up with a fellow NPC to perform searches (and notice if the other goes missing), flanking, and converging around a specific object to communicate with one another.

Furthermore, the developer has included the notion of stress levels, which depend on the number, and nature, of stimuli. "We're thinking that it's very important that the NPC cares about their life," expands lead programmer Dany Lepage. "Of course, if they care about their life they're going to be behind cover very often because they're not going to stay in the open and that's going to [make it] difficult for you to kill them." However, Lepage is keen to point out that the team has ways of ensuring the AI isn't so clever as to make the game frustratingly difficult, such as limiting their accuracy and cone of vision: "I think they're going to be much more interesting opponents. You're going to have to think a lot more to get the same result because the behaviour is going to be different depending on what you do - they're going to have a strategy that will adapt quickly to what you're doing and you'll need to use something different to get rid of them." To prove the point, Lepage demonstrates a firefight between Fisher and a soldier who appears impressively sharp in his use of surroundings for cover. In practice, we find the enemy's perception level to be a little overenthusiastic, with the slightest look around a corner resulting in immediate detection, but the team readily admits there is still fine-tuning to be done.

Which seems like an appropriate time to leave them to it. In terms of atmosphere (the combined result of standard-setting animation, graphics, audio and AI, plus a novel coop mode), Splinter Cell Chaos Theory is a determined leap forward for the franchise and, potentially, the stealth genre as a whole. It's not quite there yet, but come November you'd expect the detail-obsessed Ubisoft Montreal to deliver something special.

Splinter Cell Chaos Theory is out on Xbox and PC in November