NOTES // This article was originally published in Xbox World issue 77, back in March 2010. This is the original, unedited version before the Prod Editor cut it to fit six pages of magazine, complete with boxouts (which I’ve inserted within the body copy) but minus the concept artwork and assets Ubisoft kindly provided to illustrate the article. I’ve tweaked it a little, but otherwise you should read it as a product of the time. //
The rules were simple. Splinter Cell’s multiplayer mode is a two versus two asymmetrical game where each side has its own strengths and weaknesses, its own job to do, and its own way of doing it. Mercenaries defend a series of nodes with guns, grenades, remote mines and detection equipment; Spies attempt to hack the nodes using smoke, chaff, a taser and a lethal neck snap. Spies play in third person; Mercenaries play in first. Spies infiltrate; Mercs defend. Spies are lethal at close range; Mercs are lethal from afar. It was, and still is, the smartest, most tightly-designed multiplayer game ever made.
Now Assassin’s Creed’s Lead Designer at Ubisoft Annecy, Arnaud Mametz was the Design Lead on Splinter Cell Pandora Tomorrow’s multiplayer game. Splinter Cell’s multiplayer component was built at a studio far from the Shanghai-based solo game’s developers, and even used a different engine.
“After Splinter Cell’s debut at E3 in 2002, Serge Hascoet – Ubisoft’s Chief Creative Officer — wanted to see what a multiplayer game based on the Splinter Cell universe would look like,” says Mametz. “The task was given to Annecy because of some online prototypes we worked on, and because we were finishing Largo Winch: Empire Under Threat. A very small team was dedicated to it – just two people including myself – and one month later, we had a first rough game design, explaining the game pillars and foundations, and five minutes of gameplay explained with comic book panels. We received a go to work on what would become Pandora Tomorrow’s Versus mode.
“When we started, people told us ‘it’s not possible; you’re crazy!’” Mametz continues. “Well, they were quite right because we spent six months building the game before Xbox Live was even running, and that was crazy. When (Versus mode Creative Director) Gunther Galipot joined us in January 2003, eighty percent of the elements were designed. We had one acrobatic team of Spies versus one armed team of Mercs, we had the gadgets, the vision modes, the game modes and so on, but we still had a long way to go.”
The earliest Splinter Cell Versus prototypes were little more than testbeds for the engine and character movement. “First, we looked into Splinter Cell what can be reused” says Mametz. “Then we worked on the things we can give to the player that will help him to fulfill his mission. Then we needed to balance each gadget list so every gadget would have a counter. It was important for us to have a sexy gadget list but it should remain balanced. That was the main goal. Lots of things had to be removed because of balancing issues.
“Over the years we had so many ideas. But it’s a natural process; when working on a game you always throw away some ideas.” Ideas like:
“Different character classes, with each one having abilities and powers – it was just a balancing nightmare”
Story driven campaign games where you play several sessions on different maps, the results of one match changing the next. “For example — you play a campaign on three maps, if you win the game as a Spy in map one, the lights in map two are off and some gates are opened. We decided not to go for that solution because it seemed complicated…”
A dog as a Merc gadget, to follow the clues left by spies – “far too difficult to implement”.
Thermal Jammer – “a microwave grenade” — to override the Spies’ thermal vision.
Chemical View for Mercs, to see blood prints. “I’m sure you know why this one wasn’t kept…”
Adrenaline shots to revive players.
And a rocket launcher Merc gadget you’d have to be stationary to fire. “It was so cool but so unfair and not in the spirit of the game.”
“When we started to work on the project in 2002 we analysed the DNA of the Splinter Cell brand, which really wasn’t as obvious as it is now since the game wasn’t finished,” says Mametz. “Still, it was easy to see two main characteristics (of Sam Fisher): the acrobatic moves and the gadgets and vision modes. With that in mind, we first set out what should be each side’s ‘spirit’, what a player should feel while playing both sides.”
//Ubisoft Design Document//
AGILITY: I’m a cat. Even if I can’t strike from a distance, my speed and agility makes me deadly when I’m close.
FEAR: I’m very vulnerable in a frontal fight, I must be smart.
MASTERY: I master the map. I can hide everywhere; I can jump and strike at the best moment.
POWER: I have equipment allowing me to kill in any encounter.
INTELLIGENCE: I need to analyze clues to find my targets.
SUSPICION: Even if I’m heavily armed, I can be eliminated if I’m not focused on my mission.
“But before Gunther Galipot joined, we only had a prototype. The ‘spirit’ of the game was not there; something was not working,” says Mametz.
“At the time, both teams were using an over-the-shoulder third person camera, fixed behind the character with the ability to go first person when aiming. We knew the Mercs shouldn’t be allowed to see behind them otherwise the stealth was dead, but it was Gunther who proposed we use different camera views for the different teams. That was what we needed to make everything else work.”
The path the team walked led to game of tense, tactical battles between a pair of two-man teams. On near pitch-black maps, Spies could lurk in relative safety – hanging from pipes and beams, diving into air ducts and vents, and leaping and vaulting over obstacles and traps. Mercenaries were forced to keep their feet firmly planted, but could use a torch and laser to peer into the darkness, and use vision modes to detect fast movement and Spies using electronic gadgets. Most decisively, the Mercenaries can see from which direction noises originate on their HUD, allowing them to track a careless Spy by sound alone. Should a Spy drop from a ledge without softening their fall, or sprint down a corridor, markers would point the Mercenary in their direction.
“Once we found that spirit, we had two issues,” says Mametz. “First was the controls — we needed to build the Mercs’ controls from scratch, and fix the Spies’ controls too. Sam Fisher’s controls were not online compliant. Everything was designed for a stealth solo game: pace, animation rhythm, cameras, and so on.
“Another problem — the original Splinter Cell used a version of Unreal Engine with the online layer removed. For the multiplayer, we worked with the latest version of Unreal and had to add the specific Splinter Cell elements.
“And then there were design challenges. To be honest, creating the maps for Pandora Tomorrow was a nightmare. We spent months just working out what a good map for this asymmetrical game system looks like. We had no references and no ‘must haves’, but we knew we couldn’t use the single player ‘corridor’ maps. So we tried and tried again until we found a good recipe.”
The solution was playtesting. From the moment a working prototype was active, the team played the game every day, not just for balance and bug-checking but to teach themselves how to build an asymmetrical game and how to balance weapons, maps, and basic movements so neither side became too powerful in specific situations or levels.
“Oh yes, we played a lot,” explains Mametz. “We knew balancing issues would be very important. We spent a lot of time creating the maps and finding the golden rules for the game.
“What I really appreciated when working on SCPT is that every map has its own personality. I loved Warehouse for its progression through the areas and the Museum for its size and layout. In Chaos Theory I loved Aquarius and Club House. Each one required time to be developed, tested, and balanced. It was maybe harder for SCPT. It’s easy to say that a map is not fun but it’s difficult to know what to fix when you’re doing something completely new.
“We created some tools to help us understand player behaviours and how to balance the game. We were able to analyse each player’s trajectory, status – crouched, standing, hanging — and actions — using night vision, using smoke, and so on. We generated ‘heatmaps’ of activity so we could be efficient when understanding why some maps or part of the maps were working or not.”
Splinter Cell: Pandora Tomorrow was released on March 24, 2004, and immediately the team began playing against the community. “We had no idea of how the players would receive the game. We were totally amazed when we saw the number of players ranked after one week! One little pleasure was easily defeating players in the first few days, but after about two weeks we got our asses kicked every day by players becoming better than us.”
“The mandate was different for Chaos Theory. We had only nine months so we needed to be smart on our priorities. We looked at things we were not able to deliver in SCPT and things based on Chaos Theory’s new systems like the co-op moves, things players were asking for in the forum, and of course the biggest issues in the game – accessibility and balancing.”
Pandora Tomorrow’s sole serious problem was one which blights any complex, high-end multiplayer game – accessibility. Players who got in on day one walked over players who joined weeks or months later and Pandora Tomorrow had no matchmaking system to pair up evenly-skilled teams. The sequel, Chaos Theory, had to climb that mountainous difficulty curve.
“That’s why we added several tutorial maps and an ‘exam’ before you can go online,” says Mametz. “We tried to fix the accessibility problem in Chaos Theory but we made two huge mistakes.
“First, we made our explanations outside the game in a separate tutorial, and everything is explained in twenty minutes rather when you need it. That’s all wrong. We should have designed the early stages of the player’s experience so players could have discovered each rule in some kind of integrated tutorial. Then we could add different layers of difficulty and different abilities right when they are needed.”
Arnaud points to Call Of Duty 4 and Super Mario Galaxy as perfect examples of integrated tutorials and a gradual learning curve. In Call of Duty, players at level one can choose only one class and weapon set and are rewarded with new tools for progress. Galaxy, like all Nintendo games, introduces new ideas and concepts at a careful pace, never overwhelming the player with the dozens of moves Mario is capable of on level one.
Chaos Theory was far from a creative, critical or commercial failure. The solo game ran on the then brand new Unreal Engine 2 and was immediately one of the best and most beautiful games on the 360. The Versus mode, meanwhile, stuck with the original Unreal engine and even lacked the solo campaign’s widescreen option, but while Annecy’s game ran fast and smooth over Xbox Live, Ubisoft Montreal’s Unreal 2 co-op mode could barely sustain a stable connection for half an hour at a time, even after patching. Laggy play and dropped connections blighted cooperative players while the Versus mode trucked on as it always had, but with a steadily dwindling player base.
Dedicated players from Pandora Tomorrow made the leap to Chaos Theory, and though new maps were added and old maps were altered, the near-vertical learning curve remained thanks to thousands of players who had fought hundreds of battles in the previous game. With no matchmaking system, new players took their chances against elite players whose skills seemed almost magical. With absolute knowledge of the tools and the environment, Mercs could guess the Spies’ locations at all times, could “see” them in pitch darkness, and could locate them from rooms away. Pro Spies would seem to magically appear on top of nodes and were near untouchable for a rookie Merc.
That supernatural awareness built up by the best players made Chaos Theory a terrifying battleground for newbies, even with the tutorial, map guides,and the reluctant addition of a deathmatch mode.
“Deathmatch mode…” says Mametz, “Well, to be honest the designers here didn’t want to make a deathmatch mode. For us it just wasn’t in the spirit of the game at all. But some Pandora Tomorrow players created their own deathmatch rules in other sessions, destroying the experience of players playing the ‘real game’. We thought it would be better to add an official deathmatch mode so we can have keep players from annoying each other. We just gave the deathmatchers their own space to play in.”
Chaos Theory was an alienating game for the majority of players so for Ubisoft, the solution was clear. “Our number one mandate from the head office was the accessibility,” says Mametz. “We presented our first prototype to the Paris head office and it wasn’t accepted because it was very much a direct successor to Pandora Tomorrow and Chaos Theory. It was still very difficult to ‘enter’ the game if you are not initiated with both sides.
“So we had to rework everything. We came up with the idea of the heartbeat sensor” – a gadget which detects enemy Spies lurking nearby – “because it’s fun to hide and it’s even better when the one you are hiding from knows you are close. We tested the game with only that new addition and it was working quite well, so we built the game on that.
“Unfortunately, I don’t think we achieved everything we set out to do. The core game and the heartbeat sensor broke what was so good about the stealth in Pandora Tomorrow and Chaos Theory, and replaced the excitement and surprise with tension. We could have used more time to fix some exploits or to rewrite some of the rules to reach best of both worlds. Some parts of the controls and actions required were too complicated, some rules or exceptions were maybe too fuzzy.
“The maps in Double Agent too, were — in my opinion — not as good as the Pandora or Chaos Theory ones,” he says.
Double Agent killed Spy vs. Merc, for both gamers and Ubisoft. The great accessibility experiment didn’t lead to a rush of a new players and existing players stuck with the backwardly-compatible Chaos Theory or just abandoned the game altogether.
“We knew these changes would change the game and the way people were playing, and maybe the need to start everything from scratch was too big a change. I understand Double Agent was quite a shock for fans but I think the goal – to allow more players to enjoy the game — was a good one. I still believe it’s possible to provide an excellent Spy vs Merc mode that’s accessible for everyone, but that’s a huge challenge and something we didn’t achieve in Double Agent.
“I can’t really speak about the future of the Splinter Cell’s brand, but I really like the idea of Spy vs. Merc as a PSN or Live standalone game” says Mametz; “maybe it’s something the fans can ask Ubisoft for?
“I’ll always remember Pandora Tomorrow. It was such a huge challenge but for a designer it was a dream; we were able to create something that had never been done before and be innovative. It was a human adventure because everyone who worked on the game was so motivated. Everyone wanted to help; they spent nights testing just to find a single bug. Everyone was pushing his own limits just to provide the best game we could. I know the game wasn’t perfect but I think we can be proud of what was released: an innovative game. Was it unique? Was it something which could never happen again? Only time will tell.”
//Splinter Cell’s multiplayer Spy vs. Merc mode was my favourite multiplayer game ever. I played hundreds of hours of Spy vs. Merc on Pandora Tomorrow and Chaos Theory before I joined Xbox World, and this was something of a selfish project my editor was kind enough to let me spend some time on.
Spies Vs. Mercs returned for Splinter Cell Blacklist, developed by Ubisoft Montreal, not Ubisoft Annecy.//