When I was first told about the planned Invasion Mode, I was dubious. I started at Rebellion early in the development of Sniper Elite 5, the latest in our stealth sniping, third-person shooter series, and was made the design lead on Invasion because of my experience with multiplayer games. I’m old school, though, preferring to play single-player games alone and offline, and my instinctive reaction to other players joining was one of distrust. It only took a few weeks, until the first proper playtest, for me to realise I was completely wrong.

We immediately knew we had something: having another player in the game, with a sniper’s unparalleled ability to kill without being seen, immediately ramped up the tension and changed the experience dramatically. Those of us who normally played fast and loose were hugging cover, crawling along the bottom of ditches, and scanning every window. It was an electric feeling.

The intention of Invasion Mode was essentially to replace a traditional boss fight. We have AI snipers, but none would ever reach the level of creativity and intelligence of a player: you would never feel like you were being hunted in the same way. Films like Enemy At The Gates made sniper duels excruciatingly tense. We already had this in our multiplayer-only game modes (this Rock Paper Shotgun article was a crucial reference point) – now we needed to bring it to the Campaign.

The crucial problem we found with Invasion is that snipers are bullies. If a sniper is good at what they do, you’ll never see them coming and have no ability to anticipate or respond to them, making deaths feel deeply unfair – but if you take that away from them, then you lose the fun of playing as a sniper. How do we keep the host having enough fun to leave the Invasion Mode on, while keeping the Sniper Elite magic for both them and their enemies?

We started from the basic assumption that the Invader would be the more skilled player. The host would very likely be newer, and we needed to balance the scales accordingly. We needed to actively work to push players towards tense, cool kills instead of combat that was over in one shot.

The main way we did this was by focussing on the asymmetry of the mode. The host player is pinned down by the AI and the need to complete their own objectives. As such, they need to have access to more information. They already have Focus Mode to show the precise location of nearby enemies – we added the approximate direction of the enemy sniper to the ring around the edge. We also implemented Invasion Phones, which allow the host player to request the rough location of the Invader – with various limitations, including having their own position revealed if overused. These only give the host access to approximate information, a rough idea of which way they might be attacked from, and how to avoid the Invader. They get to act normally with one eye on a specific direction, rather than needing to watch every possible corner for an assassin.

As a counterpart to our “Hunted” player, the invader is the “Hunter.” They have access to less information and have to play more actively to acquire it. Skilled players will learn to establish the gameplay state from the objectives (if the bunker has been blown up, there’s no point waiting in ambush there), but we also allow the Invader to mark the AI soldiers, revealing the host’s location when a marked AI catches them. (This also gives them something proactive to do when lost – run around marking the NPCs and building their spy network.) The host can become extremely hard to find, and the Invader can move freely around the level, searching for clues and closing the net.

When the players encounter each other, the host has an advantage in a stand-up fight. Their Skills mean they’re likely to be tougher and faster than the Invader – only appropriate for hero Karl Fairburne – and they have more weapons and items available. The Invader can move freely and has assistance from the AI. They need to use that to attack from an unexpected angle or while their target is distracted – and with the information advantage, the host probably, sort of knows they’re coming. It forces the players into an imperfect, asymmetrical combat – one that’s drawn-out enough to allow the players to compete with each other as equals, maneuvering, planning, feinting, rather than dying immediately to the first shot.

The potential for a negative experience still exists, as it always does with multiplayer. Here I was able to use my initial scepticism for good: the Invader is a guest in your home; we’re going to make them wipe their feet. It was very important to us to prevent the Invader from spoiling your single-player experience. Invaders cannot take items, block objectives, or disrupt the AI too much. We make sure to save at the start of the invasion and give you a one-click route back to that save at the end, even if you win. This all allows the host to have a negative experience and move on from it without feeling that their playthrough has been irrevocably ruined.

One particular feature that supports this is the anti-camping system, by which players who stay in one place for too long are marked (and in the case of the Invader, eventually removed from the game). The nightmare scenario was for either or both players to retreat to a single room, shotgun pointed at the door, and wait for their enemy to approach. An invader with knowledge of the map could effortlessly camp an objective, and a nervous host might feel the need to turtle up for their own self-defence. In a way, the anti-camping protects both players, both by delegitimising this behaviour and giving out the information to break the stalemate… at the sad cost of the stay-in-one-place-forever part of the sniping experience. As before, the challenge is in making this perfect ghost archetype fun to play against.

There are a few things I wish we’d done differently. We had a whole planned set of behaviour that we referred to internally as “blending in.” In the very first playtests of the mode, it became clear that a fun strategy for the Invader was to hide in plain sight, pretending to be one of the AI, trusting to the Host’s need for secrecy preventing them from opening fire on every slightly suspicious enemy. We had grand plans in the works to build on this, including allowing the invader to choose and play idle animations, but this fell by the wayside. Firstly, the resources were always needed for more crucial parts of the game, and secondly, the list of small cues separating an invading player from an AI was becoming longer and more obvious, from weapon choices to walk cycles. Fixing all of these little differences would have taken still more time and resources, and they would undermine any more “blend in” features in an unfair way, giving the advantage squarely to players experienced enough to spot them and further disadvantaging new players who didn’t know the distinction.

It's very gratifying to see the positive response that Invasion has received. I’m hopeful that future projects we might do will include a similar mode – it seems that everyone on the development team has some interesting ideas about how to expand or improve it, and I think the potential is incredible. I’m very glad that I was wrong about how enjoyable it was, and hope I’ve managed to design a mode that’s enjoyable even to curmudgeons and cynics like me.