Cooperative games have problems.

This isn’t to say that you can’t make good cooperative games – Hanabi is one of my all-time favourites – but they come with a distinct set of design challenges that aren’t present in ‘normal’ competitive titles.  To get Sub Terra to correctly produce the horror film experience I was after, it was important that it involved the players working together for genuinely friendly reasons.  It was therefore critical that the underlying cooperative mechanics were as enjoyable as possible.

UNCOOPERATIVE PLAY

What problems am I talking about? Mainly, these two:

• Quarterbacking
  If all the players are working together, they will usually have access to exactly the same information (either because it’s already public, or because there’s no incentive to keep information private).  This means that working out the team’s next move can be done entirely by a single player, without the need to discuss options.
  
  If a session has one or two particularly skilled or experienced players, then they’re capable of playing the entire game by themselves, without the need for anyone else to be present.  If they’re polite, they’ll sit there, quietly bored, while their less experienced friends slowly come to the same conclusions.  If they’re impolite, they’ll take control and start making decisions for the group, allowing the other players to become disengaged and frustrated with the experience.  Ideally, games should engage all their participants as much as possible, regardless of individual skill level.
  

• Inconsistent challenge
  Games are at their most enjoyable when they present clear challenges to players that are neither too difficult (and frustrating) nor too easy (and boring) – this is to help create the desirable mental state known as Flow.  Our responsibility as designers is to keep as many players as we can within this narrow ‘flow channel’ for as long as possible.
  
  With symmetric competitive games, this is almost trivial to do.  You simply choose to play the game with players of a similar skill level to your own, and most gaming groups will be in this position already.  If this isn’t possible, or the game is not fully symmetric, then it’s often possible to give handicaps to the better players to even the playing field again – for example, a lot of games with first mover advantages try to offset this by letting the youngest player go first, using age as a proxy for skill.  Finally, building small random elements into these games allows low-skill players the chance of defeating high-skill players, which keeps both players engaged – this is useful when crafting compelling tournament games (contrast Magic and Chess).
  
  For cooperative games, however, you can’t use the players as a brake.  Your antagonist is the game itself.  If you make it too easy, then skilled players will steamroll it and find the experience dull.  If you make it too hard, then new or more casual players will find it impossible and never want to play it again.  And even if you pitch it just right, players will get better at your game over time. What was a sufficient challenge to a new player likely won’t be a sufficient challenge to a more experienced player.

How can we fix these issues?

QUARTERBACK TO THE FUTURE

The quarterbacking problem can be dealt with in a number of ways, including:

• Information Restrictions
  If each player has access to a different subset of information (e.g. a private hand of cards), then no one player is capable of making correct decisions for the group – everyone brings something different to the table!  Unfortunately, this effect quickly disappears if there are no restrictions on communicating that information to each other, as you can just show all the other players your hand.
  
  Some games such as Shadows over Camelot and Battlestar Galactica try to get around this by vague rules saying you can’t say what’s in your hand directly, but you can give “hints” – I’m not a fan of this approach, because players will push the concept of a ‘hint’ as far as possible to get any sort of advantage.
  
  Hanabi is a cleaner example, as here restricted communication is built into the game’s resource system; you have to spend ‘clue tokens’ to reveal information to another player, with strict rules on what you are and are not allowed to say.  The problem with this approach is that the restrictions work against discussions of group strategy; you can’t easily plan things as a team, as you have to be conscious of not accidentally leaking privileged information.  You can play in total silence, but social interaction is one of the best parts of tabletop gaming, and teamwork is one of the best parts of cooperative games.

• Asymmetric Goals
  If each player is working towards a different goal, what might be a correct decision from one player’s perspective might not be a correct decision from another player’s perspective.  At first glance, this may appear to be an odd solution in a post about cooperative games – surely cooperation requires everyone to be striving for the same ends?  Well, rather than there being exactly two states ( {competitive, cooperative} ), it’s more like there’s a spectrum:
  
  
  
  Even some competitive single-winner games have cooperative elements to them, as it’s a good way for 3+ player games to constrain the currently winning player.  The card trading mechanic in Catan is a great example of this.  Similarly, small competitive elements can be added to pure cooperative games.
  
  One way of implementing asymmetric coop would be to give the players an overarching group goal that’s required for each of them to “win”, and then an additional individual subgoal that gives them points which allow the winners to be ranked.  This has to be applied carefully, though, as it can lead to issues similar to the tragedy of the commons where players become so fixated on maximising their own score that the group goal is neglected and everyone loses.
  
  Another way is to have a hidden subset of the players be ‘traitors’ working against the group goal.  This is used to great effect in a number of games, but isn’t a great fit for every environment as the game mechanics need to allow frequent opportunities for sabotage of ambiguous origin, otherwise the traitor(s) are either obvious or useless.
  

• Strategic Complexity (“Depth”)
  Ideally, you can make your game easy to learn but very hard to master.  Less experienced and low-skill players will still be able to play and enjoy it, but they’ll make sub-optimal decisions using basic strategies.  High strategic complexity means that the more experienced and high-skill players will be able to make more optimal decisions, but the mathematically ‘correct’ decision will ideally still be beyond their grasp. They may suspect that they have a better decision than the other players, but they won’t know for sure.
  
  This ‘lenticular design’ approach allows a team of mixed-skill players to all feel like they’re participating in the game, and it ensures that the game is sufficiently hard to solve that even very skilled players will be unable to do so perfectly by themselves.  This is something you want to build into your game anyway, but it’s particularly useful here.
  
  In cooperative games, this effect can be bolstered by increasing the asymmetry between players.  If everyone’s character does a different thing or is in a different situation, then this naturally gives each of them a clearly defined chunk of the shared decision-making problem to work on.  This is the approach role-playing games like Dungeons & Dragons take – each character is usually too complex for more than one player to fully reason with at any one time, so the overall group decision-making computation is distributed among the whole team.
  
  (Appropriately, depth is a deep topic – I’ll discuss this in more detail in my next post)
  

• Time-limited Decision Making
  This is related to depth, but is approaching it from the other end.  In untimed games, players can (theoretically) take as long as they want to make a decision; rationally, they should take as long as possible until they’re sure their chosen move is optimal. In practice, however, players are constrained by social factors such as the other players getting bored, or the fact that it’s now 3 a.m. and they have to go to work tomorrow (today?).
  
  If the players are under enough time pressure (either socially or by using an actual timer), then the focus shifts from making the ‘best decision’ towards making the ‘probable best decision in the time you have available’.  There’s not enough time for one player to solve the problem completely, so it’s better to distribute the computation among all the participants.  Everyone gets to participate!

UNIVERSALLY CHALLENGED

Inconsistent challenge can be avoided (and flow maximised) using:

• High Scores
  This is the challenge behind ‘endless runner’ mobile games or classic arcade games like Tetris.  The game starts out in an easy state, but gets progressively harder as time or turns advance.  Eventually, you’ll reach a challenge you can’t overcome, and the game ends.
  
  Here, you’re not competing against the game, but against previous versions of yourself.  In a cooperative game, you can strive as a team to get the highest score you can in order to put all previous and future teams to shame.  You make the game as hard as you want it to be.  Unfortunately, this doesn’t work with games with a clear ‘win condition’ and end state, as this puts a natural cap on the score.

• Difficulty Levels
  This is the more common solution.  Here, you provide a selection of difficulty modes that players can choose at the start of the game.  Less experienced players can choose an easier experience, while hardened veterans can opt for a serious challenge.  This allows the game to cater for a wide variety of audiences and handles player improvement over time.
  
  Unfortunately, it requires players to somehow know in advance which difficulty level is right for them, or be patient enough that they’ll try again if their first guess is wildly incorrect.  Video games have a massive edge here, as they can attempt to detect player skill as they play and tailor the experience to match.  The Left 4 Dead series is a great example of this adaptive difficulty in action; players basically choose a ‘comfort level’, rather than a difficulty, and the game spawns enemies and obstacles to maintain this.  It’s harder for board games to emulate this, as the players will be necessarily aware of the difficulty-shifting conditions and can use this knowledge to game the system.

Not all of these solutions will work for every game.  A lot of them change the texture of the gameplay, or don’t mesh well with the thematic goals.  But they were a good starting point when working on Sub Terra…

AS ABOVE, SO BELOW

Sub Terra deals with the quarterbacking problem primarily by maximising strategic complexity, with a little bit of goal asymmetry thrown in.

One of the key parts of reinforcing the horror theme was cementing the link between player and their in-game character.  To help with this, the endgame scoring made the most sense as (roughly):

“If you escape, you get a ‘medal’ corresponding to the number of other players that escaped – ‘gold’ if everyone escaped, ‘silver’ if one player didn’t make it, and ‘bronze’ if two players didn’t make it.  If you don’t escape, you get nothing”.

This scoring function is still strongly cooperative, as you’re incentivised to help get as many other players out as possible.  But it also adds a personal element to your decision making – do you risk your own life to rescue a friend?  Do you sacrifice yourself for the good of the group?  It’s also interesting because the reward is not numerical, which allows players room to value ‘medals’ differently – what a ‘silver medal’ is worth to one player might be far less than what it’s worth to another player, which can produce legitimate strategic disagreements which require negotiation to resolve.

(I’ll cover how the strategic complexity was handled in my next post – I’ve found that even experienced teams have a lot to discuss, as the game is not easy to solve perfectly while still remaining accessible for inexperienced players)

The inconsistent challenge problem was dealt with using difficulty levels, as the theme very clearly wanted ‘success’ and ‘failure’ states rather than a final score.  The hazard deck gives the game a natural clock – one card is resolved per turn, and the (short) endgame starts once the deck is empty.  The longer this clock is, the more forgiving the experience and the more mistakes you can make as a team before escape is impossible.  Modifying the difficulty by adjusting the size and composition of this deck was an obvious solution, and one that lasted all the way through development.

As an aside: originally, the names of the levels were “Easy”, “Medium” and “Hard”.  It eventually became clear that these were inappropriate, as those words describe the level of challenge you’d expect to face, and no-one wants to play a game that’s too easy or too hard.  Taking a cue from Left 4 Dead, these were replaced with “Beginner”, “Normal”, “Advanced” and “Expert”, which didn’t make any implicit promises about the level of challenge and also guided players to levels based on experience.  Names matter!

COOPERATIONAL RESEARCH

Cooperative games may have problems, but hopefully the above shows that with some careful thought these problems can be solved without compromising gameplay.  Did I miss anything?  Let me know in the comments!

In my next post I’ll be tackling the related but distinct concepts of complexity and depth, and how I tried to make Sub Terra both accessible to new players and remain rewarding for experienced ones.  See you then!