Keep it Together: Encouraging Cooperative Behavior During Co-op Play

[Music] okay I'm gonna get started thanks for coming down again this is keep it together where we're gonna be talking about how to get players to cooperate with each other when they're playing in co-op games so I'm Patrick Redding I'm currently a game director at the new Ubisoft Toronto studio we opened a little over a year ago prior to that I spent about five and a half years six years at the Ubisoft Montreal studio where most recently I was the coop game director on Splinter Cell Conviction and then before that I was the narrative designer on Far Cry 2 and then played various game design roles on the last generation of consoles far cry games but today I'm gonna talk about player cooperation and I mean this in the sort of most obvious explicit possible way

I'm I'm not talking about cooperation in the sense of when players get together and agree to abide by the same rules I'm not talking about stepping into the magic circle or any of that stuff I mean actually players working together to try to overcome the challenges that are posed by by the computer so so co-op play before I get into that I just want to step back a couple weeks ago to an article that I was reading on greatest Egan's blog which has played this thing and there was an article and posted by one of his regular blog contributors goes by jz w and jzw was asking he was sort of thinking about the current situation in the Middle East and some of the recent popular uprisings that have taken place in places like Egypt and Algeria and whatnot and he was asking if someone was to make a game that was about a modern nonviolent revolution what would that game look like and this

you know this prompted kind of a lively discussion on the comments thread in which a variety of people brought up different games that they had seen over the years they thought kind of fit the bill games this is a force more powerful from a few years ago and pretty quickly the author had to kind of push back and say well hang on a lot of the games that people are talking about our games in which the player is kind of a Revolution tycoon he's somebody who's sort of taking all of the resources that make up the populace and figuring out how to allocate them in support of a revolutionary movement in order to take down the man as it were and his point was that in real life as we've seen somewhat vividly somewhat tragically in some cases real revolutions are started from within they're started by people whose necks are every bit as much on the line as the people who they're organizing and that furthermore the nature of this kind of movement is that it started by people reaching out to

their friends and to their family and to members of their immediate community the people with whom they share all of the common interests that lead them to have a grievance in the first place that it's not about just simply sticking a youtube video on the internet and then having 20,000 people show up in a public square the next day and there has been kind of a popular misconception I think in the last couple years that somehow modern nonviolent revolutionary movements come about as a sort of smart mob with slightly higher stakes that the fabric of the Twitterverse has somehow just kind of given rise to these movements in some kind of magical emergent way and in fact this doesn't turn out to be true at all it turns out that the same kinds of strong social links that were needed 5,000 years ago or 500 years ago or 50 years ago are require are just as critical and important now as ever that

it's not gonna be your 958 Facebook friends that are gonna help you when you're up against the tanks and the barricades and the reason why this is ought to be relatively apparent but there's some non-trivial consequences to it basically if you're dangling at the end of the string if it's an existential stake that you're up against then those strong social links that you're depending on are a mechanism for negotiation you're depending on the fact that you have license to negotiate action in the service of some kind of collective goal and what's interesting is that the negotiation actually helps to reinforce the strong social links and then the strong social links in turn help to reinforce and enable the negotiated action when we apply this to other types of collective goals say the collective goals associated with an alternate reality game or even an online

multiplayer game something like World of Warcraft what we find is that the same effect is true obviously slightly smaller stakes but players become invested primarily in the success of their collaborative partners because of that ongoing process of negotiation because of those leveraging of strong social links the negotiation process converts their individual selfish motives into the motives that are associated with the collective goal now it's probably too dark but that's okay like a lot of people within about five square blocks of this auditorium I've occasionally been playing minecraft in the last year mostly I've been playing minecraft on a multiplayer server maintained by a small group of friends of mine who are scattered all over the world in different time zones and what's interesting is we hardly ever actually see each other when we go on this multiplayer server we show up we do our mining we do our crafting we construct massive monuments to our personal hubris

but we don't actually see each other's avatar very often and we don't even really chat that often so interestingly a couple weeks ago probably around the same time I was reading that article on play this thing I had to sort of take a bit of a hiatus from playing Minecraft to be honest it wasn't a hard hiatus to take as much as I enjoyed mining and crafting and building extravagant monuments to my hubris the truth was that you know I had a point where I had other things that I needed to take care of like creating GDC talks but one weekend I decided to fire up the server because I was kind of to break and I was curious to see what had been happening in the two weeks since I've been on there what you see here in this kind of dark images is kind of an elaborate urbanized area with elevated light rail systems and waterworks and lava pipes and various barricades for the mutual defense of the inhabitants against marauding mobs none

of that had been there two weeks before it had been desert and waterfront and mountains it was like visiting people in limbo and inception all of a sudden there was like these giant cityscapes that had emerged from nowhere and I was almost instantly overwhelmed by the need to devote the next 48 hours to working on my territory it was I felt this powerful sense that I was not only missing out on all of the cool collective things that had been going on in the meantime but that maybe I was actually not holding up my end of the bargain and I think the point here is that players respond very very powerfully to a sense of collective agency that's at work in the game space it's more powerful and more potent than maybe any one individual game motivation that they might experience as a single player and for scientists who study the psychology of people in games in games that have multiple players this is known

as shared intentionality and it lies at the heart of what we're gonna be talking about in terms of cooperation so that's fine that's minecraft that's eight weirdos playing a game that doesn't actually have victory conditions a game that hasn't technically been released yet for those of us that are working on kind of these mass-market commercial mainstream games that have a co-op component how do we achieve the conditions in which shared intentionality becomes an important factor in the enjoyment of the players and therefore causes them to want to work together and to want to cooperate now this isn't trivial because by definition and if it's an online game the players are separated geographically if to a large extent our statistics are correct and we think they probably are probably at least half of them don't even know each other they're playing by a matchmaking and various other mechanisms also let's face it as

game developers as designers we really prize and prioritize individual player agency right we have huge fights about it's probably the one fight that every single designer has had at some point with the rest of their team as we as we argue that regardless of what the production realities are regardless of how polished the scripted events might look in a particular section that if we're eroding somehow the agency of the individual player if we're taking away from interactivity that were fundamentally degrading the play experience and we're actually correct about that however sometimes I think we develop myopia because we stop don't stop to consider that also multiple players need to be able to have a kind of collective agency and that we need to support that systemically so I'll talk about some lessons that I learned with my team while I was working on the coop part of Splinter Cell Conviction well the first big lesson is that players will work together if they

determine that there's a way for them to use that cooperation to optimize the output of the game systems to talk about that I'm going to just refer very very very very loosely and very broadly to the detection model that we used in Splinter Cell this isn't going to be a technical discussion of this by the way so if you're in the AI business I apologize it's not very deep so the number one lesson that we obtain from observing our players is that stealth is a difficult enough thing to achieve in a single player game but when you have more than one player running around in the game space stealth becomes an extremely fragile condition okay well what what do I mean by stealth what am I talking about well stealth for us is really a condition that we achieve by allowing the player to either be detected by the AI or not be detected by the AI and it's all the various actions the player can take one way or the other in that direction and to understand how that works we just need to sort of see that

our AI has been kind of built to be tricked but also been built to incrementally detect the player so there's a number of states that permit this we have kind of an idol default state which is everything that the AI does between the time that it's instantiated and the time that it's killed or unloaded whether it's systemic or scripted this is going about its business but if something happens to disturb the AI wall it's in that state then it can move into this kind of perturbed condition where the behaviors are all about searching for whatever the source of the disturbance was and if enough disturbances occur or if the player just simply outright allows themselves to be detected then the AI will enter a state where its behaviors are those that we would associate with combat with trying to close with the player trying to kill the player or more specifically and accurately trying to entertain the player while looking like it's trying to kill them and an important point here again not technical at all is that these

transitions that I just described between states these are thresholds basically this represents an accumulation of events that cause the AI to move from one state to another and if we look at just even a single player that's operating in the game space what we see is that it's actually kind of difficult to know exactly how any one little action or event contributes to those transitions taking place I mean you can do it you can search you could build a giant list or a log you could build a big matrix that shows exactly how one particular action by the player under certain conditions will cause us to take place there just be a lot of work but if we add a second player to the mix well we have to do all that as well we have to see how the second player can contribute actions that will cause the second player to be detected but then very quickly what we realize is that actually it's even worse than that because then we have to account for how player one can cause actions that will cause player two to be detected and then player two can do things that will cause player one to get to two so you can see I mean it turns into a nonlinear and

i'li messy combinatoric expression that I wouldn't want to have to draw the matrix for it bottom line just because you have two players in the map doesn't mean you only have twice as much chance of being detected in fact it's possibly a whole lot worse than that but the flipside is that if the players are functional behaving intentionally and they are observing the AI and learning how the AI operates and learning how the detection model works they can actually work together to reduce the chances that they're going to be detected substantially and the point is here that players are acting with their knowledge and understanding of the game systems not because the game has told them that that's what they should be doing but because as they study the game and as they play the game whether they're playing cooperatively or not to begin with they realize that they have a better chance of success if they work

together under these conditions and what's interesting maybe even paradoxical is that that power that comes from that knowledge and understanding of the game systems that kind of procedural literacy actually is giving the individual players more permission to express themselves as individual players and this is something that we actually observed when we were doing our play tests and when we were looking at other games as well is that self-expression becomes a much more interesting and deep and rich thing when there's multiple players around at the high level what we saw was that players working together will do so to develop strategies in the pursuit of the games objectives that at a mid level those players will work together to create planning and to delegate individual low level tasks in order to support those plans and that when they carry out those level level actions as individual players they're more likely to take

risks and to try things that they wouldn't have tried if they were playing as single player players and that over the course of the entire game as they are developing this kind of mastery this kind of literacy in the game systems their understanding of how the AI works how the 3d navigation system works all of these things that they are willing more willing than they are in single player to explore optional paths and try alternate play styles because they know that they have a reasonable chance of regrouping with their team and trying another approach if it doesn't work out and what we find is that if both of these conditions are present if the players have an understanding that they can use their cooperation to optimize the output of the game systems and that the game is letting them express themselves as individual players more because of cooperation then they actually enjoy the game more in fact their enjoyment from that from this from what I'll call meaningful cooperation is so good and so high that they're willing

to forgive an awful lot of flaws and an awful lot of lack of polish in the game that would maybe bother them if they were to scrutinize it while playing as solo players ok so I use this term meaningful cooperation I want to examine meaningful because it's a little bit loaded but basically what I mean by about by this is that it's important in terms of their actual success in the game it's not just something we're telling them you ought to do this because it says co-op on the on the box but also even though we're not telling them to jump through hoops it does matter in the context of the game systems the game systems recognize that they're doing something in this case cooperating and it rewards them for it ok so now we have this idea that we would like to try to create circumstances for conditions for for meaningful cooperation now we need design tools as designers that are going to allow us to support that we need specific strategies what now today I want to talk about strategies that are

hopefully usable by the largest number of people so I want to talk about strategies that don't depend on any one particular aesthetic wrapper or particular genre I want to talk about strategies that are hopefully scalable and are replicable across the whole game that are not dependent on a particular special case implementation that are practical in the context of most productions so I'm gonna talk today about what I'll call cooperative dynamics which essentially just means I'm talking about enabling cooperation through the games dynamics ok another loaded term dynamics actually it's good that I'm talking on a Thursday because those of you who've been here for the week will likely have now had several opportunities to run into this usage of the term notably acquaints talk for me today and if you've attended the game design workshop you've heard those guys talk about it basically its dynamics in the sense that we meet it in the MDA model and in mark mark and Robin and Bob's paper in 2004 they said described

it as the runtime behavior of the mechanics acting on the players inputs and each other's outputs over time which is a way of saying it's what the game does when you're actually playing it dynamics are what is delivering the final player experience in this case it's cooperation meaningful hopefully but cooperation meaningful or otherwise and what I'm going to talk about today are seven of the dynamics that we were looking at on Splinter Cell this list I'm not going to pretend is complete but what I am gonna try to do is is demonstrate or claim that all seven of these dynamics which I'm going to go through sit on a continuum of what I would call coerciveness that they run the gamut from the top to the bottom of dynamics that are highly prescriptive in the sense that the player can't avoid them down to those that are entirely voluntary the player retains control

over whether to participate in them or not an interesting gameplay emerges out of that okay so let's look at the first of these what I'll call gating or tethering this is probably the one with which the majority of players and the majority people here are familiar it's kind of the most brute force essentially is just saying that no player is allowed to proceed in the game unless all players are allowed to proceed in the game and this is should be familiar to anybody who's ever run into an invisible wall because there was a teammate that was straggling behind them or has been abruptly teleported forward because a teammate hit a checkpoint in advance of them it should be pretty evident that this is a highly prescriptive approach towards keeping players together it's either straggler centric or it's sprinter centric in the sense of relying on these outliers to determine whether it works and most importantly maybe is that there

are lots of ways to tart this up and make it look like something more and deeper than it actually is but even if it is glossed over with a lot of production value it is just going to be an invisible wall or an invisible tether mechanically now that doesn't mean there aren't good implementations of it in uncharted 2 if you're playing gold rush mode and in their co-op you'll encounter situations where the two teammates need to work together for example to move you know a burning obstacle out of the way so that they're able to advance forward it's a good implementation it fits the fiction of the game the kind of action-adventure sort of fiction that's been used there and it's consistent with the level of production value that you find elsewhere in the game on Splinter Cell we used I have to admit a lot of different gating strategies some of which were I'd say heavily realized and

some of which were just straight-up gamey and visible collisions we had things like exotic doors that needed both players to pry open but we also had our holding points between different sections of the map where one player would just simply be forced into an idle position until the second show okay the next I'm gonna talk about is what I call exotic challenges and the term exotic in this case means specifically that the normal camera and controls the players using during core gameplay have been altered and substituted with something else the unique perspective and an unique set of controls that are for different different types of gameplay and what's interesting about this is that it creates the condition in which the player who is required to use these different controls is still susceptible to all the same kinds of threats and all the same kinds of defeats that are that he'd be confronted by when he has his core gameplay controls so it's moderately coercive in

the sense that it's not a binary threat it's not something it's immediately gonna snap and then it's kicked into effect and now you're unable to proceed what it means is that the risks associated with the player playing by themself grow sharply over time and that furthermore you need more than one player to avert trouble in order to survive because there are other players who have retained the core gameplay controls and are therefore able to support you so classic example of this is the sort of wizard performing astral projection while the muscley fighter guy you know hangs out and keeps track of his lifeless body the modern warfare equivalent to that is I equip my laptop in order to control the predator drone while my friends continue to use their tactical movement and and shooter mechanics in order to protect me what's interesting is actually in case of modern warfare 2 and the spec ops levels that they have is the flip sides also true right I could be in an ac-130

gunship and I'm actually the one helping to avert trouble for my friend on the ground who's using the core gameplay mechanics he's outnumbered he's fleeing a larger force and I'm firing shells in order to help protect him and cover him and play it run interference a conviction we had a few different exotic gameplay moments probably the best realized was the wireless hacking sequences where one teammate would need to go up to a particular computer interact with it and then had to remain within a certain radius of that object in order to be able to maintain the signal strength to download the secret data which meant that they were constrained they still had access to a lot of their core gameplay but they couldn't move away from that point or it would interrupt the download teammate who retained free movement would have to run interference move around try to protect the two of them against the inevitable horde of security forces that were responding to the hack so the next two systems I'm going to talk about are

kind of two sides of the same coin in the case of punitive systems what we're saying is that really one player is trapped one player is trapped and another player needs to come to their rescue or it's game over for them I think the obvious thing about this is just that it requires rescue by another player in order for the first player to survive the less obvious thing is that it actually represent and punishment and negative feedback to everyone involved not just the player who's been trapped player who's been trapped is being punished because he was careless and wasn't paying attention and wandered away from the group all the usual things you would expect the other players are being punished because they let the player wander away but also they're punished if they don't respond in time and don't take the necessary action to to recover the situation but the less obvious thing is that it's not a particularly coercive mechanic because it can be avoided if you're smart and if you've mastered the game you're able to recognize that the dangers are there and

you're able to to get away from them probably the best implementation of this that has we've seen in recent years would be in the left 4 dead series with a smoker I think the smoker is perfectly balanced in the sense the the signs of feedback associated with it the distinct sound that accompanies the smoker appearing on the scene sort of sets the entire party on edge they quickly look at each other to make sure they know where each other are they quickly scan the rooftops to make sure that they can see where the smoker is and and maybe kind of disgustingly the smokers tongue is like the perfect progress bar right it's like it's like the ultimate organic timer as it drags you to your doom everybody you can see exactly how much time they have left in order to try to save you we had a couple of punitive systems in Splinter Cell Conviction co-op we had the the usual one that you associated with these kinds of games where the one teammate has been wounded is gonna bleed to death unless the other teammate comes and rescues them but we had an additional one which is at higher levels we started do encounter enemies that were capable of of grabbing an

unwary player and holding them in a chokehold at which point didn't start to asphyxiate and then the the second player we need to run to their aid and the first player could throw the elbow and duck down and create the window of opportunity what's interesting is that these two systems actually ended up working together and are really surprising and delightful way which I'm going to talk about a little bit later the flipside to those are would be buffing systems buffing systems are pretty familiar it's just the idea that one player is mechanically making another player more powerful and as they are doing that it's creating this kind of back and forth movement between the team being sort of unbalanced in a way and because the buffer is probably rendered vulnerable by this and the Buffy becomes very overpowered one thing that's interesting is it can be done implemented in a way that's either passive or intentional passive in a sensitive as long as you stay close to me you'll be tougher less visible or it could be intentional and that I actually have to target my teammate the benefits are almost always

conditional and usually temporary in order to avoid an permanent unbalance and it's entirely voluntary players can opt in or not opt into this an interesting approach to this on a game that's not only pretty realistic would be in the Army of two series the aggro system which literally allows one player to buff the stealthiness and the and the invisibility of his teammate just by standing up and making a bigger target of themselves and this is a system that's actually grafted above and beyond the normal behavior of the AI accompanied by a bunch of signs and feedback in the hood and in the game now we had a buffing system in Splinter Cell Conviction co-op and people and people definitely used it however if we're being honest with ourselves we probably have to admit that a lot of them didn't realize that they were using it we had the mark and execute system which was developed for single-player which have enabled the player to pre mark targets and then under certain conditions take

them all out in one attack to sort of enforce that idea that you're the elite spy who's a crack shot and you're able to do this and co-op what we allowed players to do was jump in on each other's execute if they shared line-of-sight on any targets but also if any other marks were available nearby so that we could essentially extend and boost the the range of the the area of effect of the mark and execute and it was a great system and people definitely were using it but when quizzed about it afterwards a lot of them weren't even aware that they were using it so fail asymmetric abilities are I think a pretty straight forward notion to people who've played for example you know rock band style games games that are supporting different sets of abilities for the characters and we're talking specifically about the characters in this case the the idea being that the the group is constantly moving back and forth between conditions of being sort of overwhelmed by the the challenges of the game or overpowered just as they sort of reconfigure

themselves to tackle those challenges in the right way it could be an entirely built out orthogonal class system like you might see in an RPG game or it could be the consequences of a customization system players spending experience points in order to specialize in order to give themselves unique abilities but the one common assumption here is that no one player is going to be able to max out everything right they can't just turn themselves to do a super generalist but what's interesting is that players are still free to progress with their own limited subset of those character abilities they can try it and in some cases they can succeed at it it's just that all of those complementary advantages and disadvantages tend to lend themselves in a fairly obvious way to the players to working together for mutual support Borderlands is a game recently that I think has done a very good job of this because at the start of the game while we have different stats the different characters have different stats nevertheless they do share a kind of core set of mechanics that they can

kind of learn to use at the same time and after about an hour of leveling up they are able to unlock their unique abilities and then become very powerfully motivated to explore those unique abilities even while they've retained their understanding of the core systems the fact of the matter is that on Splinter Cell Conviction we didn't explore this particular dynamic very much the both of the players in spite of the fact that one of them supposed to be an American Splinter Cell the other one supposed to be a Russian boron agent have identical abilities identical movesets they use the same animation the same fighting ability they just look different and sound different when they talk so the onus was entirely on the player to use the pec challenge points in order to customize their weapons loadouts to give themselves either arm or more weapon ammo capacity or a gadget capacity and the sad reality is that most players didn't explore that system very much at all so next is combined actions now combined actions on first blush is going to sound a little bit

like asymmetric abilities it's the notion that there different player skills that are implicated in the gameplay and that players need to understand that no one player is going to be able to tackle those challenges by themselves or at least it's going to be very very difficult for them to do it and the key here is just that players are able to recognize the nature of the challenge break it down into its individual component challenges and then divvy up or delegate those roles between them at which point it becomes almost a formality now this is this hinges on an idea it's pretty simple but it doesn't it does merit some explanation so the notion that any game challenge that we describe can be defined by the specific player skills not the character abilities but the player skills that are required in order to overcome it now here's a list of player skills precision timing measurement and the sort of analog input

sense resource management tactical choices strategic choices puzzle play this might be a complete list but I'm not gonna stand up here in debate whether it should be like eight more things or five fewer things or whatever I mean the point is that if we accept the notion that there are skills and that the skills are implicated by discrete gameplay challenges then it's possible for us as designers to ensure that the tempo of those challenges is just always beyond the comfort zone of any one player at which point it's up to the players to identify what those compound challenges are agree amongst themselves who's gonna take on what and then at that point it becomes a rather enjoyable process of overcoming it Red Dead Redemption in the out loss to the end co-op does a very good job of this actually the specific example I'll use of I'm driving the horse-drawn stagecoach so I'm having to engage in precision and timing in terms of controlling the horses a certain amount of analog measurement in terms of my speed at the same time that my buddy

who's riding shotgun is having to use precision to line up targets with the crosshairs is having to make tactical choices about which of the bandits to shoot in which order this is almost a classic example of how combined actions would work in order to help promote cooperation and Splinter Cell I mean we call herself still a stealth action game and in co-op it was a mandate on the level design side to try to support multiple styles of play players being able to use action being able to use stealth being able to use gadgets in order to make their way through it the reality is is that an awful lot of the skills associated with those different styles of play are the same skills they differ maybe by one or two skills in a couple different areas but we didn't really fully explore or take advantage of this approach but it's definitely something that we could the last dynamic I'm going to talk about is one that I think is probably familiar to people who are have played a lot of the bullet health style shooter games it's notion of survival and attrition and as a dynamic it really is a kind of

oscillation in terms of player motive so it it's almost at the level of meta game it's the notion that play can continue for as long as there is at least one player alive which is essentially saying you could be a hero but you can also be dead in a sense many players are possible or many survivors are possible that's all the players could survive till the very very end but if you die you're out for the round you're gonna be waiting for however long the round lasts until your friends are done and you get to listen to them laughing and screaming and yelling and enjoying themselves what's interesting about this dynamic is that it makes survival itself a kind of victory it just makes it kind of a slightly lower tier victory because at the end of the day there can really only be one most valuable player and that pressure to say well I've survived the last few levels I'd really like to be

the hero I'd like to be the MVP means that players will over time tend to stick their neck out they'll Rhea the way the risks and the rewards and they'll give it a shot and some of them will do it and that'll be the hero and some of them will die and then the next round they'll probably play a bit more conservatively this particular dynamic lends itself really really well to sort of the squad military style game because of that whole aesthetic the whole conceit of we're a military unit we have certain mission objectives we need to achieve but at the same time you know we have to keep the squad alive we leave no man behind that this kind of concept works very well with this dynamic and the recent Ghost Recon games draw one and draw two did a very good job of leveraging this approach now I can say with absolute certainty we did not use this dynamic in Splinter Cell Conviction coop at all because in Splinter Cell Conviction when one player dies it's game over for both sides now in our case I think it was the

right choice because it fit the fantasy of the two covert agents placed in harm's way unable to be able to complete the mission unless both players are able to do it so if it and it wouldn't necessarily have made sense for us although something we would explore okay so those are the seven dynamics and I'm gonna just go through some of the lessons that we learn on conviction as a result of seeing how these things worked in practice the first thing that we found is that more players in co-op we're tending to use the mark and execute system intentionally than they were in single-player why because we let them share marks we let the players mark individual targets on their own and then we let either of the players execute on those marks now initially players would mark targets because it was a great way to communicate with each other it was a great way to say oh hey there's a guy over there he just walked around that quote there okay there he is and then

that created an opportunity as players were earning execute tokens for each other by taking enemies I hand-to-hand for the for the players to use mark and execute in exactly the way we designed it next and this gets back to what I started to mention before we got real drama like real the sort of true archetypal water-cooler moment out of the interaction between our punitive systems because it was possible for a player to be wounded and to be immobilized on the ground bleeding to death to have their teammate come rushing to their rescue and to be within steps of reaching them and then get grabbed by another enemy in a chokehold at which point the player on the ground was able to draw a sidearm they could scream at each other over their mics duck duck now doc and the guy would duck and say take the shot take exactly like we've seen in every movie a hundred times over pop him in a head shot with

the both of their button arrow health bars shrinking as they as the whole thing's going on and then the the freed teammate would race to his friends aid and rescue him at the last second we knew it was possible when we were designing the system but we thought yeah it's probably kind of an edge case we'll be lucky if we see it once in a while happened in our first play test it was honestly probably the most ecstatic moment I've ever had as a game designer here we had these players who up until that point had been playing a pretty to be blunt shitty mediocre build of the game struggling with all the different problems that we were having with the controls of that moment didn't know each other because they were just guys we pulled into the play test suddenly they're like high-fiving each other and you know they're best friends it's good stuff the fact is while they were useful and they helped us to in a variant pre-planned way author moments that fit the spy fantasy the fact is exotic

gameplay moments exotic challenges did have the effect of exposing limits on the interactivity of the player because up till that point the players using core gameplay mechanics is in the groove is developing a kind of familiarity that kind of literacy that we talked about the ability to optimize the outputs of the game the ability to cooperate with each other and communicate with each other and then all of a sudden we're imposing a very specific constraint on their action it's still a tool it's still a useful tool but we need to be aware the fact that it it's a big sort of screech in the in the flow of social interaction in the course of the co-op game something that we we realized and this wasn't released a big surprise because it's it's also true in single-player James is a lot to a large extent is that high-level plot is never going to be as important as the kind of low-level interactions that we support between characters and in our case between players and in our case we were able to exploit that we

built a lot of the dialogue between our two player characters around systemic output so in other words when certain kinds of gameplay outputs would occur the two characters would acknowledge it they would share a bit of dialogue and we allowed the nature of that dialogue to evolve over the course of the whole game so these two characters who in the in the plot aren't supposed to like each other at the beginning of the game are kind of slowly beginning to respect each other and this is reflected in their systemic barks in there and in their gameplay dialogue players appreciated this it was in fact it's the only thing they talked about when it came to story when we read the forum's they they'd long since forgotten what the plot was something about umps and Russia and mobsters and things but they talked a lot about Archer and Kestrel and the fact that these guys didn't really like each other but they seemed like they were buddies and then and I'm not gonna spoil the ending so just to kind of go over some generalized best practices in

order to wrap up I think the first thing is if you don't have to try not to pour production values into highly prescriptive systems all right you're not fooling your players if you stick an invisible collision or a glowing wall or whatever it is your players will understand what it is they'll realize what you've done and they'll understand that you had a reason for doing it they might not like it but they'll move on if you put if you attempt to make it look as though everything is normal and okay then they are going to bump into that collision in a more painful way because they will be trying to use their collective agency they'll be talking to each other wait no maybe we need to go around to the left okay you try going to the right okay see if you hit that okay no actually no you're just not allowed to do that so better to go the gamey route and embrace it and have them deal with it than to try to convince them something else is happening whether or

not you agree with any of the dynamics that I described just please my appeal to you is give the players tools for mutual support you can find your own language for it you can find your own way of doing it but that's really the key here this is this is what matters recognize that when you have more than one player in the environment that generates noise it generates all kinds of noise it generates the actual noise of multiple players characters moving around in the game space it generates of the sensory noise it generates a systems noise of multiple players perturbing the system the AI and all these other things in lots of different ways that that make the environment more chaotic and you have the social noise of these two people who might be sitting on a couch or might be talking over over their microphones distracting each other so when you're designing the ingredients at the level of kind of aesthetic level of level design focus on readability on the legibility of those ingredients first then make it pretty to give you a good

example of this on conviction in the single-player maps which were beautiful lighting was initially done as a kind of an art pass basically the lighting fit the logic the architecture the the intentions the mood of those maps and it was beautiful and once it was beautiful then as after multiple iterations looking at heat map results all this other stuff the team was able to go in and tweak and adjust the lighting to work well with the a eyes detection model in the case of coop we had to do it a hundred and eighty degrees in the opposite direction right we built our lighting was giant circles of light like you'd see in a Bugs Bunny cartoon and we would sort of string these circles of light to create our shadow pass and once we knew that we had enough of a shadow path to accommodate more than one player then we do a pass to try to substitute those light sources with something maybe it felt a little more logical and then we would go through and do these iterative passes to make it more discrete and more subtle and have it look a little bit nicer and

the message there is players will forgive that even if they were to do a side by side comparison go whoo geez man co-op lighting looks a little old-school it's okay they are more interested in having the ability to do what they want to do than they are in weather that the lighting pass was done exactly the same way between a single player in koa finally and this is sort of the high-level takeaway I guess is you need to trust that your players will find a way they will if you let them work together and if you give them a good reason to they will find interesting solutions to the gameplay challenges that you present them with that maybe you wouldn't have anticipated so after everything else I've said at the same time don't freak out don't feel like you need to control the player experience for these two people or three people or four people their numbers will give them the added collective intelligence to tackle these problems now there's some things that I think would merit further investigation they're outside of the scope of this talk they were almost within the scope

of this talk until I realized that it would be a two hour long talk I think we've talked about dynamic solutions that are meant to be relatively accessible to lots of different types of games but it would be an interesting discussion to see how specific aesthetic Direction specific level design solutions could also help make a particular game type more conducive to mutual cooperation and meaningful cooperation there's an entire discussion about how we can use presentation design in other words the HUD system cues and prompts and visual signs that feedback inside the game environment in order to give players a stronger sense as to situations where they can cooperate or where cooperation might be a good idea and then I think there's an entire talk or two on the subject of tools for communication and coordination which is really communication coordination being

something entirely distinct from just straight collaborative working together to solve problems but it feeds into this and there's so many dependencies here that there would be an entire discussion on it so these are all topics that I encourage any of you who are involved in the development co-op games to think about and maybe offer suggestions on but it's a it's a GDC talk for at least a few people in this room if you want to give a shot and that's where I'll end it so I'll start taking questions we have two mics you know

yes absolutely the the it's when you say that Sokka Appa Titian which is the horrible portmanteau that we use for this is is absolutely has a role in probably the majority of cooperative games because at some point when you reach that level of mastery where there's so much shorthand and these strong social links are allowing players to use shared intentionality and then and that you start to want more and and what's fascinating about that is we can either again throw that entirely to the players discretion and just try to cede the world and cede the game system space with lots of opportunities for one player to kind of elbow aside his his fellow citizens and in his own pursuit of his own glory or we could do it in a

really deliberate and intentional way my suspicion is the more deliberately and intentionally we try to support coopertition the more will screw it up but the long assured of is I absolutely think there's a role for that I mean I think there was there was clearly evidence of it in in Splinter Cell and I think probably we would love it if we could support that even better than we did yep that's a really that's an excellent question so so how do we avoid the kind of the kind of turret tyranny of the of the elite player versus the new band in a co-op situation I think one thing we

need to understand is that okay we are not ourselves going to be able to avert bad behavior right I mean be any attempt we make to police that or to sort of aggressively kind of punish that is going to fail badly it's just simply going to degrade the experience for everybody so it's a certain amount of this that just needs to be self pleased needs to be people trying to act somewhat like civilized human beings although that's always a tall order online but the other aspect to it is just that we need to bear in mind that co-op is often the easiest way for a new player to get into a game so when we're designing our co-op modes with co-op modes we need to be thinking about that it's there is often a temptation to think that oh well because it's multiplayer it must be more hardcore right but maybe we don't need to think of it that way maybe what we ought to be trying to do is imagine that it's the perfect opportunity for an experienced player to kind of help bring up to speed a less experienced player help them level up a bit faster kind of twink them along a little bit so that then they're

able to get into playing single-player and really get to know the system so I don't know how to change human nature but I do know that maybe we as developers need to be careful that we're not deliberately making our co-op modes the more hardcore mode that would be my solution uh-huh yeah right I mean you know that it's a super good question I mean we had a we had local co-op on conviction we had split-screen for example and and we felt like it was super important to support it I think the the main issue like there's a bunch of a number of usability issues that come about when you're sharing the same kind of screen real estate which is again outside of the scope of this but I think again that almost is the perfect example of where you have two players who are co-located they're geographically sitting on the couch with each other right or at least

in two different rooms if they're playing over sis link or something like that but presumably they know each other right they're talking to each other in like neat space they're not on their microphones they're like throwing beer cans at each other and all this kind of stuff so in a sense that's a classic set of conditions in which the player in a sense is able to supply most of what's needed in order to do this to facilitate the communication the shared intentionality has spilled out of the game and into the real world into the living room or wherever so that they're able to kind of do that negotiated action in a way that's very human and you know ancient so to a certain extent that's an example of where we almost don't want to intervene because they're gonna figure it out on their own I know it probably sounds like that's my answer to everything is just don't intervene but but in this case it's true it's it's like it's much harder for us to negotiate these things with online play because we just don't know what the conditions are and you might have a guy in in Lithuania playing with a guy in Buenos our ice they literally have no common framework in which to speak to each other and one guys you know been up since 3:00 a.m. and the other guys you know just woken up so it's it's I mean

those are the hard problems for us to solve in terms of social interaction Thanks right I so so I think I mean for me when I'm using the term exotic challenges and exotic moments I'm really talking about situations in which we have fundamentally altered the sort of interaction space of one or more of the players like we've changed their camera we've changed their controls we've changed the way they move we maybe even have altered their perspective completely so the situation you're describing where maybe one play the players have been forcibly separated you know one guy's fallen into the pit and is now fighting the Sarlacc well you know so-and-so is trying to whatever call in an airstrike I mean those those

what you're describing are situations where players still have access to their core game mechanics and to the normal level of agency that they do in the game world they're just they've just been confronted by a level design constraint that causes them to be separated now in those situations I think that you know that creates an interesting variety I wouldn't want to abuse that I wouldn't want to force players to go through that more than four more than maybe five or ten minutes before they get back together but I think that sort of situation works out rather well I'm talking about situations where you know you're the first guy to hit the trigger and so now all of a sudden you're the one that's flying the by point around and you can't do any of the things you normally would do and you're vulnerable and you're you know those are the situations that I think we need to use sparingly or at the very least understand that players will recognize that it's imposing some kind of limit on them Thanks yep awesome right yeah yeah yeah yeah no

it's really really good point and I don't know if you noticed like for combined action yeah I had my screenshot from Starcraft 2 that was like my token RTS mother and I didn't even talk about but but you're no you're correct in fact I mean I find it actually weirdly like sports games and RTS are like it's like these two sort of subgenres of game that we and everywhere in everywhere else land often fail to look to in order to find lessons but you're right we absolutely looked at that and actually I'd say that that over the course of the last few games I have worked on and dot dot we have spent more time paying attention to real-time strategy for the lessons

that we can we can extract from it and and yeah it's it's true like I think you know I think for like combined action in Stargate or a Starcraft 2 is a perfect example of this right it's that kind of ability to share resources as you know in a kind of intentional way that I'd say is perfectly exemplified by that thank you yeah yes yes absolutely yeah oh I see I mean I mean okay now you're getting into stuff that's under the hood that the players maybe either don't perceive or aren't sure about I mean it's interesting we've had people ask us okay so if one guy's fallen does that mean you're more likely to get grabbed in a chokehold the answer is no

I mean the only the only way in which it's more likely is that the second player is running to get to his friends so he's automatically being not as careful it's not as conservative in his movement and he's heading straight for the place where there was enemy activity so it really is an emergent outcome of those two systems and we didn't have to do anything to Monkeywrench it and I probably would be very very cautious well as no big surprise there I don't think we should intervene I wouldn't I wouldn't probably try to force the issue what I think is interesting though it's a more interesting challenge is can we come up with like a third system that fits in perfectly and it kind of rocks paper scissors away with those two that's that's a tougher problem and but you're right it's AI it does seem like it gives us a more organic moment that players really feel ownership of and I think I'm getting the eyeball from everybody okay yeah last question yeah

I mean anything which depends on incomplete information for one player the other is probably not going to be very effective in split-screen or couch co-op but I mean I would have to look at the whole list and probably do a kind of build a matrix for it but yeah yes yes cool I think that's it thank you very much everybody [Applause]