Master of the Katana: Melee Combat in 'Ghost of Tsushima'

[Music] hi i'm kristen merman i'm studio head emeritus at sucker punch i did a lot of work on melee combat and ghost of tsushima so for the next hour or so i'm going to try to pass along some of the things we learned while working on ghost since i'm going to cherry pick a few things that worked well and skip all the things that didn't work well at all it's pretty much a normal gdc talk here's how the rest of the talk is laid out i'm going to outline five basic problems we ran into trying to get melee combat to work i've got a case study or two of a particular instance of each problem type and after going through the case studies i'll summarize what we learned with the catchy sounding rule about game design at the end of the talk i'll give you a slide with all five of these rules for game design also i did a lot of the programming on this stuff so there are some technical details the technical details are going to be uninteresting to many of you so i'm not going to talk about them in the talk instead i'm going to take advantage of the special nature of this year's gdc by dropping them as texts into the chat window when we hit

the appropriate part of the talk so watch the chat window for the for the notes next time i'll remember to do the slides i'm guessing most of you have played at least some of ghost if you haven't i recommend doing so first of all it's an excellent game second a lot of what i'm going to talk about is going to make sense if you haven't but a quick recap just to remind you of what the game's about ghost tsushima is an open-world action rpg set during the first mongol invasion of japan in 1912-74 this invasion really happened its historical fact including the subjugation of the island of tsushima in real life the 80 samurai on the island were all killed in our story one samurai jin sakai survives the opening battle and decides to fight back against the mongols he's massively outnumbered in order to fight back he is forced bit by bit to abandon everything he's been taught about being a samurai jinn learns to fight from the shadows sacrificing his samurai identity to save his home so that's the setup here are three primary goals that constraint or

goal slash constraints we adopted while building the game first the game is immersive our overarching goal was to build a time machine to transport the player back in place in time to ancient japan so that everything melts away and you just feel like a samurai on tsushima island fighting off hordes of mongols second the game's grounded while we were trying to build a historically accurate game we were trying to build a believable one that means no monsters and no magic just mud blood and steel the catch phrase we use to keep ourselves honest and third the game's accessible for just for ghosts to justify it's not insignificant budget it needed to appeal to a broad audience we didn't set out to make a niche game we set out to make a game that millions of people would enjoy these three goals slash constraints shaped a lot of the design so keep them in mind as we move forward so how does melee combat fit into the game well melee combat and specifically katana combat is central to the samurai fantasy without katana combat that looked and felt great we couldn't succeed furthermore melee combat is the backbone

of the story we were telling the whole story relies on jin sakai being forced by the sheer number of enemies to adopt underhanded tactics if you can just hack and slash your way through melee encounters without fighting dirty then the story falls apart the whole story hinges on melee combat overwhelming you as a player while still feeling fair just as it overwhelms jen to force you into these decisions jyn doesn't want to make now the samurai fantasy about melee combat is broad fighting in a duel with a single worthy enemy is certainly part of it so is a desperate fight with one surrounded by a group of opponents and so is fighting alongside your fellow samurai in an epic battle ghost melee combat needs to smoothly span this whole gamut from pure brawler action to a challenging one-on-one duel and that's a tall order especially since the last mainly focused game we built shipped 15 years ago and featured a talking raccoon so let's be more precise about what constitutes the samurai fantasy for a ghost we just decided the core of the samurai fantasy is discipline and precision precise control with every

movement having a purpose a sense of stillness and readiness despite danger and a burst of explosive action at exactly the right moment and discipline in all things discipline in the moment to hold back on action until the perfect opportunity discipline over time as your extended practice and hard work leads to mastery over the katana and the other soldier the arts that you study and progression is part of the samurai fantasy too determined disciplined hard work over time leading to mastery that's not dissimilar to the experience of playing the game actually so keeping the player's journey in tune with jyn's journey was an important goal above all the samurai in our mind's eye was skillful a master of the katana so we need to decide how the player should skill there are lots of ways for players to show skill and games precise use of sticks in a shooter precisely judging distance between two characters as is the case in many melee games we chose two things first precise timing to echo the precision the samurai fantasy and two second a quick and accurate reaction

to enemy actions both having to be demonstrated and performed under high pressure from your enemies but there's a problem with this choice human reaction times are slower than you think for a simple visual stimulus and a simple player response it's maybe 300 milliseconds that's slower than you can swing a sword slower than jin's fast attacks in the game which take less than 250 milliseconds you can work around this though there's no limit on enemy reaction times in many circumstances they can and should react instantly to player action in fact we did some early work to give npcs more reasonable human reaction times for instance we had npcs react to your movement with a human scale delay if you dodged to your left the npcs took a few hundred milliseconds to track and this was a disaster it felt wrong sloppy buggy didn't feel right at all and i think this is because our expectations for what a sword fight looks like are formed by movies and movies aren't very realistic everything is choreographed so the actors know what's going to happen next they're anticipating they're not reacting real sword fights like real

fights of any sort are incredibly messy but the fantasy we're trying to hit is the movie fantasy of a sword fight clean and precise so enemies track your movement instantly tech note it's worth noting that human reaction times are slow but human anticipation can be close to perfect it takes a while to react to a stimulus but you can all clap along to the beat of the song um and one main goal one final goal slash constraint for combat is that we wanted the game to be purely skill based and part of this decision was that if you play the game perfectly you won't get hit no matter how many enemies are attacking you um if you do the right actions at the right time you will not get hit this felt like the right fit for the samurai fantasy that if you played perfectly you could dance between the raindrops and make it through any situation unscathed so where does that leave us well initial enemy attacks have a wind-up so that you have time to react follow-up enemy attacks in a combo can be fast since you're anticipating instead of reacting and player attacks

can also be fast since npc reaction times aren't limited but we ran into an immediate problem with this basic design quick attacks with square were too effective we assumed i have the reasonable assumption that stabbing someone with a sword would interrupt their action um we were trying to stay grounded so this seemed like a reasonable decision but with slow mpc wind-ups to let the player react and fast player attacks you could just poke guys during their wind-up and win this did not test well but mashing was an effective strategy but it didn't feel very skillful and it wasn't very fun and honestly we might have rebooted the whole approach at this point if i weren't studio head as well as the guy working on combat which gave me a certain amount of latitude to stay the course most immediately though we needed to work out ways to blunt the squarespan problem we looked for natural limits believable ways to stop quick attacks from being too effective so we added a shieldman archetype that blocked quick attacks this made sense to players they didn't object later on we added brutes big guys guys

were so big that poking them with a sword wouldn't interrupt their attacks and this also worked but more importantly we started overlapping enemy attack sequences we recognized that we could start an npc wind-up while the last npc was still finishing their attack sequence as long as all the attacks landed at distinct times the sequences could overlapped this turned multiple enemies attacking to you effectively into a single sequence with a long wind up only import only happening early at the beginning of the sequence in the end there wasn't any magic bullet here we wanted quick attacks to be fast and powerful we wanted you to feel like a samurai when you did them it took lots of little fixes to help keep to keep the heart of this quick and powerful attack intact while damping its over effectiveness the first thing that we added that actually made you feel skillful wasn't the quick attack it was a timing based parry delaying your block until just before an enemy attack lands to get a better result most npc attacks are

blockable especially in the early parts of the game it's easy to block attacks you just hold l1 and you're safe you're able to stand there and you'll block attacks but this simple and straight safe strategy doesn't provide any counter-attacking opportunities with perry's though the enemy is spun past you interrupting his combo and leaving him vulnerable we even interrupt any wind-ups as buddies have started to give you a clean chance to counter-attack and this felt great player connection was excellent you felt like you actually parried the blow and remember that's our goal our goal is to make the controller disappear to make the player feel like they actually swing the sword or parry the strike not just press buttons to make those actions appear skill-based pairs were our first real immersive success it felt dangerous to delay your block and skillful to pull it off but there was an exploit with this there was a cheese opportunity you could flutter the block button press block very quickly and parry every single attack you should press the block button fast enough that you were guaranteed to hit the wind up an enemy attack so while we loved the pairing mechanism when it worked there was a spark of

magic many players couldn't leave the cheese alone once they discovered this exploit they couldn't stop themselves from using it even though they weren't happy and this was an internal play test these were professional game developers they knew that this was a bug they knew that this was an exploit they knew we would fix it they knew they weren't having fun but they kept doing it we eventually found a way to keep the core magic of perry's intact while eliminating the exploit and there's a tech note about this but to simplify we track an effective block time that ignores button mashing and perry's end up being pretty foundational to ghost melee combat here's a general game design rule that follows from these two examples and many others like them you might think of this as the hide the cheese rule most players will play the game and what they perceive as the most effective way to make progress even if that leads to them having less fun even if they know that playing the game that way isn't fun they'll still play it the way they think

is most effective and they won't be happy they kind of have two choices right either they'll play the game in the way that is effective but not fun and resent you for making enough fun game or they'll play the game in the not effective but fun way and resent you for making them into saps so make sure that the way the players feel is the most effective way to play the game is also the most enjoyable way to play the game otherwise you're going to get yourself in trouble so it was pretty clear from our testing with players that varied uh play styles were the key to enjoying the game war if players who mixed up the way they played the game had more fun and they scored the game higher if they used all of jen's abilities melee ranged combat ghost tools stealth they consistently reported higher scores we wanted to nudge more players into varied experiences because we thought that would lead to more enjoyment and it did turn out this way but it might not have right the causality arrow could have pointed the other direction it was plausible in early tests that

what we were seeing was that players who were enjoying the game were trying out more parts of it rather than the other way around but in fact this did work that as we encouraged players to mix stuff up our scores went up people had more fun so there are two basic ways to create a varied play experience you can vary the situations the player finds themselves in so that they're using the same abilities in different contexts or you can provide lots of options for player engagement hopes that they'll use different abilities in the same context or most likely you'll provide both with a rich set of player verbs and a similarly rich set of game provided experiences let's start with the player's ability set providing variety by giving players the ability to do different things well providing a rich verb set that gives players agency on how to interact with the game's challenges is a great thing letting the players choose which verbs to use and how and when adds agency and that's great but the problem with player

agency is that players tend to not use it the way you want them to in particular players tend to strike stick with the strategy that they find that works the first strategy they find that works the psychological term for this is satisficing their approach works so why experiment with stuff that mostly doesn't work that's especially true in a game like ghost where you're under pressure the whole time you don't have safe spaces to experiment the only place to try new combat techniques is inside of combat so you stick with what you have so generally ghost players needed a nudge to try new things there are two new ways to there are two main ways to nudge players you can discourage them from the strategies you think they're overusing or you can encourage them to try new strategies the ones you think they're under using so you might try to discourage an overused attack by introducing an enemy who blocks those attacks perfectly but is vulnerable to the attack you'd like to see more of that's frustrating the sort of locking key design breaks the immersion we were trying to create players can tell that they're playing a game and they don't like it so much

similarly you could limit an overused ability by having it consume ammo or adding a timeout or having its effectiveness to grade over time or when overused but why would you do this players were using the abilities they thought were effective and enjoyable why take that away from them by making those abilities less enjoyable and less effective so we focused instead on encouraging use of underused abilities rather than discouraging the use of overused ones we focused on carrots instead of sticks if you want to think of it that way in our early tests it goes to shema for instance we did not see varied ability use uh in our play tests we track every attack launch by the player and early on player attacks were overwhelmingly light attacks um quick attacks like 90-something percent for many players players who didn't just use quick attacks though who mixed up their play styles scored the game higher they had more fun we wanted to nudge more players into that varied play style our goal was to make heavy attacks more appealing instead of just discouraging light attacks first we had to make sure that the value

statement for heavy attacks was clear what is the point of using a heavy attack as a player why do i do it in ghosts the reason you use heavy attacks for the most part is that they let you blast through blocks for players who want a direct approach who want to just walk up to somebody who's blocking their attacks and knock down the block and then and then hit them with their sword and that describes a lot of players this is pretty appealing the ability to have a direct path through a block and this was mostly a communication problem we could know that heavy attacks were breaking through blocks and in fact we tried an approach where it only took a single heavy attack to break an npc's block but this felt pretty mechanical we decided we needed to have um to require multiple heavy attacks to break through a block but communicating the the player was making progress during this was a challenge and unfortunately our best solution was adding ui to the screen which we didn't really want to do since we were trying to minimize ui we added a meter that showed remaining block points but that's not its real purpose the real

purpose of the block meter isn't to show how many block points are left it's to flash when you hit the guy with a heavy attack um in effect it's just a signaling mechanism to say the thing you just did was a good thing and it had an effect um you can tell how many block points someone has left but that's not really the point and in the end this helped this was what we needed to do it it did help people understand why they're using a heavy attack and so we were willing to accept the compromise and actually add some more ui to a game that we really didn't want to have ui in next we improve the effectiveness of heavy attacks we made the attacks faster to make them easier to use we increased the amount of block damage they did to make the direct path more appealing and these are easy changes because they both make the game more grounded at the end when we ship our light attack our or quick attack usage was under 65 percent of all player attacks with heavy attacks a little over 20 percent the rest of the smattering of super attacks go stance attack standoffs and so on and we were okay with that that was an easy that was pretty good mix

similarly it was a challenge to get players to use ghost tools and this was a problem since ghost tools were a big part of the story we were telling a story that jin sakai had to adopt underhanded techniques to win players were comfortable moving through the game with their conventional tools my katana attacks in range combat luckily most players were happy to use assassination so we weren't completely out of luck on underhanded technique underhanded tactics but again players that did use ghost tools scored the game higher so we wanted to nudge the players in that direction again we could have nerfed all the conventional player abilities but players were enjoying them and we didn't want to ruin that we could have added lock and key designs npcs that were only vulnerable to a particular ghost tool but that also seemed like a bad direction so we focused on making ghost tools more appealing the most important changes were to make the tools easier to use we started with kunai the throwing daggers and ghost because those were the most popular tool the most important change we made was to

have the players start by throwing multiple daggers not a single dagger giving the constraints in our aiming model that it had to be fast that you were using the joystick that daggers had a reasonably long range we knew there would be some imprecision in the aiming that players wouldn't be able to perfectly snipe a target so we leaned into that by throwing three daggers we tripled the chance that you'd hit the target you were meant to hit we do give you feedback on which target we thought you aimed at it's the middle one of the three but players didn't care that we misinterpreted their input they only cared that the target they wanted to get hit actually got hit there's a tech note about this the second big change was to lock you into one ghost tool at a time previously we thought that allowing you simultaneous access to all of your ghost tools would increase usage and there were people who passionately believed that they wanted to be able to instantly throw a smoke bomb or a dagger or a sticky bomb the last version about this simultaneous input model was chorded you would hold r1 and then you would press a face button with r1 down so r1 plus x would

throw daggers but we weren't seeing people use ghost tools and we knew we needed to change stuff so we switched we switched to a one ghost tool at a time model where you only have one ghost tool assigned to r1 at a time pressing r1 will fire that ghost tool it's easy to switch but fundamentally you only have one thing at a time and someone counter-intuitively this created immediate uh wins like immediately we saw 30 to 40 more use of ghost tools and this is sort of weird right that like we've made it harder in some sense to use ghost tools by making you assign the one you want then use it and yet the usage went up so why did this happen well you know if you look at what we've talked about earlier in this talk it's pretty obvious why um we had added too much cognitive load to the players right it was there were too many things for them to think about there were too many options for them to consider and too many controls they could hit they were under a lot of pressure and so they reverted back to doing the things

that you cannot they hit square and triangle and made it through encounters that way by simplifying it down by reducing the cognitive load they were able to think quicker and they could keep up with the game and they could hit r1 and fire the kunai when they wanted to the third thing that we did is a little bit more subtle it's to make sure that the ammo model works out we we restrict your use of uh ghost tools through ammo and the thing about ammo limited models is that people will only use them if they feel confident that more ammo is easily accessible that if they're not if they feel like they might run out of ammo they'll just hoard it so first we had to make sure there was lots of ammo in the world which we did and then we had to play some sneaky games so in ghost if you're low on a particular kind of ammo you're more likely to get that in an ammo drop we'll lean on the dice rolls a little bit but and this is important even if you're full of kunai we will still drop kunai for you occasionally and that's important because if we stop

dropping kunai players will conclude there's not a lot of kunai in the world and they'll stop using them so even when you're full you'll still get stuff um to so that you think there's lots of it in the world and this all works out we end up driving ghost tool usage up to being like five percent of the attacks and that's not bad so what's my takeaway from these two case studies well uh first if you're trying to get a player to use some ability don't try to force it don't make them use it don't create enemies that are only vulnerable to the new ability especially don't hamstring the players other abilities to get them to use the new one instead work to make the new one more appealing make the value proposition obvious be sure players understand the reasons why the new ability would be useful make sure the ability is easy to use and obviously effective if you must give players an external reward for using new abilities like an xp bonus but don't hit players with a stick they

don't like them they don't like that hand them a carrot instead let's switch to the other way of creating variety having the game present new situations for the player to deal with and in particular let's rotate back to one of the first issues we talked about the limits of human reaction speed reading situations and reacting quickly and accurately is a core part of the skill challenge ghost presents for players but player reaction speeds aren't constant they depend on a lot of things for a simple stimulus and a simple response you can expect a quick reaction time think of a line of sprinters all in their starting blocks waiting for a starter's pistol to go off it's a simple stimulus and a simple response adding more complicated stimuli slows down reaction times so watching for attacks from multiple npcs unsure which of them is going to attack will slow you down or needing to look at different cues different wind-ups for different attacks from a single nbc that will slow you down as the complexity of the situation goes up player response times get longer

the same is true for complexity and player abilities as the player considers more options their reaction times slow down those sprinters in their blocks only have one option to start running but the player has many block dodge light attack heavy attack super attack jump attack ghost tools all kinds of stuff right remember that quick and accurate reaction to this situation was one of our core skill tests when it works well when we're dancing at the edge of what the player can do then successor is rewarding if we undershoot if we don't provide enough complexity then players react too quick find the game too easy and find exploits if we overshoot we have too much complexity then players are frustrated and feel unskilled and so finding the right balance is important we have to ladle in the right amount of complexity so that it feels challenging and fair but is a challenge the players can overcome let's look at one way in which we ladled in more complexity to create a more enjoyable challenge and it's archers we couldn't very well make a game about an army of mongols without including archers our

our early integration of archers was problematic especially in the intersection of range combat in melee combat which is to say what happens when you have archers firing into a melee fight or what happens when you get into a fight with archers around well you have some options you could let archers fire into fights and we tried this and it was not fun you're in the middle of a high intensity sword fight you've got your hands full dealing with people with swords and spears and you keep getting shot from off screen by archers so what do you do as a player well you start every single encounter by locating and eliminating all the archers you stealth their way around and kill them all and this forces into a single repetitive pattern in every encounter which is not what we are after or you can decide that marchers don't fire into melee scrums and we tried this it was better but still not fun once you got into a sword fight the archers that were nearby no longer had an effective footprint they didn't have any effect on the fight you could just ignore them until you killed the last enemy melee enemy at which point the archers popped up now if you remember they were around

this wasn't a surprise but pretty often you'd forgotten there were archers around during your melee fight and your first uh reminder of this was getting shot with an arrow so neither of these two solutions are are awesome neither uh shooting into fights or not shooting the fights none of these worked very well um so uh we but we kind of needed to have archers nearby right and we needed to have them influence fights we because of the constraints we're under uh we don't have very many ways to mix stuff up right we don't have monsters we don't have magic so archers have to play a role a melee fight with archers has to feel different than a fight without archers so what's our solution our solution was to turn archers into an audio queue instead of a visual one when archers are about to fire they yell nosh that's duck in mongolian by the way anyhow when the archers yelled doge you'll see the other mongols duck which amplifies the queue and players believe it or not we're quick to accept this

in-game fiction you would think that maybe they'd say this doesn't seem right but in fact they're happy to take that there's a tech note about this as a player you're asked to hear the cue then react by dodging and if you do you won't get hit with the arrow direction doesn't matter you don't have to track where the archers are if you hear dosh you can dodge and you're not going to get hit and this gives you one more thing to juggle during a fight in addition to watching all the nearby mongols for visual cues you need to listen for the audio cues for incoming arrows and more cues means more complexity means slower player reactions which means a more challenging and rewarding fight and it means that a fight with archers nearby has this added element that a fight without archers doesn't have giving us another way to vary the player experience so what's the takeaway here well there are lots of other ways we mix things up and some work better than others so we added falcons to make stealth more of a challenge or we created different mixes of enemy archetypes or we created enemy archetypes that had really different

kinds of attacks or we had you fight in areas where your mobility was more limited or we had to fight alongside allies in a lot of ways it's easier to create a varied play experience through very content because players can't avoid it they can choose not to use the abilities you provide them but the encounters you place them and they they can't avoid uh it's also a great lever to nudge players into mixing up their play styles in natural ways a well-designed encounter can nudge the player who mostly engages in melee fights into getting into an archery battle or can nudge players who really like to fight things straight up to use stealth occasionally so mix up your uh so externally for changing the situation is a great way to provide variety so let's talk about progression the samurai fantasy is about discipline and hard work leading to mastery that's progression and that's a journey we wanted the player to go on too one of the convenient things about ghost for us as game developers was how tightly the player's journey matched jyn's journey

in particular the player gets better at playing the game with practice they're more precise with the controller their timing gets better their decision making becomes faster and more accurate adding more complexity slows players down so there's a temporary setback whenever we add a new concept but their decision space getting more complicated their decisions getting slower that improves with practice responses migrate from the slow parts of players brains to the fast parts they do less thinking they react instinctively they get better at playing the game and the players knowledge increases as well they get smarter they learn new enemy types they learn or invent new techniques they spot opportunities they would have missed early on becoming more skilled becoming more knowledgeable both of these mirror jin's journey in the game and that was an important consideration for us and they're both part of progression now when you saw a slide about progression you probably thought we were going to talk about how the gen character learns new abilities or gets new gear and that's important but it's only one leg of the progression tripod here's how i

propose thinking about progression it's not just character progression your character is gaining new abilities or gear that's an important part of the overall progression model but so is players getting better at playing the game and so as players getting smarter about how they play the game about players knowledge increasing we thought about each of these separately for ghost we wanted each to play a distinct role in the player's journey through what turned out to be a pretty long game the balance for your game will be different the way you assign to each of the three areas will be different than it was for us in ghost but thinking about them as distinct forms of progression will lead to a richer player experience and here's the thing about progression players are reluctant to try new things not all players but many players maybe even most players so if you introduce a new player ability part way through your game and let players decide whether to use it or not many won't here's the key make the first experience awesome the first time players use a new ability it should be easy to pull off obviously effective and emotionally rewarding if it is they'll use it again

and you're in good shape if it isn't they'll go back to what they know be careful that your efforts to create interesting progression for an ability don't ruin the player's first impression of that ability it's easy to decide when you're working on progression that what you need to do is weaken the first ability first version of the ability so that the player has a reason to upgrade it don't do this if the first version of your ability doesn't grab the player the second version won't matter and in a related note be sure to front load your best ideas in your game don't save your best stuff for the end of the game the most important 15 minutes of your game is the first 15 minutes the most important hour of your game is the first hour if those two things aren't great if the first 15 minutes the first hour aren't great nothing else matters front load your best ideas if the beginning of the game isn't great no one will make it to the end let's look at an example about making the first impression a strong one in the evolution of ghost melee combat one of the most important things we added was stances stances weren't there to begin

with we added them midway through the development of the game in a series of incremental steps actually because we were frightened of adding more complexity to the game which was already too complicated obviously stances do make the game more complicated for the player there's a whole new set of things for players to worry about more options to deal with in every situation and we were worried that players would be overloaded well we weren't wrong the extra burden of stances did slow players down players were overwhelmed and the first times we tested this the take-up was lower than we wanted it to be but players really like the idea of stances they just felt clumsy trying to use them and no one wants to be a clumsy samurai many players just stuck with the first stance the one they were familiar with we offered a new stance they didn't even try it they knew how to win with the first dance even if it was a struggle at times they satisfied to use the word i defined earlier other players did switch to the second stance but then just stayed in it for the rest of the game and that's also not a great outcome so the key thing was to make sure that the first impression was

a strong one the first time they used water stamps which is the first stance you earn you start with stone stance you have to have a great experience so this first step was to make sure that the value proposition is obvious water stance is really effective against guys holding shields you will blast through their blocks twice as fast as you will with stone stance the number of hits to to break a shield men's noticeably lower players can count they can tell and we do subtle things too to make sure that the value proposition is clear like earlier we talked about how we sold heavy attacks by having the block meter flash when you did a heavy attack damage let you know you did the right thing well if you're in the right stance and you do it you get an even bigger flash we don't tell you this it's just sort of subtle reinforcement that you're doing the right thing step number two is making sure to put the player in a good situation we're not in control of the timing of the player getting water stance so we have dynamic training for it once the player gains water stance we

look for opportunity good opportunities to train it to oversimplify we wait for you to be one-on-one with the shieldman then the training kicks in clock slows down we walk you through switching to water stance and using water stance um and that's really effective it's a little bit heavy-handed especially since the training for stone stance is going to kick in right afterwards until you're in the first when you're in the first one-on-one situation with uh swordsman so uh but it but it is effective um and the third thing uh the third step for to getting people to really engage with stances was to add more anticipation uh to them we actually uh make a big deal of making progress towards your next stance it's hard work you know it doesn't come easy it doesn't come out of the natural play through we we make you go after we make you spy on uh leaders we make you kill leaders and when you do we make a big deal the fact that you've made progress towards the next stance and all this puts the player in the right headspace it prepares them for thinking that the

science is going to be a big deal so at the point when when they get it they go through this elaborate training they're predisposed to think of it as being an important thing and to engage with it and that was really effective so what is the takeaway from this um well we followed this real pretty closely through all the new things we introduced in the game we tried to make sure that the first impression you got was a really positive one so the first ghost tool you get is kunai because it's the easiest to use and most effective and we make sure that it's great out of the box and that you're trained and that your first experience is really a good one the first samurai armor the first samurai armor you get is better in all ways than the stuff you start with it's more effective it has better perks and it's much sharper looking the first time you assassinate somebody it's useful and effective there's no phase of clumsy and noisy assassinations when you're learning how to do it even though that would have made way more sense for the story because it would have made the first impression not as good the first experience with something new sets the bar for everything that sets the tone for everything that follows if it's not great then the player won't engage with

your feature they'll just set it aside and go back to what they know all right let's switch gears again it's pretty clear at this point if you've listened to this talk to this point that ghost evolved a lot during its development it was six years we changed our mind about a lot of things a lot of times our design process historically has been very focused on prototypes and iteration we did on the order of 35 full team internal play tests during ghost development we did about 25 external play tests and they were the external tests maybe half of them were took a week long or a week long and natural playthroughs of the game that's lots of testing and lots of iteration and if you're going to prototype and iterate your way like this if that's your way of getting through to a successful game then you have an important question to answer which is what level of execution is necessary to fairly judge an idea how developed does everything need to be before you can tell whether the idea is good or not how developed do the controls need to be but the animation or sound or camera design vfx haptics

writing acting like how good does everything have to has to be the most important obvious problem you can sorry the most obvious problem you can have is giving up on an idea before it works and that is a real problem but the worst problem is not giving up an idea that's never going to work because counting on a doomed idea is much better much worse than giving up on one that never reaches its promise once you start counting on an idea working out you start building around it and then it becomes really hard to drop the bad idea without reworking everything else so you keep funneling resources into it and it never really works but if you're going to prototype an iterate you can't expect you're going to implement everything all the way to shippable quality before you make decisions you just don't have time most of our ideas don't pan out so we have to build a prototype and test it and somehow we have to figure out how to get reliable uh data from that test so this can end in disaster right if you're not careful you'll get play test data

that pushes you in exactly the wrong direction and it's because play testers aren't great at projecting it's hard for them to play a prototype and imagine what it's going to be once it's once it's done once it's fully realized even our internal play testers even professional game developers aren't great at this it's a really hard talent to squint in exactly the right way to see what the prototype will become rather than what it is and there's one particular failure mode that's really dangerous once the play testers will round up on under implemented ideas if something is really obviously a placeholder or a stand-in they'll imagine something awesome in its place the platonic ideal of whatever it is that you did they'll say this is obviously not done it's going to be great once it's there and secondly they will round down on stuff that just needs polish no you'll hear things like uh this animation doesn't look right or the sound effects garbage or i'm not doing enough damage in this case or here's an exploit i found this i don't like this

here's a graph of the way this can run and we've seen stuff that looks like this graph way too often it's very painful for me um it shows player reaction to something you're iterating on time's moving from left to right here as you're making progress on an idea and like the players play testers perceived value for the feature or the ability is going from bottom to top so what happens well first you test the feature as a rough prototype and you know it's intended but none of your play testers get it uh we'll call that in comprehension the very beginning of the graph then you improve things you leave some placeholder elements but you get enough right to get people excited they round up on the parts of your idea that look like placeholders they imagine perfect solutions in all the empty slots this is going to be great they say that's irrational exuberance then you start finalizing polishing things you replace placeholder animations with final animations and placeholder effects with final ones you put in some haptics you play with the camera the feature is much closer to done so you test it again

and people hate it the probability function at this point is collapsed instead of possibility your play testers are just seeing disappointment and disillusionment that's despair now things can go one of two ways either through hard work and insight you overcome the disillusionment you find ways to hit the rewarding version of ideas people who are imagining during the irrational exercise phase and everyone ends up happy you end up with victory or you don't playtesters were imagining things you can't deliver and things never improve you end up in defeat and your project won't survive too many graphs that end in defeat uh and by the way from personal experience this happens at all scales in our play tests we see graphs that look like this for individual features or individual components or individual missions we also see grass for the whole product that looked like this and wow is it scary to see the scores for your game going down as you get closer to shipping so this is bad how do you avoid grasping

and defeat well there are a few things first you have to make sure you understand your experiential target you've got something you want the player to feel you've got some experience you want them to have for most of the stuff we did in ghost we wanted them to feel like an awesome samurai in different ways second there needs to be some spark in your idea that makes them feel this way some bit of magic the heart of your idea the best part of your idea we'll talk about what that bit was for a couple of ghost ideas in the next couple of slides but you have to know what this is when you're prototyping a feature then you have to figure out what will get in the way what will distract players from this magic what will confuse them about the key part of the experience what will cause them to not give you the feedback on the magic but give you feedback on unimportant stuff all of this stuff has to be representative enough to not be a distraction representative here means close enough to shippable that play testers don't imagine replacing it with something better

when they're telling you how much they like the thing that you're testing they imagine what you've shown them not something they're imagining for most of the stuff we did in ghost that means that controls have to be close to done because that's the hardest thing for people to imagine animation has to be representative you have to have representative sound effects the camera has to be representative sometimes other stuff like visual effects or haptic effects especially if they're playing a key part of the experience and this may seem like a lot to have to have done at a almost shippable level before you can count on your testing data that's a lot of stuff to polish but there's even more that i've left out there's no mention of animation variety or progression or character and environment models which don't need to be polished in our experience the ui is unimportant unless it's the thing you're testing enemy variety unimportant how the ability is costed or how it's constrained not important we generally don't worry too much about robustness the ability of failing in some weird circumstance doesn't affect scores unless it's an

exploit so don't worry about it backstory integration of the world all this stuff just doesn't matter it'll also get finished at some point of course but you can safely test without it you can assume that you're going to be able to fill in those blanks and not get not get caught out too often my phrase for this is clear path to victory if you're not super confident that you have a clear path to victory that you know how to get from the prototype to the finish line or aren't super confident in how players will react to the ways you're going to get to the finish line then those things need to be present and representative in your prototype otherwise you're rolling the dice anything else can be omitted just leave it out seriously don't even put a placeholder in you're just wasting your time creating stuff and you're confusing the players so just leave it out so let's look at a couple of examples let's look at standoffs for instance um so the inspiration for this is the sakura crossover movie sunzero at the end of sunderland the two main characters retreat to a hillside for a duel they stand each of them with their hand on their katana intense

anticipation each waiting for movement while the grass waves about them and it's still for an uncomfortable time then one samurai moves but the other moves quicker and the villain falls to the ground dead this is an awesome scene in an awesome movie which you should all watch um and we really wanted to capture that experience in ghost well what's the key to this what's the magic point of this and it's tension if there's magic it's going to be capturing that feeling of tension this sense of stillness and incipient danger that explodes into action danger and opportunity all wrapped up into one so everything in the design of standoffs is focused on increasing tension we take our time setting things up to let the tension build you watch the characters walk into place the control model isn't our standard attack model usually we have you tap a button to attack but here we have you hold a button to ready yourself then release it to attack which is inherently a tense thing and we use an e jutsu attack an attack where your katana starts inside the saya and that leaves you vulnerable too because you

don't even have a katana drawn and finally since it's a simple stimulus and response the enemy movements can be faster than their normal attacks they don't have long wind-ups in a standoff they just attack and you know this going into it you know that you're going to have less time to react than you do in normal gameplay and that makes it feel more dangerous everything points at heightened tension during the standoff so we prototyped this and there were glimpses of magic it really was tense but rewarding even in an empty test world so we had then the question we have to ask is what needs to be representative before we can trust the magic and in this case animation sound effects visual effects there's some time slow down which is important so we do that the control model and the camera but everything else is missing at that point we don't need to worry about we did this in test world it didn't matter so this was actually a point of commitment for us because we showed a standoff as part of our demo at e3 2018 so we were sort of committing ourselves to doing this and at this point it was only a prototype and this is not work in the game in general the

stand up in the demo was special rigging it was real you played it but it wasn't systemically there in the game we hadn't really tested it in the context of the whole game but we were confident we had a clear path to victory for everything else we trusted in the magic we trusted the prototype and we didn't worry too much about when and where you can do standoffs or why because the standoff model itself was so strong now as it turns out some of the questions we had to answer were hard to answer and we were a little bit overconfident and there was a lot of work involved to getting standoffs to the point where they work systemically in the game but it was mostly engineering work it wasn't finding the magic it was just making the magic reliable there is a tech note about this with more details so here's another example this also comes from e3 2018 chain assassination in that demo we showed jin leaping down from the rafters of the temple and assassinating three mongols in quick succession and in a very flashy manner it's a great power moment and the nice part about this move is that

it's naturally limited you can't do it all the time you have to have three guys standing near each other uh and they have to you have to be able to maneuver to get above the jumping range of one of them so you really don't have to worry about it being overpowered you can really crank things up the magic of this is pretty obvious there's this uh tension to get into a position without getting spotted uh there's a feeling of mastery once you're there and you're choosing your target and here for once we actually went with a very slow targeting model and ghost most of the targeting has to run really fast because we want the game to be fast paced we want there to be lots of pressure on you but in this case we're trying to draw out that sense of anticipation that sense of mastery that you're about to drop down on these guys so the targeting model is slow and precise and finally once you trigger the move you get this spectacle of this chain assassination with flashy moves and flashy camera and flashy some time slow down just spectacle so what had to be representative here for us to trust the magic animation uh getting the time slow down visual effects getting the camera in

this case we needed to have the environment be pretty representative at the temple in the demo because this is a stealth move and the environment usually plays an important part in stealth abilities but everything else is emitted we don't worry about anything else and that was all we had to worry about getting right once we had that stuff right we had a clear path to victory we knew we could get it to ship and form in the game and in fact in this case we were right this one was easier to finish than standoffs the areas where we left out things we didn't have done um in fact we're pretty straightforward to get working we have bugs to fix but nothing that really ever risked the feature so here's the final takeaway then um from from this from from the last role when you're working on something new the first thing you need to do is find the magic you've got some experiential target in mind some emotional response you want the player to have identify the magic bit of your idea that produces that experiential target when you're working on the idea you might only get a glimpse of the magic right it might not work every time maybe

your timing was exactly right on triggering the ability and it looked awesome and you felt like a star maybe the enemies you had lined up in exactly the right configuration and your animations looked awesome maybe the random number generator bounced your way and everything clicked when that happens when you let a glimpse of the magic when the feature works great once you've got the spark to build from your goal at that point is just to make the magic happen more reliably and usually that's a lot easier than finding the magic to begin with usually it's just engineering if nothing feels like magic and it has to feel like magic to play testers not just to you then you probably don't have a winner you don't have something you can build your product around not everything in your game needs to be magic but you have to have enough magical moments or you're not going to succeed and that's the fifth rule so i've got all the way through five rules um and here they are all on one slide five generals of game design we learned while making ghosted sushi i hope this has been helpful or interesting and i hope that one or more

of these rules will uh be applicable to your own games um and if not it was only an hour of your time and there were pretty pictures so thanks for listening and uh by the way if you think it would be fun to work on uh problems like this we are hiring um so you can uh go to jobsuckerpunch.com and uh get in touch with us so thanks again for sitting through this talk [Music]