The Burden of Proof: Narrative Deduction Mechanics for Detective Games

[Music] okay um i hope you all can hear me sorry for not being on the video feed uh my name's john ingold and i'm the narrative director of inkle hopefully you've heard of us we're a small independent game studio from the uk and we specialize in making narrative games we've been around for about 10 years and done a whole bunch of stuff so um here is some of that stuff uh some of which you may know some of which you may not know it's uh quite an eclectic catalogue in some ways but the one that i want to talk about today but i'm going to talk more generally about a kind of genre of games and some design strategies but the one i want to special especially focus on is the one that's in red at the bottom corner there which is overboard now overboard if an agatha christie style murder mystery game with the neat twist that you're the murderer and your role as the player is to try

and get away with it that sounds like a spoiler but it's not um that said the talk has got quite a lot of spoilers in it i haven't really been cautious to avoid them so hopefully you've explored overboard a little bit for yourselves and already formed your opinion about whether you like it or not so before i start talking about detective games which is what i really want to do i better talk a little bit about overboard just to ground the conversation so what kind of game is it well it declares itself as a visual novel it looks on the surface like a visual novel um it on every screen of the game pretty much it has these two characters within a location and they're talking to each other right and there are dialogue choices and those choices affect the way the conversation goes but in other ways it's really an adventure game actually there's a map screen which you can see in the bottom corner there and you choose which location you're going to visit and not all the options that you get in a given location actually are dialogue there's lots of actions to take you can pick things up you can drop things there

are keys and there are locks and things like that um all of which is much more similar to monkey island than it is to you know choices or episodes on the on the app store but in reality i don't really think it's a visual novel or an adventure game and for me it's much closer to a 1980s text adventure game uh like these two which is not surprising because i grew up playing 1980s text adventure games and i grow into my position as a designer by writing text adventure games um no one really remembers them very well anymore some people do but not all that many um but they were routinely surprisingly complicated so these two that i've got on screen at the moment deadline from 1982 when i released when i was one and suspect released a couple of years later both of them have multi-location environments that you can wander around freely they have an in-game clock which affects what state those locations are in when you get to them uh they obviously going to different places at different times yields different results and that's a big part of the game play

they also have casts of characters really quite large cast of characters um suspect is unreasonably difficult it has so many characters in it and these are characters with memories and knowledge who react to the things they discover during the course of a game play and they move around independently they don't stand in one place waiting to be spoken to like a bevester guard they move around and they do things and you can see them doing things if you hide and things like that um and generally speaking the idea of the game of both of these games is not to proceed through an adventure through a linear story from beginning to end proceed through a script like you might do if you're playing grim fandango or something but rather to find information from the location from the environment or from one character and then use that information to cajole another character into telling you more and you wouldn't really play the game from beginning to end much more you'd play them over and over again until you had a picture of the world and the story and whatever the secrets of the of the

locations were and then you would use those to solve the game so these games are definitely not visual novels obviously they're not really adventure games even by the standards of text adventures from the 80s so i don't know what to call them but the most obvious label seems to be the simplest one which is they are detective gains now there's no agreed definition that i can find of what makes a detective game into a detective game there's a category on steam and gog but it tends to cover a pretty wide range of experiences and is often more about setting and the tone than the actual mechanics so if we want a mechanics-based concept of a detective game i would suggest it's something like this it's a kind of short-term game loop in which there's some kind of core discovery which feeds deduction which feeds the discovery and we go round around like that so to make that a bit more specific so i can say exactly what i'm talking about uh firstly we go out into the game and we find a thing great we found a thing secondly we think about the thing and then thirdly because the game can't

tell what we're thinking we have to prove that we thought about the thing so for example in overboard which i think is a detective game as well by this metric in the morning on your very first playthrough of the game the player is strongly encouraged to go to breakfast and act normal because they're trying to pretend like they don't know that their husband has been murdered that's not a very standard um detective game prompt but most players will do it because it's the tutorial and they want to see what the game is supposed to be about and if they do go to the restaurant and have breakfast they'll bump into another character who will engage in an idol conversation and will remark that you are only wearing one earring that your protagonist is only wearing one earring this is a thing it's a clue we've discovered a clue um but it doesn't tell us what to do then we have to think about it and the game prompts you quite heavily because this is the opening but still our deduction is it might be on the upper deck which is where the murder

happened right we might have lost this hearing during the murder and our action to prove that we thought about it is to go to the upper deck and check it's very straightforward it's very simple but it fits this loop of finding a thing thinking about the thing and then demonstrating to the game somehow that you've done it so there is this hand over hand process between the game and the player with each one doing a little bit of work on their turn to push the game forward to push the story forward so the game will give us a piece of explicit detailed information not your detective quality is increased by one and not you've gained seven police points but rather there was someone in the library at 815 with a knife something really specific and concrete and then we the player have to think about that and express back to the game i understand i understand what you're trying to tell me so the game unlocks some information and then the player in return takes some kind of action so the part where the player shows the

game that they've understood uh that's obviously the most interesting part for designers right um that's essentially basically takes the form of a combination lock right uh it requires the player to take some action that isn't explicitly detailed to them um which acts like a combination lock to express to the game that they understood what's going on so that might be simply go to this location as soon as possible or go to this room between the hours of 10 and 11 that kind of thing it's not difficult necessarily but there isn't a button for it and there isn't a quest log telling you to do it and there isn't a ui marker telling you to do it and while it is hopefully fairly obviously the correct answer the correct action to take what the game actually does is give you a map screen and say prove what you know and then as the game continues those demands of proof become stronger and stricter and tighter and more specific assuming that the game gets harder as it goes now i don't know if you're going to like that definition of what a detective game is but i'm going to run with it for now but i think it's worth noting that a few

games that are commonly thought of as detective games don't fit so i don't think the witcher 3 fits this definition particularly it tells a detective-like story at many occasions but they're very linear they're very guided and the player is never sought to prove that they know what's going on they're more asked to react to what they think about what has happened i'm not sure this definition covers a game like l.a noir either because if i remember that game correctly again you're always doing what you're told and while it does branch it branches more on your ability to punch people and drive your car than it is on whether or not you understood the point of the crime however much their interrogation mechanic hopes to be that so i think what i want to be clear about is i'm really trying to cover games that require the player to do detective work not necessarily games about detectives which is obviously to say that we're really talking about games which have detective like verbs right because the interesting bit is the

expression part it's the part where the player tells the game yeah i get it i know what's going on give me more um and detective games live or die um on the basis right that's fairly obvious i mean if the player has an idea about the game but they can't express it because they haven't got the verbs for it then the game cannot be played and equally if the player has absolutely no clue what's going on in this story whatsoever but is able to express the correct next step anyway then the game is linear or led by the nose or throw away i don't know it might still be fun and engaging but the player doesn't actually have to pay any attention in order to complete it and i think if you've played any detective games we don't need to break anybody over the coals um we've definitely all played examples that fit into both of those extremes that's fair enough it's a hard problem um the slightly more approach then is to start thinking about those detail obviously the more expressive verbs we allow the more

interesting deductions the game can allow um fair enough okay no look i didn't fly halfway around the world to tell a bunch of narrative designers to think about their verbs i mean i quite literally did not do that but i do think it's worth reviewing some of the really excellent detective games that we all like that we've all played and everybody knows and just have a little look at their verbs before we proceed so i picked on three all of which have won bajillions of awards and everybody likes them so don't accuse me of criticizing anybody because we're not here to do that um so if you look at her story classic detective game probably the best detective game we have in the detective ubra ubra ubra isn't that anyway um but the verb is very strange it's this database search mechanic right that brings up videos and it doesn't necessarily transplant very well outside of her story itself so you can't really imagine putting that into an agatha christie story i mean you could perhaps make an epistol one of those letter writing stories anyway where you find a letter by typing

in a word from letters but i'm not really sure that's a good idea necessarily and anyway sambalo likes video too much so it will never happen um obra din is obviously a very excellent detective game but it's combination of verbs being a magic self-writing book and a time rewind moment of death magic pocket watch is so frankly bizarre it's amazing that he got away with it in the first place um and it's unlikely to spawn an entire genre of detective games that follow the same model which is kind of odd really because mysteries are like one of the biggest genres that there are in fiction and there were a billion episodes of colombo and a billion episodes of quincy there's definitely something repeatable in the concept of a mystery but uh obradin's particular take on it i think is specific to obradin i may be wrong but it seems that way to me and then finally um outer wilds which i would argue by our definition is a detective game is probably actually the best example that we've got of a game that doesn't use strange verbs to achieve

its um its combination locks right it relies on location and timing for almost all of them and then mixes in a lot of the elements of its verb set but it's verb set is fairly ordinary it's not that unusual you know it's light and jumping and thrusters and things like that it doesn't have a gun which makes it quite unusual in some ways as a computer game but less so than that used to be but equally while it's extremely successful at its detective deduction loop it is the game in this list that moves us the furthest from a classic notion of what a detective game is supposed to be philip marlow never did nail a 30s pornographer by jumping into a tornado at the right moment so i want to be really clear before um before people go into tribal brain because i'm talking about some games that you might like i'm not criticizing anything this is not a criticism um they all do they are all excellent and they do precisely what they set out to do that's what makes them um superb games

but the point is if these games represent our model of not only what a detective game is but what it can be we have to ask the question have we missed anything is there anything which ought to be included that somehow isn't included in fact is there a large chunk of the art of mystery storytelling which hasn't been addressed as far as i can tell in these titles at all well i think maybe there is because so far we've only really been talking about the short term game loops so clue discovery clue discovery clue discovery and slowly over the course of that uh we piece together the back story of the game right that's how her story works is you assemble your understanding of the narrative that's how else the world works is you assemble your understanding of the narrative because these games have a secret history which you the detective are uncovering as you play and piecing together um but every time you encounter a bit of that narrative it's generally a piece of truth and you put

it together into your truth jigsaw and by the end of it you have um the complete truth of the game and you can understand the story and you can stop unless it's her story in which case you don't necessarily understand the story anyway but that's okay it fits the genre um i mean we've all heard obra din described as narrative sudoku okay but the thing about sudoku is that when you're doing a sudoku there's that end bit which is this is a picture of the end bit of a sudoku where you're really just putting numbers in you're not actually doing anything and you totally could just stop at that point because like in any meaningful sense the sudoku puzzle is over it's it's finished it's been done it's complete already and the only reason that we don't leave a sudoku puzzle in this state is because it's kind of nicer when it's been tidied up um but that's not really how narratives are supposed to end we're not supposed to end narrative experiences with a process of tidying up particularly um narratives are supposed to surprise and delight um detective stories are

supposed to have a big twist at the end a big moment of revelation that clears away the confusion that came before a moment where the light is suddenly shone i mean you've never finished a sudoku puzzle and finally put in that last number and thought oh it was a five all along of course i knew it that changes everything sudoku will never generate that experience it cannot generate that experience because it's fundamentally built not like that right so so what am i talking about i'm talking about the solve we need the salt perhaps we need the soul anyway we seem to be missing the soul the master solution to the complex crime the final pulling together of disparate elements to create something unexpected but ultimately completely sensible um hercule poirot does not solve a succession of small mysteries over the course of his one and a half hour tv movie from the 80s or his 200 page novel with a hand the full narrative to him that's not how he works his process

which is a lot more dramatic is that he brings together all the confusing and separate elements of the crime in a sunburst of solution that wows everybody that flaws everybody in the room but as soon as he does it we go ah we should have known we should have known so that's what i'm interested in that's what i want to talk about today is how do we move away from this idea of the basic deductive game loop of the detective game to something that allows us these moments of blinding revelation that can climax a story for us can it can it be done can it be done and of course people have tried to do this right it's an obvious part of the murder mystery genre it's classic it's very very classic um and one of the mechanics that's often used is the the fact linking deduction mechanic right um the fact one plus factor gives us exciting new fact 3. um now i want to talk about this mechanic specifically for a while which

perhaps seems a bit unfair because it's just one mechanic amongst many but i think it's got some really interesting flaws that that are really helpful to think about sort of quite clearly um the first thing i want to say is that it's been tried by a lot of games there's a lot of games that use this kind of mechanic some of them are more successful than others um and it's a pretty natural design to try right logical deduction is a plus b implies c so that's pretty reasonable to think that this will work very well um and moreover it comes directly from the detective game core loop we have a fact we take a combination lock style action of applying these two things that don't immediately appear connected together demonstrating our understanding and the game unlocks something for us so it's a natural part of that game loop but as i say i think that the idea has some flaws which expose some flaws in our wider thinking about detective games

perhaps firstly some fairly simple things you actually have to have facts one and fact two i don't just mean you have to have played the game enough to find them but they also have to be things that you can pick up and put in your ui screen so in a good detective story the detective is frequently making deductions based on things like their understanding of human nature or the unspoken rules of social etiquette and not just the cigar but found in the fireplace it's a mixture of those two often a detective will go looking for additional weird information on the basis of some kind of hunch they've had which gets accumulated into their facts pile in christie's the mysterious affair his styles hastings by asking by saying the most significant facts in the crime is that the temperature was 80 degrees in the shade the night before hastings is completely perplexed but it does improve indeed to be the clincher of the crime it's obvious in retrospect the problem then is not one of finding the answer it's really one of asking the right

question that if you're doing a deduction linking game your facts have to be up there in a way the player can access directly you have to have the temperature on the night of april the 15th written in a little box on the screen you have to have i don't know ladies button their jackets right over left and when you start to do that it's a bit like when my son is eating an apple in the park and decides he wants to play i spy for something beginning with an a secondly the conclusion might not be that exciting um it's quite common um to be playing one of these stories and be a little bit ahead of the narrative and think oh well obviously this and then wonder how you're supposed to tell the game this thing you already know what weird and confusing process do i have to go through to explain something that's obvious to me it's a very very hard problem to avoid because of course people are differently versed in detective stories some things jump out some things don't it's deeply annoying both for the designer and the player it's it's it's a real problem because of this combination lock gate that the

player has to get through but then again what if the conclusion isn't obvious what if it isn't immediately apparent then suddenly the fact that you've got all the facts of the case laid out in front of you ready to be quickly linked together scales up the possibilities in a way that isn't necessarily very fun if there's no curation then it can get very quickly overwhelming even if the answer that you're looking for is completely fair and completely logical and even if you understand the plot um that's a really crucial point i think actually i wanted to give you an example which is this this is from a book called where's the pair by british tech and drop and it's a book that my children have it's a book for toddlers and it's surprisingly difficult it's full of pages like this which have a big picture and you're supposed to spot the two matching fish now if i could see you properly you're a very tiny window on my computer screen um i would ask you to raise your hands when you see the pair um as it is i'm gonna wait for a bit and i'm gonna show you the answer but uh perhaps you saw it

immediately perhaps you already know exactly what it is um or perhaps you're still looking uh but i think we can agree on a couple of things we can agree that that's a problem it's entirely fair and the answer really should be completely totally obvious i think the answer is the two fish the other way this should jump straight out to us but i don't think it does i find these pages of this book extremely difficult i don't know whether you did or not and i think that's because cross-referencing is hard even if the rules for your cross-referencing are actually very simple if the rules for what constitute a match are actually a little bit obscure or opaque then you end up with something that the player is going to attack by brute force out of sheer desperation and we can't really fault them for that the overwhelm problem is real now i don't want to make it seem like i'm sitting here in my in my armchair literally criticizing a design that i've never tried that would be a bit unfair so i would like to add as well that at

inkle we built around 20 or 30 deduction based prototypes in the course of like the last two or three years and found well we found even more problems and here are the other problems we've found firstly it's cluttered there's an enormous amount of information on screen but relative to it there's not actually that much for the player to do it's a bit like the fish all had name labels and blurbs and a tooltip and you had to read all of them before you could destroy sometimes you have to read them twice or three times and check back at them just to see what they said because you can't remember because there's so much text all the clip to matching things um the clutter gets worse as the game goes on rather than better because every time you make a deduction it makes another bloody fact and puts it in your fact list unless the new facts somehow wipe out the old ones but what if the old facts have multiple consequences truth isn't tidy things don't imply one thing they imply lots of things in fact truth is a real a real gift because it's inherently broad not narrow i mean the thing about truth is that it's true

there is never only one way to arrive at a truth or to deduce any fact um because truth abides if you like um so while you can make a deduction game which allows you i suppose to arrive at truth in multiple ways so that would get around that problem but then you're putting even more facts on screen some of which the player isn't going to use or he's going to wonder why it's there because they got it over here in this other obvious way and lastly and this is very significant to me it's hard to write i mean where does one fact end and another one begin is the fact elsa is lying different from the fact elsa is lying to protect ana or is one a deduction from the other or is ana knows elsa is lying different from elsa is lying or from elsa is covering that honor or how do we know how best to chop our facts into tiny little atomic pieces we're sort of guessing but it gets very complex and muddy quite quickly to all of this poirot says no

and i agree with him it is too much so i think we need to look at the big solve concept from scratch i think we need to take a step back and try and and think from the beginning what is it we actually want to achieve in this climactic scene if we're going to really enjoy ourselves with some detective solving loveliness so goals great goals what are our goals right i want to be able to combine facts from the game i want clues i want discoveries and i want to be able to combine them with ideas from the real world with things that i know or intuitively believe um i want a curated input system so that i'm not overwhelmed at any point i want just a few fish not quite so many fish thank you very much i want to have some kind of input that makes it easy to express ideas i don't want it to be too abstract or weird i don't want to be drawing funny diagrams to try and explain something which doesn't naturally fit that particular presentation model um and finally i want it to be balanced i guess i want it to be you know feel like it's it's got a space for play

in it so that it doesn't highlight the correct answers but it doesn't hide them away either so what we really want i think is we want to turn the detective game mechanic entirely upside down and instead of hunting through a whole possibility space to find the golden path of truth the one true meaning that will lead us to the end of the game what i want to be doing is i want to be building up a truth a few pieces at a time ideally with each piece going in being examined and tested for its suitability and considered before the next one is applied i want to move away from the paradigm of the um combination lock towards something that's a lot more like solving a maze so in a maze you have dead ends but you can walk back from them there doesn't need to be a single path through a mage you can do it quickly or slowly it doesn't actually matter there might be multiple exit points in this

maze you're just forming a path there are rules you're not allowed to jump over the walls but there's not many more rules than that in short you're constructing an argument and there is no one right argument for anything um even something which is entirely self-evident like that the russian military action in ukraine is a crime against humanity there are multiple ways to prove that idea there are multiple approaches to make that convincing there is no one way because it's truth and truth can be approached from any angle you like so all we really want to be able to do is make a case that's compelling and it's watertight and it covers the available facts we want to step away from this idea that the player is the master detective in control of everything totally in charge because that's not actually that fun holmes is not really having that much fun when he just knows the answers it's a bit like sitting down to do your sudoku and being able to see all the numbers in the grid before you've actually done any thinking it's pointless at that stage i was glad earlier when my slide suggested that sudoku is pointless obviously it's the process of deduction that's fun but holmes doesn't get to do

that really because he just knows because otherwise he's not magical homes so really we want the player to be watson rather than homes feeding ideas to his friend uh which his friend will accept or refute doesn't necessarily mean that you need multiple characters one protagonist could be going through it in their own minds poiro is constantly muttering to himself so perhaps he's playing the game that we haven't yet made um and also i think if you think about the design in this way you start to see opportunities for balancing a combination lock cannot be balanced you either know the answer or you don't it's like a quiz it's inherently not fun um but an argument that's built by playing a game can't be balanced there can be multiple routes the player can redo sections the player can be encouraged to improve a semi-coherent idea into a fully coherent idea um if you are talking to another person how helpful are they do they suggest alternatives do they point out contradictions early or late or specifically or generally there's lots of wiggle room for creating

a space which is fun of course the problem with all of this it's very nice it sounds great the problem is uh verbs again we need to somehow present the player with a method of actually doing this which means ui which is extremely annoying to build and no one at inca likes building ui i like to think we're good at it but i'm not sure that we necessarily enjoy it um unless unless for overboard we weren't thinking about any of this we were just trying to make the game as quickly as possible and we wanted to replicate the agatha christie structure um as best as we could as closely as we could so we knew we wanted a few hours of in-game exploring and talking to people and mucking about followed by the detective gathering everyone together in the room and going through the crime step-by-step because that's what she does so that's what we wanted to do but our insight and we had it sort of halfway through building the game was that the detective in the room scene that christy likes so much is the process by which the author

explains the story to the reader so we thought well if an author can explain their deductions to the reader via one of these climactic scenes maybe that can also be how the player explains their deduction to the game so let's steal it so we stole it which led us to implement our big solve scene using the same buttons that we had in the game anyway basically conversation uh we've tried this before actually i should say on a smaller scale there's a murder mystery in 80 days if you've played that game um but it's a particularly ruthless implementation there it's got a very narrow golden path if you've got the right idea there are traps on all the other options if you haven't and if you've got half a picture of events then it's kind of luck as to whether you pick the right path through the conversation or you pick the wrong one and of course you need to have collected the right information over the course of the story up to the this point um to even have the correct answers available so it's extremely easy

for this sequence to feel extremely unreasonable even though there are a few multiple routes through the conversation so for overboard we needed to do better than that i mean the whole game rests on the final conclusion scene feeling intelligent and responsive and above all fair right if it doesn't feel fair the game is impossible to play and because overboard is a game about being a murderer not catching one the protagonist spends most of her game time majority of the game either destroying evidence or falsifying it or entirely murdering other people so the golden path model doesn't really work for the conclusion anyway because it might rely on a character saying something who's actually at the bottom of the ocean or having a piece of evidence which you can't have or the non-existence of a completely fake piece of evidence that veronica has managed to bring into existence so that presented us with quite a challenge how do we have this big solved moment that we want um given that we have a very very flexible input state to that final scene

what we really needed to do was create an ai system that would deliver a conclusion based on evidence and the people available to it in a logical order presenting it rationally to the player while giving lots of change for the player to kick and affect that storyline as it was being developed so effectively we're now saying on our super fast projects that we want to get shipped this week please uh we want to build an ai hercule puero who will ignite all the evidence and formulated conclusion okay so we gave it a try let's break down overboard a little bit more what actually happens in overboard are two phases right there's a high agency low impact phase where the player is making evidence and lying to people and setting up the parameters of the accusation scene they get a lot of freedom they can do a lot of things but the individual actions don't have very many consequences there and then then at some point the hammer falls the accusation begins at that point the player then can argue and lie to tilt the accusation scene but they don't have much agency anymore all

the initial parameters have been set they're now just trying to adjust things and tweak things but those tweaks might have a very high impact because everything matters now if you can just nudge a conversation away from being something incriminating then that's a that's a clear win that makes a difference it's not a delayed consequence it's a consequence now and then finally an ending is created and one of the good things is because we're not interested at all in creating a puzzle or a single solution or a combination lock we just want a coherent conclusion there isn't even a proper murder story to be uncovered so that's kind of nice we mostly want to make sure that what the player is doing makes sense it has to be interesting and the game has to respond intelligently and ideally we reward creative ideas and we punish banal ideas but we don't really need the game to be particularly clever or ingenious so long as it's fun it doesn't have to be hard we don't have to make sure that the player can't accidentally brute force it or find a solution by accident we don't

want them to but we don't have to go too hard on that because it's such a wide open parameter space so the whole game rests on the implementation of this accusation scene of this constructed argument um so we approached that by building it in the most straightforward way that we possibly could mainly so that it could grow barnacles and sidetracks and sub content as we playtested it and developed the game we wanted a simple core so that it could get more complicated later so we approached that in a top-down way this kind of inspired by the weave syntax of ink so um it's a it's i don't know if it's unique to ink anymore i think yarn spinner has something similar but the idea is it starts at the top and it drops through content and it always reaches the bottom and there might be choices and branching but it's guaranteed to hit that final beat the idea there is we want to make sure that whatever happens we definitely reach a conclusion um that right there are two ways that an interactive story can fail right one of

them is it runs out of things to say and the other is it doesn't make sense and of those the first one is a critical fail because then the game can't be played anymore and the second one isn't a critical fail because it means some players will have a less good experience than other players and that's normal in games so we use this drop down syntax that always hits the end and then we work on making it and optimizing it so that it's as good as it can be for every playthrough that we can generate and we can make we tune it essentially so into this top down process we basically get the characters to talk one topic at a time the topics are introduced in priority order they are whatever we say they are a topic has no formal definition so it could be a broad concept like malcolm's gone missing your husband's gone missing or it could be a specific detail like veronica's earring was found on the upper deck at the crime scene we deal with really important uh topics first uh things like uh you somebody saw you killing someone well okay that's a

straight game over we want as little redundancy as possible in each play through we don't want people going through an enormous scene you know only to find out they lost straight away um fine that's straightforward so in a given topic uh what we do is a very similar pattern for each one it's introduced by some character that's not always the detective character it might just be another character who discovered or learned something during the course of the day it's then discussed by the group which includes veronica the player so you can affect that discussion you can be very significant in that discussion you can be called out for lying you can be not called out for trying to adjust the truth you can be the person everybody listens to depending on your relationships with people um the conversation is pretty free form there's no need for hard rules here but whatever happens the drop-down idea again we will eventually reach a conclusion and some kind of outcome that's accepted by the whole group it's it's considered now to be definitively true that this topic causes

this conclusion here and that conclusion will accumulate against somebody it'll be it'll be some kind of evidence towards some kind of overall conclusion to the story so for example because that's quite nebulous uh the topic of malcolm is missing might result in some conversation where veronica says oh he was with me last night but the steward says that the bed wasn't slept in so that means everybody concludes that veronica is lying and that's a black mark against veronica it goes in her ledger and it's it's one one point against her so that's we do that for each topic we kind of write though in compartmentalized ways this topic this topic this topic if we introduce a new piece of evidence or a new storyline we can just write a topic for it for the conclusion bung it in the conclusion um and then we stitch the pieces together topics are introduced whenever it's sensible for them to be introduced there is a spine but if it makes sense you saw this well i saw something like that then they'll jump in and say that

so we can shuffle the topics around um if we want to if it makes better sense there's a tuning the story to be as good and as tight as we can make it um the uh there is a tiny bit of ink just to make sure that a topic isn't discussed more than once obviously but that's really a minor detail topic discussions work really nicely i think for the replay of overboard because they teach the player what was relevant on your playthrough they pick off details one by one and say you might want to think about the earring you might want to think about the sleepy graph you might want to think about this so that's really good but another thing that comes out of it is that the system doesn't need to be rigid and an outcome of one topic can always if it wants to recontextualize a previous one so um you know if we start talking about clarissa and malcolm having an affair then veronica might say oh i knew all about it and i felt terribly ashamed i couldn't believe it what am i supposed to do and the outcome of that if you pull it off might be that oh veronica

was justified in her lying so we might take that previous outcome which was a mark against her and turn it into a mark against clarissa by virtue of having this topic appear after the previous topic so the order of topics can be significant if we want it to be and the way the player plays them can be significant if we want it to be finally when we finally run out of topics to consider or the evidence has got so overwhelmingly that you're definitely definitely guilty that the detective just says yeah okay fine we're done um then we pass it to our detective characters um an npc who weighs up the outcome and at this point the player is no longer free to affect anything they percent set out and now they're going so you know this is fairly straightforward we don't do anything particularly complicated because it has to work right

so if veronica has three guilty outcomes against her clarissa only has one our detective character will assert that veronica is guilty um which leads us to an ending where she's arrested the ending itself is window dress with lots of content based on how the scene went so while internally major singh is simply saying veronica has two more guilty points than clarissa does what he actually says might be well it seems to me that you were definitely on the upper deck and we all know that you hated malcolm because he was having an affair and we all know about the life insurance therefore you're clearly the criminal and that conclusion that puerto style conclusion is is driven by whatever you set up in the game so hopefully the way that plays out is people bring in evidence you try to struggle against it you fail you fail clarissa gets a little thing you think you're going to get an advantage there but that goes against you too and at the end the collective slams it and says no you're definitely guilty enough to go to prison and it feels dramatic and fluid um and at the beginning of the scene we don't know how it's going to go and in

the middle of the scene we might start to get a sense of how it's going to go and by the end of the scene it's completely fair enough how it went so hopefully we get a dramatic scene which is entirely driven by the player's high agency actions during the main game loop but that feels deductive and detectivish and full of content and context and with lots of opportunity for us to chip in with conversational ideas or theories or counter theories or comments about what clarissa might be like as a person or whatever it is you think that might get you off the hook and they will sometimes work if they make sense so because it's not a formalized abstract system we can muck about with it basically and put in stuff that we like similarly uh we might play a game in which we're so good at this that veronica has no guilty outcomes and clarissa has three guilty outcomes in which case the detective will spin round and accuse clarissa of the crime and nothing she says at that point will save her and that's fine and you can gloat all you like um and again his dialogue will explain how the particular bits of evidence that were gathered during this playthrough show the incontrovertible truth that

clarissa is definitely guilty and of course that's dynamically constructed but that's just a text generation problem there's no actual design implications there so we can black box that and again just make it as good as we want and because overboard is a game about throwing someone off a boat we have a nice full-back position you always need a full-back position right you need a way of dealing with sort of meh outcomes where the player hasn't really done anything very exciting and in our case it's well if veronica's got one guilty outcome and clarissa's got one guilty outcome and really there's nothing very exciting or dramatic going on um then it's probably a suicide and we have a narrative reason why that's a not very satisfying ending to the story which does the job that we need it to do which is to say to the player go away and try harder so that's um cool so the basic principle of all of this is a bit like the holmes and watson model that i mentioned right even though it's a game about falsifying a conviction

about faking evidence and constructing a crime you are essentially constructing an argument you're just doing the evidence first as well um but the crucial thing is that it doesn't cast the player in the role of the master detective or kind of the phrase master detective is sort of polluted by sherlock holmes really because the point is the player is the detective the player does go out into the world asks questions discovers information brings it to the attention of other people goes to the right place at the right time and solves the various combination locks that the game has to offer but the detective doesn't get to be the judge the detective is not there to assess the quality of the conclusion somebody else does that and by the time they come to do that the player has set up their initial conditions and is just trying to push an argument through to construct an argument that the judge will be happy with the player is feeding inputs into a system to get the outcome that they want and so it's not a combination lock because it doesn't have an input that

you are trying to hit it's a bit more like giving a set of directions to a little rat before it runs into a maze and seeing which door it comes out of the other side there are lots of solutions in fact we don't know how many solutions overboard has because why would we care about that information at all it's just so long as every solution that we've seen makes sense and feels fair then the game works fine and that's good enough for us so um that's the sort of broad idea of how we approached it for overboard and while we were making it it was mostly just a quick christie on march but the more i look back on it the more i wonder about how generalizable this is as an idea for creating a bit of drama in the story of the detective game because i think a point that gets lost a lot when we're thinking about detective stories is the thing that christy was so confident about in her writing that there are two stories in a detected story not one but two um one of them is the story of the crime that what happened and why

and then the other one is the story of the detective and the characters affected by the murder dealing with the aftermath of the murder or whatever's happened and solving it and that is a valid story and it's an important one too chrissy novels always have a love affair in them you can always tell people who are not going to be guilty because there is always a young couple who fall in love at the end that's just standard logic for a story told in that time um actually conan doyle does this too sherlock holmes stories are structured much less elegantly there's like 80 of it is sherlock and watson running around having a fun time on trains and hitting people with sticks and then the last 10 is the guilty party usually sitting down with homes over a nice cup of tea in a pipe and telling him some story about a thing that happened in the colonies which is why he ended up sending orange pips to this person or murdering this person or writing stick figures on a well the two stories are completely separated in conan doyle's writing but they're still both there there's the adventure we enjoy and then there's the weird colonial rant that conan doyle would like to go on now um and they're put side by side christy layers them together and then brings them together

in a conclusion scene with poirot doing that and the reason that's important is it gives us depth to our world because we have a secret history to be uncovered but it also gives us just a tub thumping good adventure we get some characters to be with we get a protagonist moving through a world exploring and having an adventure and being a real person in a room full of people in a dramatic intense situation which is generally speaking why mystery stories are enjoyable it's if you think about the mysteries that you watch on tv especially the ones made by abc studios you've got to remember that the backstory part is never as interesting as the story part we're rooting for the detective we're rooting for the characters and oh yeah there's this mystery and we kind of want to know what happened but it's rarely the focus of of our engagement so i think it's pretty cool i guess that's what i'm saying i think it's pretty cool um a few other notes on the design uh firstly it feels quite natural like writing it was not a big theoretical complicated exercise there's no weird ui to be assimilated we don't have to

really tutorialize anything because it's just people talking to each other the player gathers some material they present it it's assessed and if it's not good enough they get sent back to find some more until they have something which is convincing um it feels to me weirdly like a really familiar design except i can't actually think of anywhere i've seen it before except possibly in the two infocomm games that i mentioned at the start the talk there are no new ideas in game design right um the second thing i want to know is that overboard is very definitely replay to win it's structured to put this conversation at the climax and if you go wrong you get thrown into jail or throw yourself off a boat and you try again with different starting parameters until you get it right but it doesn't have to be that way there's nothing about this design which is inherently a replay to learn game you could for example play the detective who phones up the police chief to explain what they've learned so far and lays out the cases they see it and the police chief says no i'm not sure about that bit you need more evidence for this we can't yet arrest because we haven't got enough data on that and that go away

and get any more and then you come back and you get some more in that case a fail state isn't being sent to prison it's just let's start from the top or let's look around some more call me back in an hour and it doesn't have to be dialogue like i did it through dialogue because i like dialogue but um and it's fairly natural way to present this idea but um for people who like ui and they like connecting boxes and filling things in you could imagine a pro forma where you say so i don't know blank went here and did blank and then you drag in the blank and you drag in the next blank and then that unravels into a next bit of conversation and the story unfolds depending on exactly how you feed it right um i have a suspicion this is what failed better's mask of the rose deduction system is going to be like but i haven't actually played the demo enough to be sure about that so we'll see um [Music] such a system if you built it that way would still be curated it's still filtering the options down to make sure that i don't get overloaded it's still building an argument organically from

the ground up it's perhaps a little bit more constrained in its structure but it's not very constrained it's certainly not so rigid that it has to be fact one plus factor equals factory or nothing um and it is capable of expressing different kinds of ideas or bringing in ideas from the real world or bringing in new ideas there's no reason you can't drag in ladies button their coats left over right if it seems like a relevant thing to drag in and it will be a surprising moment if perhaps a fairly obvious one in the context of that that instance of the game finally the correct path can be very broad which i think ultimately we have to agree that's probably a good thing because frustration in detective games is a real problem and truth is a big broad concept there shouldn't really be a single solution path it doesn't make a lot of sense and every time i try to write one it just goes wrong because it's ludicrous right um so we're effectively talking about something which is a kind of simulation though it depends how you build it exactly but a simulation of an argument not a branching tree narrative in which you die on these instances and you don't die on that instance

so to summarize all of that to put all of that together i think it's like i said earlier i want to get away from this idea that the combination lock is the core model of the the of what the detective does that the detective is constantly cracking the correct answer from a set of impossible answers away from sherlock holmes and his um when you've removed everything that's impossible then whatever remains however improbable must be the truth which you know fine yes that's true but that's not the most interesting way to solve the detective narrative it's not the most sparkling conclusion to a narrative it is the five in the middle of the sudoku which is a perfectly good solution it'll do it sort of says well when you've tried everything you might as well go with this one it's a bit like i don't know yeah it's not it's not very it doesn't fill me with the excitement really the great thing about the end of a puerto novel is that when pyro turns around and accuses someone it's always the person who obviously was from the beginning and para's argument is usually i knew that you did it you're the only person in this room capable of murder but then i

had to figure out how and why and there's something really nice about that something really solid and charming about that i think so let's move away from deduction deduction means taking things away let's stop doing that let's move towards construction towards building and growing and developing a narrative that we can be part of and we can be excited about that's what i think we should be doing for detective games thank you for listening i will attempt to take questions over the magical internet but otherwise i'm done uh hi there thank you that was spectacular um i'm wondering as a person who makes visual novels and adores detective games why did you choose to go with the the set dressing of a visual novel and um would you have chosen a different kind of format um if you were going to make it again

so um i have a kind of honest answer which i feel like i ought to say which is um we made it the way we did because we really did try to make this game as fast as possible because we were supposed to be working on our highland game and like we kind of did this for fun so the format we took was because it was very easy to implement and to understand from a design point of view we saw like the classic visual novel there we were like yup we can definitely build that let's build that um and that is the honest reason why we did it why i think it works because i think it does work um and it whether that's luck or or not i believe it's an exercise to the reader um but the reason that it works is that it focuses the game entirely on people and nothing else and i suppose that's why it works in visual novels as well actually that if you were to build it as a point-and-click adventure then we get a lot of distraction in terms of the environment or just the business of walking from over here to over there and and all of that kind of stagey mechanics

that that come with adventure games are very very hard to solve and that actually detract from the naturalness of the people in your story world right if i'm wandering around a room aimlessly looking at things and this character sitting in the middle paying no attention to what i'm doing they look more robotic whereas in overboard we simply put an abstract image of you image of them we can imply any amount of movement around the room but it never breaks visually so our feeling for why it works is it's just it gets to the point um of what the thing is supposed to be and i suspect that's exactly why it works really well for visual novels actually it's like it's attractive art style you can really see what the characters are thinking you can see their faces and you're listening to their talk and nothing else is relevant so that's why i think if we were to build another one apart from the sheer fact that we have all the code now so it'll be really nice to just build it the same way but i think i really like it actually and you know i don't think it's a game where traversal is massively

important so i think you know if i was to look at ways of developing the art style i would push the bespoke art so you know i would have her hiding behind a sofa and we would see that as a specific shot from a different angle and we would have more cutaways of like interesting items or particular expressions or particular moments in the game because i think once you have a we often talk about um inkle about a sensible default and overrides being the best way to get coverage on a game especially a wide and branching one so you have a sensible default of characters room it's sorted you can you can make the whole game that way and then you can spend all of your fun time doing overrides for all the exciting moments like there's a bit in overall where she hides behind a pot plant and we have a picture of her skulking out from behind a pot plant and i love it and it's not important at all but it's it's just that's where our time and money should be spent i think um so yeah uh yeah no the visual novels were were really helpful as a way of just leapfrogging that whole design process and just getting to a place we could just get stuck in straight away so that was really great

thank you hello um i was wondering if while doing this sort of like uh showdown with the detective war and really any sort of um sort of the endgame part of that if you have to consider uh weighing certain victories or losses in the sort of uh topics that come up more and also have to adjust their values depending on the like uh context possibly changing on a future uh response for so for us so for example let's say you like get caught in a lie that would be say a point against you then you if you recontextualize that you'd have to um maybe neutralize that or maybe give it a plus one um do you sometimes worry that in the end you might come up with this answer that feels not necessarily unsatisfactory but very strange like say the text says well there are these three things that prove that you're uh innocent um but there's

this one very strange or bad thing that you did that we're gonna have to put you in an or or even weirder where it's like they say like this long tangent on why you did it and then there's just one final bit that meant when they get but you couldn't have done it because x y or z and that kind of feels very strange yeah so i think um what we found was that we we built that scene and we play tested a lot basically i had a random a random player and it did a bunch of things for me but one of the things it did was set up a random initial parameter space like at the start of that scene it just writes down what the player's done it's a very large list of booleans and then it set them randomly and it played the scene to see what would happen and the great thing about we're writing this in ink which we're pretty adept with now but it's generally quite a flexible scripting language so you'd get to a point where a conversation would jump a track like that like the the character would suddenly spin on a six fence and say you're so guilty you're so guilty anna you're fine actually and you when you play through you you sit you see that and you go oh wait

no no no wait no no no in the case where the player is just ridiculously guilty and he's trying this cheesy cop out don't let it work and it takes you know only a couple of minutes to write that kind of thing and it feels beautifully rewarding when you hit it as a player because it's so specific to what happened to you and there must be a thousand of those in the conclusion of overboard and because the tools make it really fast for us to just muck about with the specific cases and because the flow itself starts here goes there and doesn't go crazy in the flow graph or anything like that it's actually very stable to do you just say right you're here we're giving you this mark putting you there the game carries on so lots and lots and lots of play testing and kind of working and working and working on it to try and make it as seamless as possible and you know our beta testers and our qa people and people who played the pre-release version were invaluable for that as well because they would play and say oh i didn't like this ending this one wasn't good um so i think that to me feels like part of the crafting of the scene and it's such

an important scene we could afford to spend a lot of our time doing that um so yeah definitely a problem but one that you can solve without getting too systemic about it i think thank you uh thank you so much for your talk this is really great um you discuss the fact mechanic and detective games having to be represented by ui which then like precludes things like intuition or knowledge of human nature uh did you have any feelings about how disco elysium handled that with the thought cabinet where they had their detective go out and get ideas about human nature or about intuition is like that was their inventory was their ideas then using that did that kind of address that problem or did that feel satisfying to you as someone that makes detective games um i felt like they were doing something different um that kind of fitted with their aesthetic and their style and their narrative i'm not sure whether i think disco elysium fits

this model of a detective game or not off the top of my head it's it's it's a complicated and odd game in many ways and i'm not sure whether it has a clear closed deduction loop exactly um so i don't know i also think that it's a solution it it's quite abstract and like that works for them i don't know whether it would work in a more kind of low-key setting like you know you can't really send cuaro into the world to learn um a detail about human nature and then shove it in his mind palace it feels a bit weird but yeah oh apparently i'm out of time so thank you very much for coming everybody enjoy san francisco in the sun i'm sorry not to be with you and take care thank you thanks for listening [Music] you