Level Design Saga: Creating Levels for Casual Games

[Music] all right great welcome to GDC and I have the unfortunate first session of the day my name is Jeremy I am principal game designer King Berlin and today we'll be talking about level design saga creating levels for casual games little bit about me Prince became designer at King Berlin I'm working on a brand new saga game from King originally I was from Singapore but two years ago I decided I need a little bit more gloom in my life so I moved to pretty depressing but vibrant Berlin I'm a lifelong passionate gamer been gaming since I was six the first game that I play with Super Mario Brothers and I never stopped playing since then I'm currently in my ninth year in the game industry and I'm working in the different facets of this industry from startups to indie to academia teaching game design to serious games to more traditional triple-a companies like Konami we're going games across different platforms console mobile for different markets international and Asian markets both what excites me is that currently I'm working on mobile games and I think that mobile games is a space that it's a new frontier and we are just barely scratching the surface of reasons why I decided to do this talk because if you look out there in the articles and the books there's quite a lot of material on game design or how to formalize how to formally talk about game design but there's a little bit less than material on level design and even if you find such material a lot of it is more about core games than casual games so I wanted to come in with a slant on how we design levels for casual games today's talk will cover four different sections I named them the forties theory overview of what level designers thought the different level design principles that we think about when you're designing levels tools how do we build levels in the saga and testing how do we tweak and how do we think a level so that we optimize them to become perfect so without further ado let's start on the first section theory if you look at these games it's a big spread of games from the original Super Mario Brothers to Goldeneye to more modern hits like Halo or like even Dark Souls there's level design in all these games these games all have levels and a lot of these are examples of what we term as good levels a good level design what is level design in general level design is a composite role it encompasses our code and design and think of it like a funnel you put them together in a funnel and you create levels the difference between a game designer and a level designer is that the game designer is more it creates the rules and the systems for the game but the level designer actually implements them into gameplay experiences in our case if everything goes right and all the stars align good levels fun level lead to money in King and yeah money yay a simple way to think of it is like building blocks the game designer creates the blocks like in Lego the building blocks and the level designers are the people who put them together to form different structures like castles or spaceships or pirate ships in a more formal approach in the MDA framework where it's proposed to again designers on one end creating mechanics and the players are on the other end experiencing the aesthetic experience while this model proposes is that the game designers create the mechanics and when they when mechanics run during runtime they create dynamic situations and this leads to an aesthetic experience for the players when or something that they feel with it if you fun whether it a few exploration or something where a level design sits in my opinion it's it's somewhere here where again coming back to the concept of a funnel and it's funneling these dynamics into aesthetic experiences to speak a little bit about level design King the role of a level designer in King generally encompasses designing levels and progressions of course we create content at a regular pace like that's probably once every one or two weeks we release new levels that's in level designers are involving creating new content as well new power-ups new blockers new game modes and because of the sheer number of levels that we need game designers do level design - one thing they are like about King a lot it's that there's good knowledge sharing in King we have an our own internal app store you can go in and you can any of the other teams work in progress games you can play them and it's easy because of that is easy to start a discussion if any other designer during the regular meetups and the sharing sessions that we have so diving a little bit deeper into what makes good levels this is the section about level design principles one of the ways of how we approach reading a Saga and this is only one of the many ways is it starts a bit like this it starts as with any other game idea it starts with hey wouldn't it be cool if this or that so it's not some a cool idea and it goes to this platform it's not as well it's less famous than our other games if we have a website called Gamescom where it's a browser with website you can play browser-based games that lasts about 3 to 5 minutes and mostly is based on a set of simple mechanics we've picked what happens normally is that oh I happen sometimes is that we pick good games from royal Gamescom and we apply the saga envelope on it and we create a saga like candy crush saga some examples of some soccer games they started this way Diamond digger or bubble which they had original versions overall Gamescom and they were transformed into soccer games eventually what does it mean to create a saga it means in in king we call this the saga envelope normally a lot of you are probably familiar with this by now but what it normally means just with definition some degree of linear progression it starts from 100 and it goes to over a thousand levels for some games if you look at the original candy crush or farm heroes they have more than a thousand levels now there's new content every one or two weeks there's some like social elements like being able to see your friends on the world map being able to send your friend lives and it's intended for casual audience what is really effective about the saga envelope is that it's a strong progression based retention driver because there's new content released every one or two weeks they're new levels of players to play the players are always coming back to see something different to see something new where there's new content or new levels and because it's so effective we apply this across the different King games from farm hero super saga to blossom blasts to candy crush Dora but what this means across King is that across our games we need a lot in a lot of levels so the level design process the building process is super intensive in King Berlin one game that we were there and they follow this model is this game from rogue in Stockholm is colicky London it's a linker and I'm gonna show you a short video of the game play so to better explain what was going on in the video basically the game is a linker you link three or more lanterns to remove them one special feature about the game is creating dragons when you link six or more lanterns you create a dragon and the dragon flies in the direction of your last link you can activate bombs and rockets by linking into them which which will fire after they were after the review for cascade effects and you can collect coins we're bringing them to the bottom or hitting them with power-ups the objective of the game is to score higher score than your opponent so bearing that in mind the next question is that how do we start or where do we even start designing levels this is what the typical level design process in King looks like it starts from a level concept of course so it starts of a good idea how do you make an interesting level you go on to do a level layout how to fit this idea into the ball you create a level and a level data once you think is ready you test it a little bit and you balance it you think that okay it's ready and then you send it out for Rio testing and because we're running live games very often the suit the loop of testing and balancing goes on forever the good place to start is breaking the game down into different kind of building blocks so like in the case of lucky land and you can think of it as breaking it down into four different blocks like game mechanics the basic linking activating power-ups creating Dragons collecting coins because it's a simple three to five minute browser game there's only one game mode which is getting a high score to be your opponent the gameplay elements in the game lanterns Rockets bombs coins and the blockers things that make the add some challenge to block the players from achieving their go clouds in arguably coins because sometimes they get in the way as well if you use the same model to analyze one of our other games this is from candy crush soda so the game mechanics in candy crush soda by basic swapping creating powers by creating matches of four or five for example powerup combos when you match two power-ups together or if you're late enough in the game their helpers as well I sprinkle show I think past about 250 the game moves a lot from gamers we have in candy crush soda pop the bottles find the Bears clear the chocolate clear the bubblegum spread the jam the gameplay elements you have candies your the power-ups you have the walls you have the host things that you can use to build a board like blockers many many different blockers in the game like ice or frosting licorice locks cupcakes honey a licorice whoa the list goes on and because candy crush soda is such a big game we also have what I like to term modifiers things that we put on top of some levels to add a little bit of spice things like the soda physics for example or the gravity or the multiple gravity directions or multiple screens or scrolling screens as well so once you know your game and once you've broken it down into the different elements the next step is to build levels and when we're talking about levels these are four cones these are four concepts that I adapted from this book called game level designed by Ed Byrne the four different concepts one the first one is difficulty how to keep players challenged in the level rhythm how to vary the player experience across the game flow how to keep players moving through as the goal and hooks how to make one level different from the other but what you need to keep in mind for our for our kids particularly is that we are creating mobile games one thing that we use quite often in King is that we are creating games that have this bite-size brilliance these mean this means that our players engage with our games in situations that are different from the normal core game like what it could be on the train it could be when they are waiting for a plane on the airport it could be a mother babysitting a child while playing candy crush before other free hand or it could be friends at the dining table playing this over lunch so the idea is that we're gonna look at these concepts and how we translate them to the mobile format the good place to start of what mix of what makes an interesting level would be starting with a hook a hook is defined as a unique element that makes a level different from other levels in the game and it could be anything it could be this twist it could be a special power it could be a different gameplay style it could be a different environment a different art style anything that makes the level unique if you look at some of these exam in traditional games for example in the first one here in kickflip in Super Mario truly will the hook here is that every time that the player jumps the Panos switched from red to blue and this gives you different one different platforms to jump to over here shared all the Colossus how many of you have played this game yeah great game one of my favorites the whole game as you know is about sixteen different it's probably just about sixteen different boss fights by each boss having a different hook where you need to use a different set of mechanics or tactic to actually be the boss this level everybody knows what this game is yeah path life - rayvin home well my favorite I wasn't high five - where the hook is you start the level and you almost have no other gun except the gravity gun and most of the level you have to go through only using a gravity gun to get to the level and this is from Bayonetta - where earlier in this year's GDC in Abba son when he was talking about creating levels for Bayonetta one starting point of which they had was to start thinking about scenarios what kind of scenarios or what kind of actions scenarios that is interesting for Pina Bayonetta to be in and then designing the levels and the spaces in the narrative around there so how does this translate to our games we start from a level concept so if you look at one way to think about it is how one starting point is that how do we utilize a mechanic how do we utilize a mechanic and interesting or a new way any of you know or having any people play this game Boston blast AHA yeah so for the people who are who haven't played this game but some bizarre guy is a linker as well and the main thing about the game is that the longer you the longer your link is or you make you feel linked you may even make a link powerful enough it triggers a blossom glass which is Prague which is a chain reaction that triggers everything else so in this level particularly is what we focus on that mechanic and the idea is that you only have one objective which is declared this wheat in the middle so the player needs to set the board up so that it comes to this situation where they can trigger Boston blast close enough to the wheat and then when they do it the tricular blossom blast and declared a wheat so this is a level as focus on utilizing that one mechanic in the diff all really specialized me that's one level concept another way the thing about it is another way to come up with interesting levels is what happens if you combine different elements in the game to form new gameplay dynamics this is a screenshot from candy crush jelly candy crush jelly is our latest candy crush game there then one of the new features is that it has a jelly Queen and jelly queen is an AI control correct that every time after you mega turn the jelly Queen makes a turn and spreads the green jelly so the objective is to spread enough of the red jelly over the board so this is introduced early in a game but what happens later is that we introduce this same element and we combine it with another game mode that was introduced before is the puffle is the puffle of our game where they run around in the frosting below and need to find them but the new gameplay dynamic here is that we took the the AI of the jelly Queen and we put it with find a puffer so you're fighting this guy instead called cupcake cow or the most hated characters in the whole candy universe and you need to compete against him to find these guys so this is an example of how we combine combine different mechanics to form new situations or from new level dynamics sometimes it starts with sometimes the idea is to create an emotional wow moment like in this case again in candy crush jelly you look at this bond layer what happens here is that there's only one one move you can do in the first term which is to combine the chocolate bomb with the orange candy and when you actually do that it triggers a chain reaction of all the other stripe orange candies at the bottom and blows up like vertical blasts to the rest of the ball and it's this feeling of wow this starts the level and the whole level carries on with just building on power-ups or coming back to the idea of gameplay modifiers this is a screenshot from farm heroes super saga one of the new modifiers that we had in the game is the wind direction the pieces fall according to the direction of a match and that's the modifier on this this is the first level this introduced when you matched it with an author to write the pieces will fly to the right and still so this is a situation where you think of a modifier and you apply it on a normal level so these are all good starting points of how to create interesting the next step is that once you have the concept the next step is thinking about how do you put it onto the board what kind of boat layout would be interesting for you or for this concept to be ready to be realized so this is example of all my sketches for example I do this on paper first my personal pro tip super useful in meetings when everybody else is talking and you'd want to pay attention that's the best use of your time that you can have so you're all ready so you think okay I have a cool one I have a cool level concept right and then I have a layout I'm already the Builder level it's gonna be fun so you go to the level editor and the process looks a little bit a simplified version or process looks a little bit like this you translate the level layout from paper onto the ball you place your objectives for example clear the ice or whatever you place your game objects so you put like okay like all these objects all these power-ups on to the ball you place your blockers clouds for example or something to get in the way of the challenge and then you run the level and you think okay it's gonna be awesome it's gonna be cool but most of the time it's not the case everything falls apart at this point when you press run why is this the case why is it that everything that I thought in my mind after this point why does it break it's because finding the fun in the level is not that easy doesn't just it's not just about one concept it's not just about one layout the moment you start combining things together you're opening the gameplay possibility space and this is where we thread into the realm of experimental gameplay dynamics one thumb that is where there's mentioned over and over again in level design is this term called moment to moment gameplay what happens from one moment to the other moment as the players play into the level in our case because most of our games are turn base I like to turn this as our turn to turn gameplay and this is where the concept of level flow comes in so flow now the concept of level flow is the concept of how to keep players moving forward towards the goal in traditional games you use concepts like signposting like goes and sub goals or some sort of hidden system what Evan says in his book is that the level designer is the invisible and the guys the players forward and them but more aware that the players are this hand the less fun that it would be so this is a little bit of a paradox an example of how of level flow in one of our games again coming back to candy crush soda if you look at this this is how the boss starts there's only one possible move which is to create a match of six in candy crush soda if you're familiar the mechanics when you create your message of more than five or six for example you created this thing the color converter so in turn 2 you actually activate the color converter and with the green one and this colors all the candies of another color to become green in this case what happens here is that all the orange candies became green and math series of matches and Cascades and then this becomes the ball state after that which creates a new powerup for you which is the chocolate bomb and then when the player in the third time in the player activates the chocolate bomb it clears all the green again and it opens up new possibilities and you can see that what happens here in the fourth or the fifth urn is that the player can now move on like many powers on the ball and the player can choose how he wants to engage in the ball in at this point but he was to go for the fish first whether he wants to go for the the striped candy or whether he wants to match the chocolate bomb the idea is that it starts from a linear one possible move and it opens up slowly so the players I got a guided story along by this visible hand they were talking about the main idea is that you want to move the players forward you want to give them choice or at least some illusion of choice that they have something that they can do in the game but the main thing is that you're still trying to push them forward as much as you can how we do this in our game one example is coming back to this level so if you look at the level this is the this is the layout that you see the level is pretty straight forward if you link these three you activate this lime blast which will hit this and will trigger all these things which triggers a cascade effect so this is the first turn and the first time we are building to us a WoW moment a fuse coolers like okay it triggers all these things it feels really great right but how do we how do we try to engineer some degree of level flow after that if you look at this if you remove the clouds and you look at what's happening behind we have these pieces which has these question marks these pieces are pieces which have a probability of during spawning of being a power I think in this case we tried 30% so imagine all these pieces the question marks have a 30% chance of being a power what we're actually doing here is so that after the first turn when you clear the clouds some of these will be power ups will open up so that you can go ahead to work towards the next stop goal this is one of the ways that we try to move the players forward but again because we're creating a casual game and please engage our games in the most casual manner sometimes all these subtle hints fail and when that happens the next thing we can do or the best thing we can do is to introduce the hand of God in our case that's the hint system which is - yeah highlight super obvious moves like hey over here move so there the players keep moving forward alright so after that point we are creating we're talking about creating interesting levels about how to move the players forward but once you have a set of good levels the next step is to put them together into into a saga so this is an example of what an episode in candy crush jelly looks like in our level data you can see all the levels laid out that yeah the only one I don't look at that and don't trust the numbers here they're all them I'm not difficulty zero but it's just a way in the level data and it's in this layout so that we can move them around make a shuffle them around to create a different one to create different progressions one important thing the thing about about creating progressions is that we want to create a rollercoaster experience in each they have highs and if they have lows it needs to have twists and they say of turns even in one episode itself and this is where I want to introduce the concept or rhythm rhythm is a way of expressing the pattern the frequency and intensity of the sequences of events in the level and the main purpose is to create some degree of variety in the player so that you give them some emotional journey that keeps players engaged in the game ideally a progression or a progression curve should look a little bit like this if the x-axis showing time and a y-axis showing the intensity of the experience it shouldn't be a straight line you wander up the intensity as the player goes along but you want it to be like this where there are highs and they are lost along the way particularly in our case we use different factors to do this things like different game was things like hot versus easy levels short or long levels or even how often do we introduce new content to break them on but not monotony one thing that is super useful for us is this having level libraries and having beat charts so this is example of one of our level libraries where you can see that it's an excel sheet that has other levels listed it shows you all the different game modes and this is the permutation of all the different blockers that we use for the level while super convenient here is that because we have this in an overview and we are creating a progression for example let's say it's about blocker X is about clouds then we can think about we can look for okay we need a level with clouds we need a level with clouds and this game or we need a level and clouds are combined if another blocker for example this gives us an easy overview that we can just copy and paste into a bit chart so that we can lay this out and easily create progressions in our minds at least or at least on paper before we actually physically go and put it out and testing another concept when talking about progression is this concept made famous by Nintendo the four panel Japanese comic concept called ki shoten ketsu it's because the Japanese is read this way from the right the left basically it means to introduce to train to twist and to conclude or two tests in Tendo uses this in their game super often and while the best examples is this one of the best games ever in Super Mario 3d well I'm gonna show you a short video of this well yep so there was a video taken from this website called this YouTube channel called game makers toolkit that shows the application of that principle and now I want to talk about how we apply that principle in one of our games coming back to blossom blogger this is the first time this new game was introduced unfortunately named bloom the big butts so why you actually have to do here is that you need to link flowers of a certain color into this big butt and you need an link enough so that you bloom in its first introduced in this and then strain in a later level where it's the same concept but you need to work your way down to it so that you can do that and the twists happens when we combine this with one of the other blockers the glass blocker the glass block up behaves like a lock when you cannot link into it from the start you cannot link from it from the start but you can have to link into it but and you can destroy it by blooming chain reactions of Awesome blasts what happens here is that the big buds that when you bloom they actually create a glass so it's a new use for the big buzz when you combine them with this new blocker so there's a twist and then in level 20 you see a level that looks a little bit like this which is just half of rapport are these buds and you need to bloom all of them in how many times was there 39 turns yeah this was a pretty tough level same thing for our game coming back to the example of lucky lantern we create a dragon by like in our case we reduced it to five so these are the boat layouts for how we introduce a dragon concept this is the first level where you introduce when you see the dragon you create a link of five and then it clears the clouds and you realize that level is super linear you can only create links suffice that you got five of you to go along and you clear more and more stuff as you go along we train this in the next level where you're creating you start the level by creating a dragon again but what happens here is that you're creating and naturally as you play the objective is to clear the ice and you create it naturally as you play you need to kind of hit these things in the middle but you don't really really really need it as well because you can lock out with like a match here for example and then we twist it here in this concept where you need to use it to hit this particular block in the middle but at the same time we are combining this with another gameplay element that we introduced earlier on we're creating a longer link creates some sort of superpower and you need to use that to clear the outer area and then in this level we tests all the concepts that you learned previously where you need to use the dragon you use the power-ups and into a conclusion here so that is well now we are at the last section which is about testing after creating a series of levels putting them together in some sort of progression putting them laying them out the next step to do is to test them and to tweak them and this is where the concept of difficulty comes in difficulty is easy to explain it's the is what keeps players challenged in the level and so it's a level design through the encompasses both rhythm and flow and by every game it's a difficulty curve because every level is the point in that curve and your game becomes hard or easy depending on all your levels within when you talk about difficulty here to talk about balancing it means different eggs for different games for our case for example in candy crush soda some of the obvious things that we use to balance a level to become easier or harder things like number of objectives so like how many you need to hit how many bottles you need to pop for example the number of moves you have or like for the people who want to go for three-star what are the tree style values but more subtly some of the things that we do to make the gameplay easier or harder things like number of blockers on the board for example the level and the number of blockers on the board or something even more subtle which is the number of colors available on the board because more colors means that the chances of you cascading or forming power-ups is less so you want to make the level easier mechanically one of the things you can do is to actually reduce the number of colors available on the board of course when you talk about difficulty you have to one of the concepts that come to mind is flow theory made famous by Jenova Chen is one of his earlier games flow where he experimental showing that in games the concept of flow theory is something that states Ned optimally you want your players to be here within flow where the abilities actually meet the challenge it represented what happens if the abilities are more than their current challenge is that they become bored and if their current abilities are not enough for the challenge they represent that they go into the zone of anxiety so it's important to keep players here but bear in mind again even though you want to do that we want to keep players within the realm of flow we also want to give them a rollercoaster experience so sometimes in my personal opinion is important to break the flow a little bit so you can see that if you're coming back to this curve of intensity you can see that it follows the flow curve roughly in this zone but sometimes we want we wonder well deliberately break the flow here so that we have levels that a little bit more difficult a bit easier like in this case like you can see that here you wanna level it super easy so it feels good and then you ramp it up so that becomes a difficult level again so that they get some sort of emotional experience but what makes a easy level what makes a hard level is something that's super hard to define it varies from designer the designer even within the same team by formally what happened recently when one example was that I create a level and have this exercise of stating the average number of attempts I expect the player to fail in the level I create a level of my yeah okay this level is hard but it's moderately hot maybe it's like eight attempts that was my guess we stand here for testing got the data back and the actual number that came back was that every sample attempts was 24 which shows that I probably designed I was a little bit too my idea of heart is probably a little bit too skewed so it's super important that we have different modes of testing so the general testing cycle and King looks a little bit like this when you create levels you create you start a level and you create it and you first yourself tested I would say an average of 15 to 20 times where the designer plays it to make sure that everything is okay and then there's some degree of internal testing where we you test with other designers in the team because if other people in the team to play the levels to see if what everybody feels about it is it fun as it easy is it hard and then once that is ready you send it up for qualitative user tests we can we use different channels to see this we send it out to actual users and they play the levels to play the game and they give their feedback a stay at the top going there talking through as they are playing the game and if this is ok if we get good enough feedback then we stand out for play test releases which is often releasing the game in a country with a sample size of a few thousand people so that we get some quantitative data that we can work with in King one of the sayings is that it's where the magician meets the mathematician so as a level designer you need to have a little bit of both you need to have the magical gut of a magician to know what makes something fun and you need to be able to look at data and analyze it like a mathematician if you're looking at free-to-play games of course with ok about the normal KPIs and measurements from retention the progression the difficulty and finally monetization this on its own could be its own kind of worms and it's the top one in itself I don't have the time to go through all of them today but the general concept is that we start from retention because good retention is the starting point and that would gradually and hopefully lead the good monetization later on so if you look at retention the things that come into mind are these two concepts how difficult is the game how difficult levels and what's the player progression some of our charts look a little bit like this this is the example of a player progression chart right the what starts here is a hundred percent so hundred percent of your plays level one and you see the drop-offs across all the levels on the right level whatever level hundred 1120 so you can see the percentage of players that drop to what to pick up difficult levels is that when the drop is steep like for example in this case it's probably a series of like easy levels and then a harder level here on this case as well and when you when you are when you plot that across different progressions we are doing a/b tests it looks a little bit like this each of the color bars here represent one curve and this is the curve of the different progressions the green one is the easiest one because as the it has the shallowest drop and these are steeper drops being harder progressions to reinforce this we also use this other curve to measure difficulty this curve shows the cumulative number of attempts so as the player progresses across the game how many attempts do they take to go level for example if this is level 60 like this is the different curves and how many how many lives of the players spend to get here of course this is normal upward trend we reinforce this again by analyzing the level in this with this graph that number average number attempts to pass a level so like each of these points are a different level and this shows you the average number of attempts so for example if this is like an easy level maybe it's just like two attempts and this is probably a hard level which is average of like nine or ten attempts probably and when you look at data it's one of those things that because you're running a live game it just goes it's something that we have to add coming back to the concept of balancing and testing it just goes on and on and on and we are constantly checking it over and over and over again as you're running the game so that we optimize it and to suit it for the new players as well how many of you have played original candy crush how many of you know or have past level 65 how many of you think that level 65 is the worst level ever strike so it's not that way the history on level 65 is that it started as in the original saga version of candy crush in a very very first version it was the hardest level and there was before there was it was the last level and there was before but that there was before my time and what happened was that my guess is that the designers were thinking okay since it's the last level let's draw everything in this level let's make it insanely hot so the players have a real challenge when it became a proper soccer game over a hundred levels it kind of remained that way and players were stuck there for super long time if you look at some of the comments this is taken from some of the websites eighteen percent of people think that level 65 was the most difficult level ever I love this comment a lot about some feedback about it this level just sucks there isn't any way to describe it really it's not you this level just sucks so but what happened later was that because of the feedback and because we create a mall I was want the players to progress the level was tweaked over and over and over again with the data that we got with the matrix that we have we tricked it using different things that removing the chocolate in the corner reducing the level of the frosting for example some of the different things that we did and then we finally got to this result which is yeah well these players are saying that this used to be one of the most difficult levels and he was modified and now it is one of the fun levels this is an example an infamous example how we managed to use data to tweak level that was not obviously difficult to become a fun level and with that I think we've covered the 40's so just to recap what we're talking about today the section of theory where we give it overview level design thought where we talk about a follower design principles the tools I'll show you how we use the tools to build a saga and testing how we test to tweak the levels we gain a recap of the level design process it looks a little bit like this start from a level concept you create a level layout and then you create a level in the level editor you balance the level you test it and it goes on forever and finally the four concepts again difficulty how to make love how to make fun levels so difficulty how to keep players challenged in the levels rhythm how the very experience so that is emotional journey flow how to keep players moving towards the goal and hooks how to make one level different from another and we're done [Applause] yep questions there Mike yeah please stay yeah let me just give you a mic so be great if you introduce yourself and where you're from before you ask the question yeah my name is Warren um from Bigpoint in Hamburg yep that's a really good talk with me thank you so much for sharing so is each and every so basically from your presentation each and every level on the saga games like handcrafted right which is really really cool but is this some sort of an algorithm or some sort of machine learning that goes hey player X has failed that level 71 X number of times so it's going to automatically tune the number of tones he has to give him a few additional tones maybe yep so it's a good question interesting is an interesting question so if I think if I got it right the question is that is there any kind of machine learning for for to know the levels are easier they're difficult right so to be a little bit of a tabular bit of a self company pimping hall there's a talk later today I think where it's talking about how AI is using candy crush and that's exactly what we do we actually have BOTS in the different teams where they play the levels to where the where the bots places levels to see the average number of attempts of passing or failing and we use AI for this having said that what I know and I'm not expert in this but what I hear from the other teams is that the bots are normally super inaccurate they are good for predicting some degree of difficulty on a curve but there's always some degree of offset as well because a lot of it still comes out the baba visnu comes from actual data and we send the levels out live and we take it from there hi my name is Sophie and I come from Google I had a question about more when you have levels live so you said very important about testing balancing and so on they worshiped yep I remember I saw in some levels of candy crush where people can rate the levels so how do you use that data as well in the process of rebalancing the levels so where people can realize yeah so you finish the level in your as a player so how did you enjoy the level and you put star rating mm-hmm all right so this is an interesting question and I cannot really directly answer it because I'm not I didn't work on candy crush soda but I think it's probably one as king because we're running live games and we're testing things all the time most probably that was probably when you saw it it was probably either a bee test or something that was just running for a period of time but from my best knowledge I think that the main data that we still use are some of the stuff that we that I mentioned in the talk things like the normal drop-offs or things that the average number of attempts so maybe is factor in like how fun it is but fun is always like a super subjective thing and it's really hot a lot how do they use that kind of data to fit your levels compared to more like more hardcore numerical data yep was there another question yep yeah just a straightforward question how do you sup launch levels like do you four end players do you have like really them first in some other countries or is there a smarter way to do that so I think different teams do it differently but one of the more effective ways is really a be testing so when we're launching a new episode we often just a be test it if we have two or a be testing in a smaller country in a smaller group and then see what the data is so again coming back to but of course these kind of processes are more qualitative than quantitative so even though we try to do that but a lot of its - I would say it turns out quite differently when we actually reduce the levels live to a larger volume so it's a super tricky process which is why it is this process of going on about doing it almost like forever so for every chapter you release you a be tested and then you keep the best levels hmm not every chapter but sometimes depends especially depends on what the purpose is if there's a new blocker sometimes it's important to do that but most of the time if it's just one progression we also try to do some sort of preliminary testing by sending out a smaller sample group even it's one progression just to see like some initial feedback hello I'm mark from Bohemia Interactive my question is how you control the random aspect of the levels like especially considering the casual audience hmm why there are no dangers all right so this one is super tricky and depends from game to game right if you talk about game depending on the core of the game if you talk about game like blossom blasts which is very focused around chain reactions the degree of randomness from the core itself is a lot higher that's something can happen from one thing that the other turn right in candy crush similarly as well because after you make a match even if it's a small match the chances of the cascading into something more or like creating a power of straight cascades that's pretty high and so most of our games are built with as some degree of luck in the core mechanics and to do it like coming back to one of the previous examples that I was showing the the slight way has this power-ups they had the chance of some objects being power-ups we try to do this semi deterministic approach to some extent where some of these things have the chance of spawning there's an hour game because we don't create power-ups on the fly so they're different factors to try to do it and really depends from from game the game yes there's so again different games that have different methods of doing it like I think in candy crush they use a seeding system to determine certain seats and they use that to to filter out what's a good seat versus a bad seat but yeah there's so many different ways to do it like it's one of those things that because of the sheer chance of it and yeah because of the chance of it is just so hard to always be able to predict what's gonna happen so it's not something now it's some degree of we have some degree of control to it but very often it's also at a point where you started bored and you look at it and you're like oh crap it's there's no way I'm gonna win this level in you know that and five turns so that's when you have a bad sorry I mean yeah so we try to avoid that but if you see that if you play enough like especially I saw the candy crush games like soda like there's some levels where they have blockers on top if you do enough wrong moves at the start of the level you can come to situation where there's no move and normally that becomes like a heart filled scenario we try to avoid it as much as we can but still it still happens so there are scenarios like this and the opposite of that is the positive side what happens when you create so many power-ups there cascades forever if you notice this and this is pretty subtle in can I cross over again if after I come in a number of rounds of Cascade so but what happens if you can't get too much is that if fro a six can be the the Indigo candy into the mix so that the cassia is just stop so yeah there are fail states on both sides to prevent things from yeah whereas too good about is too bad thank you yep any other questions so as a company we are trying to experiment with this especially right and when we start that linear progression is one of those things that is new it was new it was easy for casual players to understand and it's it's its age all right if you look at traditional even from the old Mario is about linear progression is about moving seeing new levels to the other right so this is a super strong driver for for casual players to always senior levels to play to move forward having said that candy crush is like four or five years oh by now and we are experimenting with different things to make like nonlinear progressions as well but I think there's a long unannounced project and we're experimenting a lot a lot of these internally but we haven't found or we're still trying to find the best way to present this the biggest problem here is that how you get players from being so used to linear progression of just moving forward to see the next level how do you get them into for example what happens if we do branching paths or alternate paths or where they can choose why the requires becomes open-world I think it's a it's a you it's more like a UX problem of like trying to figure out how to push players into this there is definitely a strong motivation and a strong driver for nonlinear approaches but we're still experimenting with the best ways to actually do it for our audience yep how do you keep inspiration Tyrrhenian when you from the designer point hmm the heart the heart ultimate ultimate of the ultimate ultimatum solution is transferring to another theme so that you start something new but yeah it's difficult because it depends different designers 1 different things some some people are really ok with just building new content and they want to specialize in there I mean and some people are just like they kind of just wants to do new stuff all the time so the good thing about King is that the company is big enough so that we have a lot of movement of internal like main power source whether it's the common so where is transfers or even within the studio moving to another project so that everybody is equally motivated I mean say that there is to definitely some degree of monotony maybe to this whole thing but I would love to say that yeah I would love to say that knowing that our players are enjoying our games makes me happy but I'll be lying so it's it's there it's something is there you have to do it and you try to find a fun in it like game design is not always fun there the good parts and the bad parts right so you find something that clicks any other questions do designers like for candy versus up do they have to keep monetization and mind as well or is that a completely different it's mmm this is where it becomes a little bit tricky to speak about because I'm not sure which point would leave Miley go come and stop me and like so the focus is our intention first for sure right Maya's ation becomes its own problem later but quite often we have cases where we start a game thinking about building fun levels and that's always where it starts right and you are you build fun levels it's fun players love it but sometimes it's maybe it's too easy maybe it's too boring maybe it's the same thing over and over again and as a monetize right so normally the process is that we look at different matrix matrixes at different point of where we are and play tests so we always start from looking at like short-term retention long term retention or mid term retention and then we look at monetization and there are some things which I probably cannot really super share about what are some of things that we do - or what else are they say King knows of King tries to maximize my ization yeah but there's definitely different stages and different things that we apply thank you mmm this is different from depends on how big the game is and depends on where it is I think if i'm not wrong candy crush at one point had four level designers per team and this is just level designers i think not even considering the game designers like on our team currently because our game is still going potentially going to soft launch we have three designers but two of them but we all do level design yeah when it comes to rhythm what do you think like how many things can you introduced in the early pack of the game mmm because you have so many different level types respond across the game but what should be the is there a standard that you do well depends on whether the game again depends on whether it's a new game if a new mechanic awareness is one of the sequel's right if you look at some of the current examples that we have farm hero super saga or like candy crush soda what these the sequel's the approach is normally to introduce your new features as early as possible because it's the five-minute experience you want players returning players to come in and see something new so that they get hooked right if it's a new game I draw some blast for example I think if I'm not wrong remembering the beach shot is throughout the whole game there's only three game was for the first hundred levels I think it's called the bloom the big butts and remove the weeds for the first hundred levels but every twenty levels they have a blocker or something so depends on again depends on the game and depends on the content but when it comes to rhythm I think some of the things that we try to do is to not use the same game mode for five levels in a more than five levels in a row or three to five or three levels in a row so that it's not the same game over and over again and again alternating between high and easy level so that it's not too hard I have also three hot levels in her in a row they're frustrated players and make them churn so we tried well we try to monitor these things a little bit more closely alright my time for one more question if there's any final questions while having like hundreds of levels is there someone who guides the over progression like the overall curve of the whole some of the levels or is it like you have individual level designers and everyone is doing his own thing or is there someone who stirs like listen we would need like I don't know ten harder levels with this kind of like constant in them or is it just individual so I think depends the question I think the question depends on individual teams again in my in in my current project like we have three designers and because we are still early stages I can I try to oversee the whole progression to know that each level was like yeah Michael my designers are always saying that like I like I'm just a freak because I can memorize all the levels because I just know we should I've always wear like mysteriously but I think at some point it cracks I don't think I can keep these on like maybe this was for the first fifty or the first hundred levels maybe but when you go to a thousand levels it's hard to memorize all of them right so I think different designers work on different things as you get later and later in the game the first the most important is the first few levels the first fifty levels for example because this is where a lot of players decide whether they want to play the game or they want to drop or the game right so I think I for me I'm very particular about these personally I'm super particular about the progression in these levels and I think in the later levels once you get past a certain point the players are used to the difficulty and then it becomes more it becomes more self-contained every episode has its own highs and lows they're getting more difficult gradually towards it per episode but sometimes it's just for example the number of difficult levels in the episode versus the previous episode and not so much about absolute levels being absolutely harder than they were previously it's called a difficulty budget for every episode how you can how you can ensure that curve you have shown because like if everyone is doing I'm doing my level and yep more difficult yep so normally I think one process is that the team makes me sits down and when you play it let you try to play a progression through you know roughly which are the easy levels I wish at the hot levels and you bring this up for an internal review but at that point before I sign up for any kind of data is really basically just a designer's magical magicians cut guess of like okay I think this is fine I think this is easy now for this is hard enough and very often we need to use the data that comes in data to treat them which is why the tweaking goes on for a super long time thank you alright I think we're out of time but one favor that I have is that can I take a selfie of everybody so that I can show my mom alright thank you [Applause]

Channel: GDC

Views: 22,242

Keywords: gdc, talk, panel, game, games, gaming, development, hd, design

Id: LuNH9Rz2e2k

Length: 57min 21sec (3441 seconds)

Published: Wed Jan 30 2019