Building the Content that Drives the Counter-Strike: Global Offensive Economy

just a reminder before we start to silence your cell phones I'm Bronn I'm a technical artist at valve and today I'm going to be discussing the what's how's and WIS of the content that drives the cter Global Offensive economy so what is this talk I should first start by telling you what it isn't I'm not going to spend a long time talking about the advantages of adding an economy to your game I'm going to assume that the reason you're here is you already have an idea about that if on the other hand that is something you want to learn some more about um we recently gave a series of talks at the steam developer days on economies all that content is freely available to watch on YouTube so just have over to steam dev.com for some links so instead this talk is about how we generated a large volume of content with a really small team in order to Kickstart our economy I'll be discussing the things that our customers value about the economy content and how that information is helping us continue to make good decisions going forward Counterstrike is an online multiplayer game game it's a team- based

first-person shooter and it's had several uh versions over the last 14 years Counter-Strike Global Offensive is the latest greatest version like we do with our other products we use steam to update CSO regularly as we strive to create the best experience possible in a multiplayer game a big part of having a good experience is provided by other players our customers generate a huge amount of value for each other just by participating in the game we want to encourage them to keep doing that the way that we encourage them is uh just by providing a bunch of fun and value to them and more importantly by making it easy for them to provide a bunch of value to each other involveed products like Dota 2 and Team Fortress 2 we found that a great way to help meet all of these goals at once is to add an economy to the game we can measure the success of our economy by checking in just on a few broad metrics and they're maybe not the ones that you'd expect after all adding an economy isn't a goal in and of itself we're adding an economy in service of our other goals

so the kinds of things that we're going to check on are things like player numbers player retention and individual play time because after all if our customers are having fun they're going to keep coming back for more since Team Fortress 2 and Dota 2 both have thriving economies we were able to crib a bunch of ideas from them they both use multiple channels for Distributing items to their players this gives customers lots of options for getting items having lots of ways for customers to interact with the economy is a big positive because it means it brought a participation and economies are essentially multiplayer games the more people participating the more fun it is in both Team Fortress 2 and doed to not every item is available through every channel this is perfect when the goal is to provide lots of different ways to interact with the economy and to see how our customers want to consume content but it's really difficult to directly compare the value of items in these different categories and at this point in csgo's economy design discussion figuring out the value of individual items was actually really important

because leaving aside the how for the moment we still weren't sure what our customers wanted to consume so let's focus for a second on the marketplace this is one of the ways that our customers can get out the content customers can list items from their inventories for sale for steam wallet funds the listings are in real currency amounts and they're set by the Lister because identical items can be listed for different amounts there's a constant downward price pressure as um the price pressure makes the supply and demand at a given price point equalize so as a company we had a unique opportunity this was the first time that we were launching an economy since the market went live we decided that csgo would become our first game in which every item can be placed on the marketplace and individual items can't be bought any other way in fact the only thing that we plan to sell directly were the keys used to open our case drops so in this proposed market- centered economy the supply of items comes from our players they come from in-game drop so the more players we have playing the

more Supply there is the demand also comes from our players they're only going to purchase things that they feel are being sold for a fair price and with enough listings the downward price pressure of the free market is going to ensure their fair price gets found so this sounded like a great way to Simply ask our customers what they value so what did we decide to ship we could make weapon skins character customization we could do new weapons or we could just do cosmetic mesh changes to to existing weapons all these things have been modded by CS players in the past in fact all the examples that I just showed are Community mods from game Banana so clearly there is a demand for customization whatever it was that we decided to ship we needed to create a lot of items in order to fill out the economy so it's important that we make it as easy as possible to do so and in keeping with our goal of helping our players generate value for each other we'd like to make it possible for community members to contribute items to the game and make money from their in-game sales we also need a lot of

variety in order to figure out what our customers want we know that economies are more fun when there are more people participating they're also more fun when there are lots of items to compare and lots of different ways to compare them so we didn't want to just create items with a broad range of value we also wanted to create items that could be valued in different ways because players with different priorities can actually care about different aspects of the same item so we'll want to keep these goals in mind as we figure out which thing we want to start with the way we choose which of our options to go with is to assess the benefits associated with each type of product and then compare that against the risks associated with shipping it and the resources required to actually complete the task at the time CSCO was actually a really small team our art team consisted of two technical artists and that's the whole list we pretty resource limited so let's take characters for example we know that customizable characters in TF2 and Dota 2 are really valuable and popular but they work best when there's lots of chances to see the customization in Dota 2 the game is topped down so you get to

see your character all the time and in TF2 although it's a first-person game there are taunts and kill cams that give you lots of opportunities to see your customerization or show it off to other players CSCO is also a first-person game but we don't have taunts so the chances to show off your customization are actually a lot fewer than our other products and so the value here isn't maybe as high as we' first hoped there's also a lot of challenges here and there's a big risk that we can make team identification difficult since csgo is a team- based game team ID is a really big part of having a good experience so it's possible that someone could have a customized character be really happy that they had something different and unique but they made everybody else that they were playing with less happy because it was more difficult for everyone else to tell friend from Fox it's also a big resource investment not only do we have to take time to try and solve our tmid problems but there's a lot of content that needs to be created and it's fairly complicated content with our small team it would take a really long time to make enough content to be able to ship the economy

that's bad for us because we'd like to ship something as soon as possible especially something that we have doubts about because then we can get data back that either validates their decision making or indicates a different direction that we ought to go so on the whole it didn't seem like characters were the best place to start what about new weapons it could be really fun having a lot of gameplay variation on the other hand our players practice intensely with their favorite weapons right down to learning The Recoil pattern so that they can learn to compensate they're not going to throw away all of that time investment and try something something new just for the sake of variety that means if we make new weapons they have to have a unique tactical value that's going to attract players to them and really there's only so many different options that we can provide that are meaningful in variety and so actually the value of shipping weapons goes down the more of them we ship which is directly counter to our goal of providing lots of items there's also a risk anytime we ship new weapons that were going to destabilize competitive play that could potentially make our customers so unhappy that just leave the game and

never come back and that's the worst outcome that we can think of there's a fair amount of resources required to do weapon balancing to attempt to mitigate that risk on top of just the content work required to create the weapon meshes so again it looks like new weapons are maybe not our best place to start so what about weapon mesh changes there's actually a surprising amount of benefit here after all if you're changing your weapon it's something you're holding in your hands so even from the first person view you see your customization all the time it's also something that could make other players more happy that you have it not only do they get to see something new and cool because you're walking around with the weapon but if you drop it into the world someone else can pick it up and play with it and have a positive experience we did have a couple of doubts though so let's just see what those could be so we actually got pretty far along this path we customized the hecka of the AK-47 uh we played with it in game and we did find that our concerns were validated the first thing is that weapon identification is made a lot more difficult with all these different Silhouettes it's super hard to tell just by looking at these guys that

the functionality for every one of these weapons is identical and that brings me to my second point we were a little bit worried about how it would feel to have different components that didn't actually affect gameplay and the answer is that it felt really artificial in the real world if you changed out things like stocks and grips the weapons recoil and accuracy would change and the experience of getting weapon components that didn't affect the Gameplay at all was actually kind of disappointing so we reassessed the risk even upwards from where we thought it was there's also a bunch of work here obviously just to make the components so that leaves us with weapon skins at this point we were getting actually really concerned because how many different camouflaged weapons can we make that people are really going to care about but as we got further along Gathering reference we discovered that there's a whole world out there we weren't even aware of people make crazy things for their weapons everything from uh intricate Victorian scroll work right up to straight up graffiti these things are essentially just blank canvases and that means that there's room for Whimsy

that's not the case with our weapon mesh changes where those things have to be based on real world components for weapon skins we can make pretty much whatever we can imagine so the benefit is a lot greater than we first thought they're also not very risky there's a little bit of risk that the color changes would make the weapon identification a little more difficult but on the whole it's nothing compared to our other options and then there's just the work of making the textures Which is less than any of other our other options it's also really easy for the community to make these easier than anything else so we have a pretty clear winner in the real world weapon finishes come in a vast array of styles and colors so we already had a really good indication of how far we could go with creating variety now we just needed a way to make a lot of things really quickly since both our artists were technical artists we did a very typical Tech art thing instead of starting to concept what the weapon finishes might look like we jumped right into the task of trying to make them easier to produce we started with a basic idea painting detail into Tex Tes takes a

long time it's a lot of work and doing that on a large number of items seemed really inefficient so we wanted to automate the process since cso's art style is fairly realistic we started by taking a look at real world reference paint wears off metal in a really predictable way so there are certain effects that we know we want to reproduce for one thing it chips off the exposed corners and edges it wears quickest where it gets the most contact and grime and discoloration to create in the cavities and another thing is that when people paint their weapons they typically don't paint the whole thing they're really careful not to get paint in the mechanism for obvious reasons so each weapon has its own characteristic areas where the substrate that's the original surface of the weapon is always visible so in this case only the the receiver the grip and the butt stock have been painted there is one thing about this example that isn't ideal though and it has to do with scale this has a really durable paint on it and so the way the paint chips off is in these kind of uh larger chunks that looks pretty cool on a rifle but if you try to apply that to a pistol it might look out of scale and

that's an important distinction because we want to ship as general a solution as possible whatever we make it has to apply to everything so we kept looking through our reference and we found an example where the wear pattern is filled with these smaller size scratches this indicates that this has a lower quality paint but this scale of detail is going to work well on every type of weapon that we have so we're going to try and Target this look it's pretty easy to replicate all of these features so long as we have good enough information about the topology of the mesh in fact the most important things to know are just the cavity and the occlusion that's going to tell us our low points and our high points for the mesh so that we can figure out where the paint should chip first for our scratches we can just tile a texture over the entire weapon we can start by just combining the cavity and the occlusion the brightest areas are the places where the paint should wear off first this is a pretty soft blend though and our scratches need to have hard edges so we can just use a smooth step in the Shader to adjust the contrast for the artists in the audience a smooth step is really similar to photoshop's levels operation

so this gives us our blend between our paint and our original substrate Now's the Time to bring in our C our um Grime we can just do that using the cavity that tells us where to place the grime and the grime itself is another tiling texture similar to the scratches so next we'll add in the ambient occlusion to get the paint and the substrate to sit in the same world together that's already looking pretty good there's uh one small problem though and that's that there are a couple areas where we have some markings and etchings on the original weapon that really ought to show through I don't think you can see it because it's a bit dark but trust me they're there we want to see those even over the paint so the cheapest thing we could think of was just to add them to the ambient occlusion since that's just getting multiplied over the surface of the paint anyway we can get our details back in dark over the top so another thing that we saw in our reference was that the paint wears off quickest where it's touched the most there's no way to know where things get touch just from the topology of the mesh it has to do with human interaction so we need to have an artist's control over this we added a wear influence texture that

is artist created for the most part we just grabbed details from the original Source texture psds and then we just painted a couple of Blobby details over the group in the magazine with these big soft brushes we spent hardly any time adjusting this mask at all and this is the effect that it produces you can see especially on the grip the paint is wearing where it's touched the most so let's add lighting so at this point we've authored almost no new content by hand the only things that we've created are really The Grind texture and the scratch texture and those things are going to be reused over every single weapon in the game anyway the fun part is since the Shader um has a wear input we can just animate that input and see the wear crawl over the surface this is great because it shows us that we can get a wide variety of looks just by editing a single Shader parameter it was around this time that we started testing the content in game under real game lighting and what we found was a little bit concerning so here's a typical first person view of a weapon

um I'm showing it without the world or the view model arms just so you can focus on what's important you can see that the way that this texture has been authored it calls out all of the edges and plane changes so that even if we disable uh directional lighting all of the plane changes are obvious and you can understand the form really well but what happens when we add our paint onto it a lot of those plane changes just straight up disappear even with directional lighting we still have a problem surfaces that have the same normal but different spatial off set still look as though they've been merged into one one of our concept artists is a big model making nerd so as we were figuring out the paint system he recommended that we check out work by mink Jimenez I recommend this to any of you in the audience if you're a texture artist a Shader artist check this stuff out he's got a lot of amazing techniques that are actually really helpful he uses a technique called modulation it gets painted on as part of the base coat before he adds the details and it's used to call out geometric details as as well as edges and plane changes so this isn't

exactly lighting although some of the directions of the gradients could be influenced by lighting instead the goal is to create a value contrast at every single one of the edges the details are painted on afterwards and they reduce the obviousness of the effect but the Improvement in the viewer's understanding of the form remains we tried it out truthfully this was not at all fun the process was just to grab an element and then planer mapa gradient onto it we then rotate the gradient until it's following the way that we want some of the more complicated weapons like the machine guns I swear these things are like the Death Star they have all these tiny little gbly bits on it some of them took up to 2 days on average about 2 hours per weapon the result though it works adding in the modulation shows us our plane changes gain that means that we can move forward without having to worry about the paint destroying the viewer's understanding of the form we obviously don't want to just apply solid paints to every weapon

though it's a pretty promising start but uh we researched a bunch of different finishing techniques and decided that our first technique would be inspired by hydro dipping this technique uses a film with a printed pattern on it it's floated into a tank of water and then sprayed with an activator the activator dissolves the film but leaves the pigment behind and then you take your primed component and just roll it through the pigment to pick it up onto the surface we imitate this in a super cheap way since Hydrographics have seams in them from being dipped from multiple angles and because the weapon's disassembled before it's dipped anyway we can just tile a texture over our unwrap if we have seams in the results I just dialed up the realism so You' think that the first thing we try would be to replace the flat color from our prototype with a bunch of different textures but that seems so single use and in the long run it didn't seem like a good first investment because we'd still have to create a lot of different images so in order to be more efficient we tried procedual coloration of patterns I'm

showing this to you really small because it's super hard in the eyes authoring these again is not super fun but once you have one you can use it in a bunch of different ways so you can change its scale its orientation Drive different colors through it and create a bunch of meaningful variation again just by driving some Shader parameters we looked at a bunch of different examples of hydro dipping and we found that for the most part people don't dip the entire weapon they'll take certain components and just use an accent color on them we thought that this look was more appealing than the head to toe camel in order to replicate it we paint a bunch of uh paint by number textures for our weapons so they combine with the patterns in order to create a weapon specific color mask the colorized paint ends up looking like this everywhere that's blue gets the same color everywhere that's green gets the same color it transforms our head to toe gam camun into something that's a little bit more appealing the results of the composite can be

randomized uh by varying the rotation and the offsets of the texture the wear is randomized also and that means that every single item we generate with this system is per pixel unique so this is all looking really promising but this is a pretty expensive Shader and it has a lot of texture inputs meaning that it's pretty heavy on memory as well so obviously it would be better not to run this thing every frame in fact we'd like to run it for just one frame and then save out the results I've been showing everything so far in the the context of the weapon meshes but actually all of our inputs are just texture information and so we can run this thing in a flat plane like a render Target and still get the exact same results that's cool because we can just save this out and use it as a normal texture that gives us an interesting Advantage it means that we don't have to edit any of the materials for our weapons we can just swap out the base texture and still get all the variety that we want so our first for into generating the new texture through a render Target wasn't all that promising

we had a 2C hitch every time we gener generated a weapon which for a competitive game is plainly unacceptable so we asked for some help we stopped trying to do everything with just technical artist and involved a rendering programmer he made some optimizations to the way that we copy the information out of the render Target and save it out I'm not going to go into detail on this stuff if you're interested I can come back to the slide during The Question period the upshot is that it only takes about six milliseconds for a single frame us usually towards the beginning of a round when it doesn't matter and then the rest of the work happens in a separate thread so there's no gameplay hitch so now that we have a fully functioning prototype we felt confident enough to start creating some new Styles the first reference that we looked at consisted of spray painted weapons we started with Hydrographics because it was easier to solve those problems um here's the thing with Hydrographics we don't care if we have seams in the results but with spray paints really they should be seamless so normally we just do a tripler mapping so we project the texture from the cardinal directions and then blend over the surface normals

this requires having mesh data to work with though and far from having the position and normal information from our mesh what we actually have is just the flat plane of our render Target with no normals at all so what we actually need to do is get all of that position and normal information into a texture format now the normal information is not that complicated we just render an object space normal map um but you should bear in mind that we never authored the U for any of these weapons for procedural Solutions in fact all of the content had shipped to our customers months before so there are some areas here where we have overlaps or mirrors in the UVS this would be a problem if we were shading with this information but actually it turns out to be plenty good enough for just doing the blend between the different texture projections our position didn't turn out quite that well we need a lot of precision here in order for it to look accurate we save out 32-bit compressed textures for the position and it's still not enough we

end up with some pretty serious blockiness in the mesh we solved this by multi sampling we just use a gausian kernel to average the positions and the result is actually pretty credible in terms of visual range though the difference between Hydrographics and spray paints is not all that large so we've managed to add some variety but we want to go quite a lot further as we were looking at our reference we found that another common way that people customize their weapons is through anodization our standard Shader for the weapons does actually let us achieve this effect so we don't need to make any changes to the way we're rendering in game what we do is just tint the specular highlights by the Bas textures color and so we can get some nice juicy highlights like this so in this example the paint's colored red and baked into the base texture and then we just tint the specular highlights where the paint appears and not over the substrate it's pretty but it looks really weird having it over the entire weapon so remember that mask that we used to create accent colors for our Hydrographics we didn't use all the

channels did we we still have one left so we added another per weapon mask to limit the anodization effect after the mask is made for a particular weapon the only overhead to creating a new anodized style is just picking a new color we didn't stop there how could we we have so much texture information that has so many useful things in it we haven't taken full advantage of it yet so we added some more Styles we finally got around to creating an uncolored hydrographic we also converted our hydrographic and spray paint Styles into anodized looks we made a solid paint Style just using our paint by number mask to uh generate a paint that has up to four colors in it and lastly we created a patina style that instead of wearing down to the substrate where is to a thicker patina we didn't create any new per weapon texture inputs for any of these we just remixed the stuff that we already had using our Shader I really want to stress that this is pretty much just procedural art it's the cheapest kind of art that we can

make and it communicates the copper material really well it's the video plan guys yes okay excellent uh it communicates the copper material really well and it also looks interesting and different at all the different uh wear levels and so we' managed to create again a bunch of variety just by driving a single Shader parameter so to recap so far we have a procedural compositing system that works at runtime that we can use to generate an essentially infinite number of textures in order for the system to work we need the following inputs the original Source texture which was authored before we even had this idea the cavity AO position and normal all of these things are just generated using out of the boox tools from our existing mesh we create a paint by number texture that's done just by colorizing elements of the mesh and then rendering to texture so it takes hardly any time at all and then our artist generated wear influence map now we mostly just cribed some layers from the original sour texture PSD so again this isn't a big time investment in fact the only texture

that took a significant time at all was the Mig Jimenez style modulation this guy took on average about 2 hours per weapon so if we add all these together it took us about a half a day to create a paintable weapon even with a really small staff of just a couple of artists that means that we could blow through all 30 weapons in the game at the time uh in about a week and a half now this does look like an awful lot of texture inputs to our Shader uh some of them are grayscale though so we just combine them into a single texture by copying those into the textures individual channels the other component that we need aside from having the weapon be prepped to be paintable is the finish that we're going to apply to it what we author depends a little bit on the Finish style but for the most part we create a pattern and then choose the colors for that pattern and then set just a couple of Shader parameters like the specular intensity once all the weapons have been prepped we can very easily author a single finish and apply it to any one of

the weapons let me zoom in a random one here and you can see all of the detail that's been procedurally placed and I'll show you another variant just because they're so darn easy to make so a single weapon finish can create a huge amount of variety both by randomizing its parameters and by applying it to different weapons at this point it's time that we stop focusing on the tech and move on to the art over the course of gathering our reference we discovered that people in the real world create hideous hideous things it really begs the question what do our customers want the game's art style is grounded in the real world and we've already gone down the path of replicating some real world finishing techniques but there's a really broad range of service providers in the real world creating a broad range of things that cater to a broad range of tastes in our game World we're the only service providers in our taste might not match demand so we needed to focus Less on the things that we personally love liked and we in creating a broad range of variety

in order to measure people's reactions so we shipped things with a big range of Aesthetics and we deliberately shipped items that were contentious within the team so among all these items you can see a bunch of things that are people's least favorite out of all the test content that we made we also shipped an in-game tool for creating finishes that we call the workbench it allows the community to author weapon finishes for themselves and upload them to the community Workshop they've done that with enthusiasm creating over 30,000 finishes we've already started shipping some of the highest rated items letting our community Express their aesthetic preferences directly by just putting the things they make right into the game we called our economy update the arms deal update and it introduced the weapon finishes that people can collect and Equip to customize their play experience our initial offering contained over a 100 weapon finishes all created in house which we've since expanded almost fourfold with the help of our community contributors each of these finish has random variations created by the pattern

placement and the wear meaning that there are actually tens of millions of unique items in our economy so checking back in with our goals so far we've managed to meet the first three it's really easy to make a weapon finish the community contributes to their creation and we have a large variety of different looks our last Wills are to have a broad range of value and to create items that can be valued in different ways so to know if we're meeting these two goals we're going to have to check our data weapon finishes are by definition luxury goods they're things that our customers like to have but they're not actually necessary in order to play the game in fact they don't have any gameplay impact at all there's plenty of academic discussion on the dimensions of luxury goods I've collected together just a a few of them that we can use to identify uh what might contribute to the value of items in the economy so let's start with conspicuousness this is the public display of your status as a person with access to luxury goods this is a graph of the sale of uh sorry

the sale price of a black laminate AK-47 over its first 3 months this isn't privileged information anybody can see any graphs that look like this for any of the weapons in our economy just by visiting the marketplace you can see that the price is pretty stable around $10 so that means that the Community Values this item pretty highly here's another item the Doom Kitty Famas prices for this item started pretty high around 750 and dropped really rapidly over just a couple of days now it maintains a fairly steady price of about 7 cents so this guy isn't a huge hit so what's the different between these two items while the finishes themselves are Superfluous the weapons that they apply to aren't so this is a graph of weapon utility over several rounds of the game it shows which weapons get used uh most frequently the status from last October when we' first Shi the economy the AK-47 has the highest utility of any weapon in the primary weapon category the M4 and the op are next uh with roughly the same

utility each and then the rest of the Primary Weapons don't have a big share so our Famas is the one that I prematurely pointed to um that's the one from the previous slide that had a really low price on the marketplace you can see it gets used a little bit in the beginning rounds of the game but by the time we reach the later rounds it has this tiny little share sandwiched between the much higher utility weapons this shows us that if you have a fancy AK-47 you have a lot of chances to show it off and if you have a fancy Famas you don't if we look at the prices of weapon finishes that have the same scarcity we can see exactly this breakdown the AK is on top and then the M4 is in the op and then a cluster of the rest of the weapons down below this shows us that the amount of time that you spend holding a weapon influences its value quite a lot that's just what we'd expect more time holding a weapon means more chances to show off your customization and also to enjoy it for yourself so conspicuousness is definitely a big factor on price that means that when we create a weapon finish just applying it to different

weapons can create a big range of value the next item on our list herin and personal history encompasses things like historic value Nostalgia personal meaning it's really easy to see how those kinds of things can influence the real world value of an item things like heirlooms uh items that belong to famous personages items created as part of a venerable brand but what about in a video game maybe it's less obvious but counter strike as a product has a 14-year history this is the Desert Eagle more commonly known by its nickname the deagle it was a very popular choice in previous versions of the game where it was a weapon with a very high utility we have made some changes since but last October when we shipped our first round of content the utility of the deagle in csgo was actually not particularly high in fact the utility of the P250 outweighs it quite considerably since that's the case uh we'd expect

that the P250 would have a higher value but it doesn't the deagle is actually the one on top it retains its high value because people are used to valuing it highly so this is an example that shows that Heritage and personal history can actually affect players valuation of items in a game scarcity is easy to explain the less of something there is the more value each unit has we sorted our items collections each collection contains items from different quality tiers the quality tiers correlate roughly with scarcity so it's more common to get an item from a lower quality tier than an item from a higher quality tier we do have two different types of drops we drop cases and individual items the cases can be opened with a key and it gives you uh one item from a collection that's associated with that case so this collection here is from a case and this one is made up of individual items so we control the rate at which both the cases and the individual items drop relative to each

other the thing that we don't control is the rate at which cases are opened we've used our best guess at case opening rates before shipping the economy in order to try and balance the number of drops so that our quality tiers would stay roughly correlated with scarcity um but in our first couple of months our customers opened way more cases than we'd expected and that means that case items are actually more common than drops so as a result the quality tier of an item doesn't actually represented scarcity relative to items that have a different Providence so there are two items here that I particularly want to call your attention to this is the red laminate AK-47 and the black laminate AK-47 visually they're fairly similar the red one is actually a little bit uh punchier so that's one way that we might have expected that it has a higher value because it's more striking um it also has a quality tier two tiers above the black laminate AK-47 so that's two ways in which we could expect it has a high value except due to the volume of cases

opened there are actually more red laminates in the world than black laminates so the red lamin AK-47 which we expected originally to have a high value is consistently around $4 cheaper than the black exacerbating this effect is the fact that if you have a red laminate AK-47 you're twice as likely to list it on the marketplace as if you have a black so scarcity influence on Price is pretty extreme in fact it's one of the clearest predictors of value that we have there's one more interesting graph uh feature on this graph that I want to call your attention to it's right here where the price of both of these items got depressed this is actually on the date where we shipped a bunch of new weapon finishes including some for the AK-47 so the value of the older finishes um actually decreased relative to the new more unique more scarce finishes we have found over time that as the pool of items increases this effect gets a little bit diffused but when we first Shi the econom it was a pretty obvious drop of a couple of

bucks in the real world quality assessment would be poly sensual kinetic feel and smell would give you indications as to the quality of materials and assembly in our game World our only indications are visual and descriptive but we found to our surprise that they are just as influential we expected that we would influence desirability for one thing we thought the worn beat up weapons looked really cool because we're artists and we think that way um we also expected that the worn weapons would correlate with this kind of Veteran status a kind of fictional bonus on top of the aesthetic interest because of that we considered making our weapons wear over time so that the more you use a weapon the more worn it gets pretty much everybody we showed the tech to had the same idea it was a really popular idea we wanted players to show off their worn weapons as Badges of Honor we ended up not doing it for two reasons the first reason is that some of the paints look better with less wear rather than more or at least that's how we assessed them and so would be really difficult in order to get the uh weapons

to always wear towards an ideal State the other thing is that in order for people to show off worn weapons as Badges of Honor uh they have to represent a significant time investment and that means that the wear could happen too slowly for our players to perceive now we don't want to affect people's inventories in a way that's difficult to communicate and so we ended up not shipping that instead each item has a random amount of wear applied to it at Grant time the wear has a a bell curve so that it's um it's most common to get an item with just a middling amount of wear so the items with the most and the least amount of wear have the most scarcity given our previous examples then we could expect that the items with the most and least amounts of wear would have the highest value if that were true our graph would look like this we have battle scard on the left and Factory new on the right you'd think that there would be an uptick in value uh for both of those just because of scarcity but almost with that exception the battle scard exteriors are most BN exteriors are seen as significantly lower quality than the

factory new finishes this definitely shows that the impact of quality valuation uh affects the price on the marketplace so creating our scratch sythetic did actually give us a broad range of value even if we had originally imagined the value spread exactly backwards so let's finally talk about the Aesthetics of individual finishes Aesthetics in the context of luxury is a consumer's ability to discriminate value just BAS based on visual characteristics we created a broad range of Aesthetics from subdued military Styles all the way to flashy items that were shiny so because the game can have a stealth component to it we actually expected that players would prefer items that they felt gave them a better camouflage so how do we test our assumptions we can't just look at price on the marketplace it's way too heavily influenced by all the other factors on this list it's also worth noting that the price on the marketplace doesn't capture the total value of the items in our economy less than 2% % of our items are listed on the marketplace the rest

of them are collected and equipped and they provide a value to their owners that isn't expressed in dollar amounts so let's look at some non-market data we added crafting to the game shortly after the economy went live crafting is just taking a number of items and trading them in with the game instead of with another player in order to get a different item back we use the fiction of a contract with the Arms Dealer it's essentially a way of trading in with the game controller instead of the player and we use specific contracts that Define which items can be crafted into which other items we've organized our items into collections and we allow players to trade up inside of those collections 10 items from one quality tier will get you a random item from the next tier up we can identify our most desired items by looking at just a couple of factors for one thing we'll look at our crafting data we can see which items are traded in more frequently than usual we can also look at the origin of an item if lots of them are created VI crafting

we're either looking at the item that was desired or we're looking at an item in the same tier as the one that you hoped you got or we're looking at an item that's transitional it's something that someone crafted in order to work their way towards a yet higher tier so if we look only at the items that are created via crafting but not subsequently traded in this gives us a pretty clear picture of the favorite items from this collection we can further narrow athetic preferences by looking at equip rates items that are equipped at an unusually high rate that are also worked towards vac crafting give us our best picture of our most loved items so for example if you own a desert eagle with the Cobalt disruption finish you're 16% more likely to equip this one than any other finish even if you have lots of them so if we look only at the craft targets that satisfy these two qualifications we actually end up with a pretty small list oops sorry so people like anodized finishes a lot more than half the items on this list are either anodized or metallic

they also like saturated colors on black or very bright finishes and they also like weapons that have a Sleek kind of spy movie aesthetic the easiest thing to see though looking at all of these items is that most of them have a single predominant color I should note that this is really different from looking at price on the marketplace although some of these items do have a high price on the marketplace that's not true for all of them in fact some of them are worth only about 50 cents and some of them are worth as much as uh sorry $30 scarcity also isn't reflected very well here some of these items are really rare and some of them there's a lot of them in the world so looking at crafting and equip rates for these items gives us a new window into what players value about our economy content so this covers the high end of our aesthetic range what about the low end everybody starts with the default Loadout as you play you get drops of weapons with finishes the first time you get one you are 9 % likely to equip it at least

that's true for most weapons for some of them even when you get it you dislike it so much you'd rather have the default if we look only at the rejects we can see again a pattern these are all camouflage inspired in fact these are all items that look like our original inspiration so although we started off thinking that the military camouflage was really cool it turns out that what Community really values our finishes that look more like paint guns that was great we needed a reminder that although Counterstrike is military inspired it's not a military simulation it's a sport when our customers play they don't aspire to be soldiers they aspire to be elite Counter-Strike players so maybe it's not that surprising that the closest real world analog we've got to our preferred aesthetic comes from a sport if you looked at our content you'd assume that we understood this right from the start and and decided to put the bright items at the top of our quality tiers because we thought they

had the most value actually that wasn't what we were thinking it was a risk mitigation strategy we were genuinely worried that people wouldn't like the bright items so instead of thinking of these as being sorted by value we originally sorted them by risk after all if you got a rare item you'd like it because it was rare even if it was an aesthetic that you didn't appreciate and for everybody else if they don't have a rare item they're happy because they don't have to see the bright stuff very often because they aren't very many of them in the world well now we know we were wrong would that have changed our design it may have and let me explain how we need to properly map out our design space we can think of the facets of luxury as the axes in that space we figured out the bounds of the axes by examining our data conspicuousness varies with the utility of the weapon scarcity um oop sorry Heritage and personal history from the novel to The Familiar uh scarcity varies with Supply as well as how frequently you encounter an item um sorry a finish for a particular

weapon uh quality varies from our original scratched aesthetic to factory new and our aesthetic style which was the most difficult for us to tease out of the data varies from our original military aesthetic to visually Salient Sports inspired Styles the last thing that we want is for all of these axes to be aligned collapsing our design space it would be possible if we didn't understand how these axes were defined in fact we aligned scarcity and Aesthetics by putting most of our visually Salient items at the top of our collections fortunately our design space has more than just those two axes or we could have had a serious problem the worst scenario is that each item in the economy has all the facets of a luxury good each in equal degrees that doesn't encourage interactions in which players both feel they're getting the better end of the deal someone's always going to be ahead of you on the ladder and somebody behind and there's only one direction to trade in that makes any sense that is the best way I can think of to stifle the economy instead what we'd like is to

have ways to create value in every axis understanding not only what the axes are but how uh to place an item along it lets us determine if we're providing enough variety of value ideally there wouldn't be any items in the economy that had perfectly equal value on every axis having that range of values really important we can create a range of items that have different value on different axes and then players with orthogonal goals can trade with each other and each feel like they're getting the better end of the deal even if the total value of the items is unequal so I could for example want to collect rare items but maybe I don't care if they're bright or shiny or I could want to collect items that are bright colors but maybe it doesn't matter to me if they're Factory new those trades can happen even if the value is unequal and players with different goals can value different items in different ways so the final we slice our axis the more nuanced trading and Marketplace transactions can get by creating our content in this semi procedural way we were able to slice our axes very fine indeed creating a broad

range of value and creating value on multiple axes so with our players exchanging over 30 million items with each other on the marketplace so far certainly seems like we've met our goals so let's wrap up by checking back in with our product level goals we'd hope that the economy would improve the longevity of our game provide fun and value to our customers and make it easier for our customers to provide fun and value to each other remember that we only have to check in on a couple of broad metrics to see if we're meeting our goals it's our player numbers player retention and our individual play times this is taken from Steam graph. net it's a community website that monitors the uh numbers we publish about who's playing what on Steam it shows our Peak concurrent player numbers daily so here's some notable landmarks here the games in beta this is the launch of our game and here are a couple of free weekends this is the Steam summer sale that

preceded the economy launch so after our economy launch we had a pretty dramatic increase in Peak player numbers bringing us to regular highs that we'd only achieved previously by making the game temporarily free this last Spike here is dream hack it's a big tournament for csgo with a high viewership this game was also on uh our game was also on sale for this time period and that caused a huge player Spike nearly doubling our player count there there was a large influx of new players around this time and they stuck around at a really high rate in fact they stuck around at a much higher rate than players from previous sales and so we can see that uh not only are our player numbers up but our player retention is up since that event we've been able to reach a new plateau of about 100,000 peaking current players daily and about 2.3 million monthly unique players in the aggregate it certainly looks like we've met our goals for adding the economy and individual players are sending us the same signal they're playing more fre frequently and for longer which is our best indication

that we're providing them with lots of fun and as they continue to play together and exchange items with each other they're creating loads of value for each other as well our one remaining goal is that we increase the longevity of the game now only time will tell but with such a large community of returning players csgo's future looks really bright I'd like to open the floor for questions now so if you have questions please step up to the mics hi Ron hi can we have a little light in here for the audience likeing you mentioned that there were only two technical artists yourself and one other working on csgo to bring the economy update about how big was the rest of the team and I'm particularly interested uh in uh how the decisions were made to why they proceed or not proceed with doing the economy in the context of how many people there were so we had let me think back around the time we probably had about 12 people on the team total um

that team size has increased since then which is good because it means we can do more stuff um but everyone was was really behind the economy I mean looking at the success of TF2 and Dota 2 we really wanted to improve our player numbers um typically after you launch a product the the product cycle goes like this right you have this big spike at launch and then it just continues downwards until you have this long tail um TF2 and Dota 2 both have this big spike and then they just keep going up and that's the model that we wanted to to go for right so um pretty much everybody on the team was really pulling for the economy it wasn't like we just had technical artists going hey let's make an economy yeah it was a whole team decision I have some uh links that you guys might want to check out just one more question um you have this game mode where you uh unlock um like further weapons the more kills you earn right uh where you like start with a pistol and then you perform your first kill and then you use the machine gun

and whatnot right so there's there's an in-game economy that's a game mechanic that's totally separate from this kind of meta economy and uh what happens is if you make a kill you earn more money in game which allows you to buy a more expensive weapon later on and that that's actually a good catch because that graph that I was showing about weapon utility it wasn't actually showing the rounds sequentially it was showing how much money you had to buy a weapon at the start of a round so it's not exactly what I said that it was but I didn't want to explain the mechanic I meant the certain the special game mode where you like start with a pistol and then you perform one kill you're talking about the the arms race yeah did you like look at stats of players that like only play this particular game mode and how the like purchase of weapons differs uh for for those players in comparison to like death match players or so um so there aren't actually a lot of players that play exclusively arms race um in fact our our players here I totally have a slide for this um if I can get my cursor back from the other screen

um so we we actually have different populations of players let me see if I can find the slide it's way way at the end here um so we have players who they mainly break down into three categories they break down into players who play uh on uh casual players who play on competitive and players who play on community servers and so uh arms arms race is going to fit into the Casual section of our players and we didn't really break it down a lot further from there because the populations got really really small okay thank you okay thanks bronwin um I wanted to ask now that uh csgo is kind of in a different place than when you decided to start doing um uh weapon skin as items if you've got if you guys have considered any of the other uh kind of more resource intensive um expansions to

it um that stuff is always on the table it's just a matter of um making sure that we have a a goal for it we don't want to ship something just for the sake of it we want to work towards actually improving the customer experience and um I mean you might be thinking of things like the the character customization um until we can solve the team ID problem that that wouldn't be something that we would want to ship so there there's a lot of work to do there uh that doesn't mean that we won't consider it okay cool than on your um on your graph there for the dream hack you notice a spike in players staying online now did you guys have the Esports uh line of your economy plan before you noticed that Spike or after oh absolutely before yeah absolutely before so um for those in the audience that don't know we sold a special key for a special case it was called the Esports case um we've actually had a couple in that Series so far and what that case does is we take a

portion of the proceeds from that case and we use it to fund Community tournaments dream hack was actually the first recipient and they had a $250,000 prize pool that was funded by our community and so the community was actually heavily invested in it in order to you know go watch it and see the the results of something that they've been caring about all this time and actually we just had the second one which was ems1 in kavit uh that was just last week um our Peak concurrent player numbers during that event went up to 165,000 so and we were also on sale so we'll see how many of those players come back over the next couple of weeks I'm uh actually kind of excited because I think we're going to reach a new Plateau okay thank you hi uh I know in the past few months you've also shipped an update that allows players to apply stickers to their guns yes and uh one I'm I'm a big Counter-Strike player myself and uh so were a lot of my friends but I found that they're a lot more polarizing than the skins and I was wondering if that was something you guys knew going into shipping that update no we we had no

idea they'd be polarizing we thought people would think they were really cool um there's you have to be careful when you're talking about things being polarizing because there's some data that you can get anecdotally that isn't actually represented in players purchasing behavior and so there's a large community or actually a large community of players that that never speak up right that are perfectly happy with the things that they get and so when we're looking at um things to add to the game we're assessing you know whether something was was valuable and whether it's working we can't just look at what people are talking about in the forums although we obviously do read that stuff all the time and take it into account we also have to look at how um our players behaviors are changing and whether they're actually consuming the content and enjoying it and it turns out that people actually love stickers yeah for the most part anyway cool thank you MH um in CS 1.6 you guys had a spray paint on the world right yes have you guys considered adding U user customized spray paints which are obviously you

guys will have to monitor what goes in and what goes what goes out and the ability to maybe like uh if I didn't like a spray paint and I was playing the game I could uh actually spray something on top of that and kind of hide it and so on what you did you guys consider something like that so um there were a lot of abuses of the original system because there wasn't any sort of gating system you could just put in whatever you wanted and so there were a lot of obscene images to sprayed around the world and actually um my favorite and the worst is that people would take a picture of themselves their in you know weapon POS standing against a wall and then they would spray it on the wall and then other players would come by and start like targeting to shoot there and you know so it's like this this weird strategic kind of emergent gameplay sort of thing um we're not sure how much value sprays have relative to some of the other things that that we could do I know that people do want them back because they're well again it's kind of a Heritage thing people are are used to having them but we haven't um made any decision one way or the other all right

thanks uh hello I was wondering if you could talk a little bit about how knives factored in uh were they always supposed to be the ultra rare uh kind of item from these cases and obviously their effects are just aesthetic so um knives yes we we always plan to make them the most rare items and the reason is that these are weapons with an extremely high utility you use them in every single round of the game you whip them out all the time because whipping the out makes you run faster relative to carrying a heavy weapon right in front of you um we thought that just giving them out um would be would be kind of a waste because there was such an opportunity to to drive value by making them rare and desirable and um we even see Within These Guys that there's a value spread like uh all the different finishes the fade is the most popular and then the slaughter and then the kuram bid is the most popular knife and so there's a good range of value here too um and it's a little bit worrisome because the value of these items is so high actually

for a long time we were watching the price of them just keep Rising on the marketplace which was really scary because we we have a cap on the marketplace uh of $400 and so if you have an item that's worth more than $400 you you can't actually get the value of it from the marketplace and so people end up doing things like taking their knives and trying to sell them on eBay and they get scammed all the time and it's it's not really all that positive for players um finally the prices stabilize and now they're around like 200 bucks so so we feel a little safer but um yeah we we always plan to make those the the highest tier items and I didn't talk about them very much because they don't really fit into any of the collections or any of the other things that we talked about and they have such a high scarcity and such a high utility that they're just off the charts on pretty much all of our other ranges thank you I think that hi oh hi I was curious about uh so you have some weapon skins that are tracking in terms of of kills oh yeah and how does that play into against like the Aesthetics of the

weapon and everything yeah yeah that thing um well so we call these items stat track items they're kind of like the Stranges from TF it tracks how many kills you get so I I think in TF um your strange weapon just tracks it as as part of the name or description and you don't actually see it as like it's not on your weapon we created modules or like especially on the knives they have these little scratched off bits where you you can actually see it on the weapon and people really seem to like that um we have found that the value of these things um there is some value there just for intrinsically they're they're fun to have but most of the value of them on the marketplace comes from their scarcity which is sort of interesting so it's another way that we can generate a range of value um does that answer your question or was there thank you cool um so you have uh Dota 2 and TF to that are both free to-play games yes so when you guys develop the

economy are you able to say what some of the tension that you were running into maybe making CSCO free play or not um that that's a question that we get a lot um the one thing that we're concerned about that's stopping us from going free to playay is cheating right now there's a friction if someone gets caught cheating they have to Reby the game in order to come in and keep ruining other people's experiences uh if the game's free to to play there's no friction at all they can just keep rolling over accounts and so until we solve that that problem um we probably wouldn't go free to play but it's it's something that we talk about all the time thanks do you make the same kind of uh assessments when choosing uh map map Workshop levels oh yeah absolutely yeah um there there are different things that the the different Maps bring there are some maps that are just abs absolutely gorgeous and they they draw people into those worlds and there are some maps that are

you know slightly less gorgeous but they play really fantastically what we do find is that the um the value of the maps to the community uh is almost all about play it's like if they play really well so uh you guys if you're Counter-Strike players have probably heard of cash um it's a really popular map it's a community created map that has um a really good play style to it so that one's really popular compared to some of the other maps that might be visually more attractive um but they aren't as popular because the the gameplay isn't quite as strong all right I think we're out of time so thanks