E45: Autobattler Econ, WILD UGC Algo & A Currency Debate for the Ages (w/Arto Huhta) Autobattler Econ, WILD UGC Algo & The Big Currency Question (w/Arto Huhta)

November 23, 2025 • 01:13:44

Show Notes

What happens when autobattlers fail to monetize? We pull Arto Huhta into the cast and chat about Telegram’s pseudo-WeChat ambitions. Eric releases a distrack on Game Designer's obsessed social spaces, and Phil wants more blood from psychologists' nonsensical F2P "choice overload." Chris enleashes a model-meets-UGC experiment: a three-algorithm simulation that shows how recommendation systems distort consumer welfare and creator inequality.

\We discuss:

Chapters 00:00 Journey to London: A Game Developer's Path 00:49 The Role of Economy Design in Gaming 01:20 From Academia to Game Development: Bridging the Gap 03:16 Experimentation in Game Design: Lessons Learned 05:22 The Intersection of Game Design and Economics 10:07 Understanding Game Development Roles 11:00 Monetization Strategies in Game Design 11:55 The Evolution of Publishing Models 12:42 Transitioning to Web 3: New Challenges 13:54 The Economics of Game Spending 18:27 Introduction to Game Economist Cast 19:06 Current Gaming Trends and Preferences 20:51 Game Modes and Player Engagement 22:03 The Future of Game Monetization 27:33 The Social Hub Experiment in Fighting Games 28:26 Street Fighter VI and Social Interaction 30:28 The Rise of HTML5 Games on Platforms 32:37 The Trend of Casual Games in Tech Companies 34:42 Telegram Games: A New Frontier 37:21 Challenges in Game Discovery on Telegram 38:52 User Engagement and Retention in Web3 Gaming 39:43 Consumer Welfare and Content Creation Dynamics 43:04 The Impact of Algorithms on User Experience 49:31 Heterogeneous Goods and Their Effects on Engagement 57:35 The Impact of Algorithms on Content Quality 59:04 Understanding Algorithmic Risks and User Retention 01:00:16 Exploring Algorithm Design in Gaming Platforms 01:01:54 The Role of User Choice in Content Discovery 01:04:29 The Future of Pricing Strategies in Free-to-Play Games 01:08:10 The Debate on Standardization and Market Forces

Chapters

View Full Transcript

Episode Transcript

[00:00:00] Speaker A: What part of London are you in? [00:00:01] Speaker B: Southwest. So we've got the US Embassy over here. [00:00:04] Speaker A: It looks like a compound, doesn't it? There's like cameras everywhere on barbed wire. [00:00:08] Speaker B: Yeah, yeah. [00:00:09] Speaker C: Is that a recent addition or. Or has it always been like freedom, baby? [00:00:12] Speaker A: It's there all around the globe. Every single American embassy Looks like a compound. [00:00:17] Speaker D: Yeah. [00:00:18] Speaker B: Do you have like the evil Empire aesthetics? It's. It's bizarre. [00:00:22] Speaker A: Hey man. Freedom. [00:00:23] Speaker C: Yeah, it's. It's the cost of freedom, baby. [00:00:26] Speaker B: It's a nice area though. [00:00:27] Speaker C: The cost of freedom is restriction and non freedom. Let's start with utility. [00:00:32] Speaker A: I don't understand what it even means. [00:00:35] Speaker C: Everybody has some kind of utils in. [00:00:37] Speaker D: Their head that they're calibrating. [00:00:39] Speaker A: There's hardly anything that hasn't been used for money. [00:00:42] Speaker C: In fact, there may be a fundamental problem in modeling. You wouldn't want to model. [00:00:49] Speaker A: Episode 44 Game of Connors Cast the boys are back in town and we brought a plus one from the old country from all the way in London. It is Arto. Arto. Welcome to the game. Economist cast. We are all longtime fans. We follow you on Twitter. I've been trying to get you on the show forever. You've resisted my advances and I went through your co founder. [00:01:10] Speaker D: No, he. He just said yes. [00:01:12] Speaker A: You got a long distance. You know how long I've been trying this? [00:01:14] Speaker B: Well, either way, great to be here. [00:01:16] Speaker D: Tell us about yourself. All I know of is that you pop into discord occasionally. [00:01:20] Speaker B: So I think the best way to profile me is as a free to play product person who does do a lot of economy design. So I kind of reject that job as a job title. But it is a lot of what I do and how we got here is I started off as an economy designer in Games in 2013. I still do a lot of economy design. It's kind of branched out to other things. One of the things I've enjoyed a lot throughout my career, I think some overlap I assume with Phil is that I've had a lot of like cross portfolio roles, worked with a few third party publishing teams. That's. I really love that because like you get to work with a lot of teams and jump across different genres and work in one multiple products in parallel. That's been cool. More recently, three years ago we started our own startup. So that's been a change in pace basically. [00:02:06] Speaker D: Yeah. How many people imagine it's a big change. Yeah. [00:02:09] Speaker B: We have a nine person team. [00:02:10] Speaker A: There's web three involved. [00:02:12] Speaker B: Yeah. So I think like what makes our venture Interesting. Primarily two things. We're doing alternative distribution, which means anything but the app stores essentially. So we do target a lot of like mobile play patterns, but there's no app to download when you play. Our game is either directly through browser. A lot of it is on mobile, can be desktop, but also telegram, which is that effectively it's an embedded browser environment. We expect more of these things to pop up where you have like players entering games through these super apps. And the other part is the web. [00:02:43] Speaker A: So you said you rejected the title Game Economist. You have a master's degree in econ, you've done a thesis, you've done like nuts and bolts econ. You said you've done economy design. I know that personally because I interviewed at Flare Games at one point, your former employer with you in 2018. Why reject the term? [00:02:59] Speaker B: It's not that it's not street credible enough. I feel it's pretty narrow for at least what I tend to be interested and focus on. So I would say I have a lot more overlap with even like game designers and product managers than necessarily like traditional economics. [00:03:13] Speaker A: Take, take away the responsibilities. What's in your heart? [00:03:15] Speaker B: Game development. I think that's. Yeah. [00:03:17] Speaker A: Do you think you. Are you still using like your, your kind of more neoclassical toolset in games? [00:03:22] Speaker B: If we're honest about it, I think the biggest, most important tool I bring from my studies is just the math, like the number crunch. [00:03:32] Speaker A: So when you say math though, is this spreadsheets? Is this building models? [00:03:35] Speaker B: Yeah. [00:03:35] Speaker A: Where's the margin? What about marginal analysis? You're not using that every day, everywhere and always. Right. I mean, can you get it out of your brain? [00:03:41] Speaker B: Yeah, I feel like my first year in games, I think I used like something like Cop Douglas production function for like a combat simulation. So that was like direct application on a very weird setting. [00:03:54] Speaker D: But yeah, other than that, I use math every day. I'm not a game mathematician. Right. [00:03:59] Speaker A: I mean as an economy designer though, aren't you thinking about the world in terms of incentives? [00:04:03] Speaker B: Yeah, absolutely. [00:04:05] Speaker A: So how would you think about the world as a, as a game economy designer? How do you think that differs from being a game economist, if at all, or just like a traditional neoclassical economist? [00:04:13] Speaker B: I mean, I think it's fairly arbitrary job titles and responsibilities. [00:04:18] Speaker C: Especially in a startup. [00:04:20] Speaker B: Yeah, especially there. I mean, I guess that's an interesting one because like my prior role before a nine person startup was at King, so specifically Candy Crush was the biggest team I've worked on by far. So that's a big, really big change. In pace. And the level of or like the degree of specialization is extremely different. [00:04:41] Speaker A: What is something that you think you've seen now that you've taken like all these different roles, job titles? You know, you mentioned product management, economy design, the master's in economics. Do you think there's an insight you've been able to have that you don't find that other people around you have? [00:04:55] Speaker B: In the context of game development, this is I think, very much team sports. So you can kind of mix and match and work with different disciplines while speaking the same language. I think that that is critical. So that's why I kind of like don't like that the job titles are too narrow. I do think there's a risk for like econom designers for kind of getting too, too stuck on the simulations and modeling and the spreadsheets because you kind of need to take your eyes off of that stuff occasionally as well. [00:05:22] Speaker C: It sounds like you would say there's a pretty big difference between an economist, especially in the neoclassical sense, and an economy designer, especially for video games. You identify more as an economy designer than an economist. Or is there no distinction? [00:05:37] Speaker B: Yeah, I mean, maybe you can tell me how you see the distinction there. I think like economy designers, I think in practice I think it's a lot more execution and being a practitioner operating with live op tools and just getting like the game development done sometimes just like fast tracking things, unblocking other people rather than spending time on getting things perfect. [00:05:58] Speaker C: There's like, there's like three different jobs. One of them is a hybrid of the two. One of the jobs that I have, for example, I work in web3 games as well, is like pure design. Hey, we have this brand new system that we want to introduce. We have no data, we don't have any precedent. We're going to create a new combat system and we want to know how it plays out. So I'll be, maybe I'll do it in spreadsheets. I typically do it in Python. It's a little quicker for me, especially now with AI and stuff like that. I spin up a simulation in a matter of minutes. So, you know, I think about that as like the design, the economy designer. No data tools, no data science there. And then I have pure like econometrics, which is X post. I've collected a data set, I've got some sort of data set put together and I'm going to study an event or a phenomenon. And that's like my. More like my classic econ toolkit that I learned in grad school. And then there's like this beautiful combo, this middle space that is both, you know, trying to improve the game, trying to update the game, trying to build a new system, but it utilizes data. And that's kind of like where I see the water gets muddy. Right, because you're not just purely designing an economy in a spreadsheet. You're using data and you're using that data to predict things. It's almost like structural modeling in economics where you've got like this theoretical model which is the design side, but then you feed data into that model in order to effectively calibrate it or parameterize it. So it sounds like you, you do a lot more on kind of like the design side or at least you enjoy doing that. [00:07:33] Speaker B: Yeah, absolutely. And obviously there are focus shifts depending on the product lifecycle. So there is a point in which there is no data to use as inputs as well. If you're doing early stage pre production, then yeah, you kind of tend to use the same toolkits to rapidly like test ideas. [00:07:51] Speaker A: What about when it comes to experimentation? Not necessarily the econometrics piece, but you know, on mobile games we're able to do a lot of A B tests. Do you sound like you have insights that you wouldn't have otherwise had due to the experiences you've had, not just as a game economy designer, but being in kind of the more neoclassical sense? [00:08:07] Speaker B: Not necessarily. I think it's more kind of like. For that, I think King was a very interesting place to work at. I think the emphasis on measurement and experimenting is quite heavy and there's a very good reason for it. Right, so the playing companies playing like two major games. One is like accumulating the genre specific knowledge and second is trying to improve Candy Crush makers. [00:08:26] Speaker C: Right. [00:08:26] Speaker B: That was super fascinating working in that environment where it is a lot about experimentation and optimization as well. I think I did learn a lot from there, but that is a very specific setting. Not a lot of companies have the luxury or the incentives to go that heavy or even try to go that heavy. I think in a lot of the cases, especially like you're going doing zero to one game development, the data is just not there. There's a lot of questions you need to answer with either weak signals or lack of data. Right. [00:08:55] Speaker D: You made an interesting comment though. You said King had two goals. One was genre specific knowledge and second was improved Candy Crush. But they've got like a bunch of these like smaller match 3 games. Are those basically just like experimentation grounds for Candy Crush? They're like lower risk environments to test new theories. [00:09:11] Speaker B: Yeah, I'm simplifying a bit, but I suppose that's what you would do as well. You would use the product overlap is really huge and I think some companies benefit from it a lot if they can leverage. I think King is a really good example of. [00:09:24] Speaker D: But yeah, it's interesting the idea that testing a theory, a new design in, you know, game B, the value of that is actually applying it to Candy Crush more so than improving the game you're testing. [00:09:34] Speaker B: Yeah, because a lot of those learnings do carry over, which is the benefit of having the overlap there. Whereas, like, I think, like, I guess like the stereotypical like product manager toolkit of trying to like, apply learnings, like trying to generalize everything to the point that it gets like, almost like I've got this menu of seven different features. This, this one is for engagement, this is for retention, this is for monetization. That, that works to a degree, but kind of like where that gets more artful is in the application. You can just like plug in a daily login bonus and expect the same results across genres. [00:10:07] Speaker A: So when you were. You were at Flare Games, was your title monetization manager, monetization expert. How do you do you look at that differently from economy design? Do you think those are different aspects, different muscles? [00:10:17] Speaker B: I think that the job title wasn't very descriptive of what we actually did. So like, I think we actually do the general product style. So it is. It was a lot of like economic design, but also things you would expect any product manager to do. So live ops analytics, preparing games for Live ops, things of that nature. [00:10:37] Speaker A: Did you have the opportunity to work on Nonstop Night? [00:10:39] Speaker B: Yeah, that was a priority project for me or maybe like a year now. That setup would have worked as. Or how I found myself working in numerous companies is kind of like spend 70% of your effort on your priority game and the other 30%. Kind of like going from team to team, trying to kind of consult or help them where you can. [00:11:00] Speaker A: Did you sell me the starter Dragon bundle? Was that you? Yeah, I monetized very quickly. I gave you the $5. And I remember if you had an update, you basically changed the whole game into like a pet direction. [00:11:11] Speaker B: That's right. The story there was the game was very sticky. So like, it was a very quick production idea to soft launch to really promising metrics. Monetization wasn't one of the very promising metrics. So what we did was we implemented a starter bundle, saw really good results, and then kind of spun out A whole like pet collection system around it. [00:11:30] Speaker A: Why do you think pets worked? [00:11:31] Speaker B: So I guess like in that format of Idle rpg, a lot of the progress is temporary. Pets were the permanent layer there. So even if you reset, you still keep your pets and that's kind of like plugging a hole in the economy. [00:11:45] Speaker A: Otherwise Royal Vault 2, I remember that you guys were running that game forever. I mean there were some, there were some great hits that Flare Games had. What do you think the downfall was around what do you guys think you got wrong? [00:11:55] Speaker B: I think the publishing model lived quite a lot throughout the years. Initially we picked up games very early in their life cycle. So it was more of a co development model which is more risk, more capital intensive and that over time drifted more towards like later stage games. So very more metrics driven and games that were ready to ship and more volume. Meanwhile, the King Publishing team, which is something that very few people know existed, had a very similar model to early stage player games which was work with a small number of teams and get involved early. [00:12:30] Speaker A: Now that, that makes a lot of sense and you've, You've moved to Web3 in your most recent role. How has that changed the questions you ask? What are things that you have to consider now in Web three that you didn't beforehand? [00:12:42] Speaker B: So from just like a purely economy design perspective, there's a lot of lot of new ground to cover there. I feel like the kind of like consumer psychology aspect of it is, is one thing. So how I feel about this is like I think free to play to begin with is quite unintuitive. I'm sure there's a lot of stuff we were kind of surprised by initially and then forgot when we kind of got deeper into it. But I do remember a lot of like moments where you look into like data from heavy spenders and you're just like doubting that is true. I remember one colleague asking me if we should just go ahead and remove like the 99 packs from the store because they're just taking up space and surely nobody buys these things. So there's a lot of unintuitive elements to the, to the way free to play games monetized with microtraxactions. We're basically like it's very outlier heavy. And it's hard to empathize with those outliers because practically speaking a lot of us might know like one or two whales from our personal life. But those are, that's a, that's just the tiniest sample you can imagine. And those are maybe like whaling In a very specific way, in a specific game. And maybe that doesn't translate. [00:13:46] Speaker C: You're looking at a whale right now, Phil. [00:13:48] Speaker A: 10,000 in spend, man. And last year I did my. I did my tax receipts. I'm. I think I'm proud of myself. I'm not quite sure. [00:13:54] Speaker C: Tax deductible, baby. [00:13:55] Speaker D: Yeah. How much? I was, I was going to ask how much is the marginal tax? You know, the difference affecting your spend? Like, how much would you have spent if it was not tax deductible? [00:14:02] Speaker A: So I mean, what, you mean like when you run it through the funnel? [00:14:05] Speaker D: No, like, you know, your effective cost is like slightly lower. [00:14:09] Speaker C: Right. [00:14:09] Speaker D: What's your price elasticity? [00:14:11] Speaker A: So, I mean, it's about 50% more if I wasn't able to build this as a business expense than if I took it as a personal expense. [00:14:18] Speaker D: So basically everything's half off? [00:14:20] Speaker A: Yeah, pretty much. [00:14:21] Speaker D: Okay, and then how much more do you think you spend? Because everything is half off. [00:14:25] Speaker A: I think I probably would average around a thousand if I didn't do this as a business expense. [00:14:29] Speaker D: Probably 10,000 a year. [00:14:31] Speaker A: Yeah, I spent 10,000 this year. [00:14:32] Speaker C: Your price elasticity is. [00:14:33] Speaker B: What is that? [00:14:34] Speaker C: Wow, that's like five or something. [00:14:35] Speaker D: Yeah. You 5x by cutting the price. [00:14:39] Speaker A: Yeah, yeah. [00:14:39] Speaker D: I mean, does that suggest that these games are overpriced and they would profit more if they charged lower prices? [00:14:44] Speaker A: No, no, no, not at all. [00:14:47] Speaker D: But they did for you. [00:14:48] Speaker C: Quantity discount. It's quantity discount. [00:14:50] Speaker B: Right. [00:14:50] Speaker A: Thank you. That's it. That's exactly it, Chris. And the other thing is, you got to do special reporting for crypto, by the way, because remember, you're buying an asset that has tradable value. So like when I bought, remember Crypto, Unicorns, everyone. [00:15:01] Speaker B: Yeah. [00:15:03] Speaker A: That was supposed to be like a real Web two game team making a game that went absolutely nowhere. I was like on Ethereum, then I had to bridge it to Polygon. My God, Web three. And I had to do some special reporting for that 20 thing. And I was like, never again. [00:15:16] Speaker D: Yeah, wait, hold on, hold on. [00:15:17] Speaker B: Sorry. [00:15:17] Speaker D: I want to go back to the price point. You get a half off or double quantity discount and you spent five times as much. And you're. But you're saying that these games would not be more profitable if they offered, like, lower their prices. [00:15:30] Speaker A: No, because you also gotta. Because you also gotta remember, like, we're not holding all this constant. In my scenario, like in my scenario it's a business expense because I actually have to write about it. Like there's all these tasks that Are that are benefit. Like if we held all else constant, I'd probably only spend a thousand a year. Like if I weren't able to take it as a business expense. But I was still writing about games, doing all the normal things. [00:15:49] Speaker B: Right. [00:15:49] Speaker A: So you only. [00:15:50] Speaker D: Do you spend 10. Yeah, let's keep all that the same. You're still a game economist, consultant, you're still doing this for work and you still, it's still one of your hobbies. The half off from the taxes is causing you to spend ten times as much. [00:16:03] Speaker B: Right. [00:16:05] Speaker C: So that is the price elasticity of 5. [00:16:08] Speaker D: Yeah. So like, shouldn't they reduce prices? Assume the tax incentive didn't exist. Wouldn't they be better off reducing prices for you? [00:16:16] Speaker A: But if I wasn't doing this career. [00:16:18] Speaker D: Assume you assume you're doing the career, there's just no tax, then I would. [00:16:21] Speaker A: Probably spend a thousand. I would probably spend a thousand. That is absolutely true. I'd be more expensive to price. Excuse me. If the tax were higher. [00:16:27] Speaker C: No, no, no. Okay. So my understanding is you spend $10,000, you get a $5,000 rebate from the government for spending that money. 50% price discount. Ish. [00:16:36] Speaker A: Remember, remember, we're playing fast and loose on how the discount happens. So first of all, there is an explicit payback, which is the vat. So remember in Europe, when you're business and you go out and make a purchase, VAT in Sweden is 25%. So if I spend, if I spend $100, the government actually writes me a check for two, $25 every quarter. Like you get something called a VAT check. So there's that. But then you got to REM. Way we're framing it is pre tax versus post tax dollars. [00:17:04] Speaker D: What's the effective discount like? Like, forget the mechanics. What's the effective discount you're getting about. [00:17:08] Speaker A: 60% when you include VAT and you include basically all the post tax stuff. [00:17:13] Speaker D: So remember, compared to if you could not write it as a business expense. Let's call it half off. Let's call it. What's, what's wrong with let's call it half off? [00:17:19] Speaker A: Let's call it half off. [00:17:21] Speaker D: So you're saying there's two parallel worlds. In both worlds, you're working as a game economist. In both worlds, you're playing these games, doing some research on them. You're saying with the half off you spend ten times as much than without the half off? [00:17:32] Speaker A: Yes. [00:17:33] Speaker C: Prices elasticity of 5. [00:17:35] Speaker A: So yes, yes, I will concede that. But that's for me, that's for me though. [00:17:39] Speaker B: Okay. [00:17:40] Speaker D: Yeah, okay, that's fine. Do you think this generalizes? [00:17:42] Speaker B: Yeah. [00:17:43] Speaker D: Why are you so. [00:17:43] Speaker A: I don't think this general, but we have the evidence for this, right? It was the King Paper, right, where they did massive quantity discounting with Steve Levitt, David Nelson. [00:17:52] Speaker C: Well, what this tells me is that you should quantity, you should discount, but at the part of the distribution where it matters for whales, right? Like you should identify your biggest whales. I mean, this is what Steam does with their marketplace fees. If you're. If you make what, over $10 million, your fee goes down to 25. [00:18:08] Speaker A: Oh, the developer fees. Yeah, yeah. [00:18:10] Speaker C: If you make over 50 million, you. You get charged 20%. It's kind of the same thing. But their price elasticity. Price elasticity is much higher. [00:18:19] Speaker A: But that's not for the consumer. [00:18:20] Speaker B: That's. [00:18:21] Speaker A: That's B2B. Right. That was a. That was a competitive Europe. [00:18:24] Speaker C: You're a business. [00:18:27] Speaker A: We have an exciting episode today. We're going to be talk. [00:18:30] Speaker C: Just going to talk about me. That's. I threw together an AI slop generated slideshow based off of a blog post that I wrote about different. [00:18:41] Speaker D: I screen share it. I want to see it. [00:18:43] Speaker C: Well, I mean, we're just doing the taste and then we'll do the whole thing later on, basically computationally exploring different types of reinforcement algorithms in UGC platforms. So think Instagram, YouTube, Roblox. [00:18:56] Speaker A: Before we do, let's talk about what we've been playing. [00:18:59] Speaker B: Got a selection of good things on sale. [00:19:02] Speaker C: Stranger. [00:19:03] Speaker A: Mr. Mr. Arto, what have you been playing these days? [00:19:06] Speaker B: I keep going back to Hearthstone Battleground, so that's been a really sticky one. Played a bit of merged tactics in Clash Royale. That's interesting. [00:19:15] Speaker D: I've been playing that a lot too. [00:19:17] Speaker C: Do you like card games or do you like these specific kind of mobile implementations? [00:19:21] Speaker B: Well, there's a bit of overlap with both of those with the game we're working on. So I think that's kind of like subconsciously part of the reason. But yeah, Hardstock Battlegrounds. It's just like a very, very nice multitasking game. [00:19:33] Speaker D: What do you think of this auto battler space? It seems like there's a bunch of, you know, like super auto pets, Once upon a galaxy. You know, there's that like team fight tactics. Yeah, team fight tactics. [00:19:42] Speaker A: Fight for the Golden Spatula. [00:19:43] Speaker D: It seems like there was a big rush for it and like none of them seem to have figured out how to monetize super. [00:19:48] Speaker B: Well, yeah, I mean, Clash Mini had a go at it. Right. So Supercell, kind of the thing where we kind of tried to do in our project as well is you do take the parts that kind of don't work. You have a very intimidating onboarding experience. Something like team fight tactics. You've got what is it like 120 characters to figure out right from the get go. And the other thing is it's no persistent economy. Right. So they put it tech building. They make you play with our own content, which means that you're learning the mechanics like one unit at a time as you unlock them. I think that that is no longer with us, that game in a way, but I think that's a. That's a really good, good evolution on a mechanic that clearly sticky. But like as you said, like kind of creating like a very strong free to play product out of it. What would seem one. Right. [00:20:35] Speaker A: So when you think about like Hearthstone adding this battlegrounds feature and Clash Royale adding this tactics feature, there's like this idea of game as a platform where you're kind of like stapling on these modes. Do you think that strategy makes sense or it just becomes like a development fucking nightmare? [00:20:51] Speaker B: The latter in general terms. Right. But Clash Royale is an interesting one because, well, it's a bit of an outlier in some. Some respects it's another game that very few have like replicated successfully any parts of it. Like the direct spin offs, they've all failed, more or less. Clash Royale, if you imagine that as if you imagine a game with a Clash Royale economy but a PvE setting, what you would like typically too is you would try to push game modes and different challenge content. Pretty hard to. To try and get people to rotate their decks and collect a wide range of stuff. And my impression, correct me if I'm wrong, but Clash Rail never really pushed that really hard. They haven't gone very deep into the deck rotation direction. [00:21:29] Speaker D: Yeah, I've heard they've. They've tried. Been largely unsuccessful and then they kind of decided it wasn't worth the squeeze. [00:21:35] Speaker B: It's maybe just a uphill battle, but now they've got a whole new game mode that like effectively does that for them. Right. And the deck size is kind of like. Effectively the deck size is a lot bigger in this new game mode, which is both a positive and a negative one. From an economy perspective, it does make sense. Right. [00:21:50] Speaker D: It almost seems to me these three are all just copying League tft where League stapled TFT on due to other constraints which we could. I can talk about, but it seems like the rest of them were just like, I don't know, we got this content, let's just staple it to this existing platform. [00:22:05] Speaker B: Right. Battlegrounds is. Battlegrounds is an like, reusing the same content in any. Like. [00:22:10] Speaker A: Well, still the cards. Right. [00:22:12] Speaker D: It's Hearthstone Launcher. It's like a separate mode in the Hearthstone menu. Right. [00:22:17] Speaker B: But it's not. It's not your cards, and I think that's a pretty important distinction. [00:22:21] Speaker D: Well, TFT doesn't use your champions either. [00:22:23] Speaker B: Right. [00:22:23] Speaker D: It's just this mode that's inside of the League Launcher. [00:22:26] Speaker A: I'm surprised we don't see more games do this. Like, I mean, you can kind of say first person shooters do this. Like you talk about dmz, Call of Duty Zombies, like kind of this game on a platform. I'm surprised we just don't see more of this because the marginal cost is so low. And you solve for user acquisition. Right. [00:22:41] Speaker B: Is it trending in the casual space though? Like all kinds of minigames? [00:22:44] Speaker D: Hyper casual kind of ran the opposite direction where every mini game was its own app. [00:22:48] Speaker A: It is, but I don't think it's the same arto, because I feel like these are segmented content pipelines where like, you're going and selecting zombies, you're going and selecting tactics, you're going and selecting, you know, the mini mode. And with casual, they're basically integrating it into the linear flow. Right. Like you're playing a match three level and it's like save the queen, do this random task. It's not like an alternative pipeline. [00:23:11] Speaker C: Yeah. [00:23:12] Speaker D: Like you could go in to open the league client and only play tft. You never interact with the. It's not like woven together the way it is. [00:23:19] Speaker B: Yeah. My understanding for Hearthstone is that Battlegrounds is extremely popular, so that sucked up all of the attention in that platform. [00:23:26] Speaker D: I've heard it's also been a big money loss because it doesn't monetize nearly as well. [00:23:30] Speaker A: How are they monetizing? [00:23:32] Speaker D: They're kind of not. I think they're just struggling. [00:23:34] Speaker B: There's rerolls and there's cosmetics from my understanding, because personally I think I might have like 600 hours into it and I just don't know. I've never had an impulse to even care about. Yeah. [00:23:46] Speaker D: Yeah. [00:23:47] Speaker A: How would you change it to start to make money again? [00:23:49] Speaker B: I think the rerolls is a really good idea, but that's really recent, so it took them multiple years to put it in. If I'm. [00:23:55] Speaker A: So when you say rerolls. [00:23:56] Speaker B: Yeah. So you have a lot of different hero characters and I think people have strong preferences on them. So rerolling is attractive on paper. I just never tried. [00:24:06] Speaker D: So you re roll at the hero select. Like at the very beginning you're showing three heroes you can pay to reroll. [00:24:12] Speaker B: Or like expand the choice. I think they've kind of iterated on it so that at some point they had a premium package that then just gave you three options instead of two. [00:24:20] Speaker D: Have you spent on it? Has it been compelling for you? [00:24:22] Speaker B: No, no. I think the more the rerolling in practice is probably like just avoiding the characters you don't want to play with rather than just going for the good ones. [00:24:30] Speaker D: But yeah, I think this, this genre is going to struggle with monetization. It feels like a huge hurdle for them. [00:24:35] Speaker A: Why can't you just make it a deck? Why can't you just make it a deck that you build in? [00:24:38] Speaker D: You can. Once Upon a Galaxy does that. I'm pretty sure I ain't spent shit in that game because I feel the fact that you're drafting from a set means that the additional value, the additional compelling. Oh, here's a new exciting card that I want to play with. It's so much less direct. You're like, you don't get to play with the card immediately. You put it into a random pool that maybe you'll get in the next two hours of play. Do you know what I mean? It's not the same. [00:25:00] Speaker A: But that's like any card based game is like you take a deck, it's a random collection of cards and you're only going to play with the cards you draw though, right? Yeah. [00:25:07] Speaker D: But you can build your deck around a specific idea much. You get more immediate access to the thing you're trying to play with in a constructed deck than a draft. [00:25:17] Speaker B: Yeah. Deck size matters, right. So if you have a smaller deck, like what do they have in Marvel Snap 8 or something like that? [00:25:22] Speaker D: Yeah, it's like 12, so it's pretty small. [00:25:24] Speaker B: If you're. I think if you're 8 to 12 range then individual swaps matter quite a lot. So there's that. [00:25:31] Speaker A: And you need to go vertical at that point too, right? Potentially, yeah. You can go straight on with Clash. [00:25:36] Speaker D: Mini is kind of doing the vertical thing. There was Arcane Vortex or something. [00:25:40] Speaker C: I was. [00:25:41] Speaker D: There's some game out of Vietnam that has like, basically looks like World of Warcraft themed, but they went vertical as well. It kind of with the. For me, it, it turned me off the game because I was like oh. [00:25:50] Speaker C: I'm just losing because my numbers are. [00:25:51] Speaker D: Lower, but I think you could try going vertical. They added a bunch of meta progression systems there where, like your heroes level up and you get special talents and there's like a skill tree. [00:25:59] Speaker A: Eric, have you been playing another roguelike? Which one is it? Wait, hold on. 2K XO is in early access. [00:26:05] Speaker D: Yeah. Riot's fighter, Xko. [00:26:07] Speaker A: All right, Xko. [00:26:08] Speaker B: Sorry. [00:26:09] Speaker D: It's his Alphabet soup. The kids call it Tuco because, like, whose mouth? [00:26:14] Speaker B: It's a mouthful. [00:26:15] Speaker A: Yeah. [00:26:15] Speaker D: Riot made a new tag fighter in League of Legends characters. And for those who know, Tag fighter is like a fighting game where you can switch between characters. Marvel vs Capcom is the big famous one. And it's like, as you can expect, people are switching characters all the time. Calling assist. It's crazy. It's chaotic. There's a bunch of happening on screen and it is extremely inaccessible to casual players and their player bases, unfortunately. Because the game is really fun and it's pretty and I. I enjoy playing it. But yeah, it surprisingly has had very little broad appeal, which is unfortunate. [00:26:44] Speaker C: So is this like a side scroller or is it three? [00:26:46] Speaker D: No, no, it's like a. Like think Street Fighter, but like just cranked up to 11. They used league of Legends characters. They did some interesting. So the team developing as a bunch of hardcore fighting game people, and I think they did not want to compromise on the gameplay. They wanted to make a game they wanted to play. So they tried two interesting. They basically tried to make it more social. It's in the game X Ko, like double Ko. You can play with a friend. Like I mentioned, there's two characters. Let's say you've got like Ahri and Yasuo on your team. You know, Chris can control Ahri and Phil can control Yasuo. And there's only one person in at a time, but you can like tag in and out and like call assists and stuff. So you can make like a. Basically we're like, can we make it into a 2v2 fighting game where there's actually two people on each side? Make it social. Hopefully you get social virality and like social stickiness from that. And then the other kind of gimmick they tried was mandatory social hub. So, you know, a lot of these fighting games have been introducing like a virtual arcade. [00:27:41] Speaker B: Yeah. [00:27:42] Speaker D: Where you like, you run around, there's all these arcade machines and you've like. There's another avatar who walks up to the arcade machine and you sit down together and play together. So, yeah, basically they were trying to go into the social, free to play route rather than innovating the game design itself. The game design is like, very much a core hardcore fighting game. [00:27:56] Speaker A: Didn't Street Fighter 6 go in this direction where they're starting to play with MMO elements? [00:28:00] Speaker D: That's right. Street Fighter 6 and Guilty Gear before them had this kind of social lobby where, you know, again, there's like a virtual arcade. There's like machines and you can walk around and chat and like, if you can have your avatar, wear a funny costume and dance. [00:28:12] Speaker C: Right. [00:28:12] Speaker D: I think it has been pretty successful in selling skins for Street Fighter. One notable difference is in all those previous games, it was optional. Like, you could just go straight into Quick Match where you. You press a button and it throws you in the matchmaking queue instead of walking around this virtual social hub. But 2xko makes it mandatory. Like, you have to. All matchmaking goes through with this social hub where you walk around this arcade and I don't think it's working. Nobody ever talks to me like. Like there's all these people running around, but no one's talking to each other. No one's saying, hey, that was fun. You want to be friends? You want to duo up and play together? Do you want to add me? [00:28:42] Speaker A: And let's. [00:28:43] Speaker D: Let's have some training runs. You know, nobody ever does that. People just run straight to their match and then straight to their next match. [00:28:48] Speaker C: So the way that Fortnite does it is pretty good. I mean, nobody likes a loading screen, but you're basically locked into a room with a bunch of people during, you know, the matchmaking process. And it's like, I'm always running around looking at people. I only am able to communicate with my gun. Well, I think I could type, but I never do that. [00:29:06] Speaker D: You just jump up and down in place. Everyone knows that. [00:29:08] Speaker C: Yeah, I mean, it feels. That feels much more social than like, Ahab. I always. I always hated that kind of stuff. [00:29:14] Speaker D: Well, what's the distinction there? Because they're both. You got an avatar and you're running around in a space with nothing else to do. [00:29:20] Speaker C: Well, I mean, like, in the 2xko model, you're. You can spend as much or little time in there as possible, so you're going to minimize the amount of time, versus in the Fortnite model, you're like forced to sit there for 30 seconds to a minute. You have no other choice. Your only purpose is to socialize. [00:29:37] Speaker D: Is there voice chat, proximity, voice chat? [00:29:39] Speaker C: There might be. I've never heard it before. I don't think that it's a standard feature. I've never heard somebody's voice on there. There was a. There was a souls game. I think it was either a souls game or a Monster Hunter game that kind of did something like this, where they had a more social component. And it just pissed me off. I just wanted to get out of the social hub. I wanted to be able to do my own thing and just get right on to the next quest. But, you know, and that's like a game. I think it was Monster Hunter, and that's a game where there's an immense amount of nostalgia, similar to League of Legends. And the social aspect is huge in Monster Hunter. It always has been huge. It's been a very social game since the old, you know, the days of PSP when you go and play with people locally. So anyway, it's just kind of interesting that you can have two similar implementations of the same thing and get totally different results, because Fortnite makes a lot of money off the skin. [00:30:29] Speaker A: Chris, do you want to tell us about the indie game you've been playing? [00:30:31] Speaker C: Tunic Turmeric? I think I played that game three years ago, actually. You know what? I was just going to say a book that I've been reading, but I've actually discovered that YouTube has games now, which is super convenient because I'll pop into my YouTube and I've been watching a lot of watch videos. Somehow I've turned into a watch person. I haven't purchased a watch. But these are. [00:30:53] Speaker D: Those YouTube games are. They're like HTML5 games, right? It's the same type of game that's in Telegram and LinkedIn. [00:30:59] Speaker C: And I think it's all. I think it's the same. Same kind of thing. I just think that YouTube. As somebody who uses all those apps, my social contract with YouTube is very much like entertainment. So I get it open and it's like, oh, there's a game that you could play. And it's like, fuck it, I'll play it. I played two games. They both sucked big time. But it's a really interesting concept. Imagine I'm about to watch a YouTube video about a watch and there's like an ad that pops up. That's those, you know, those ones where they're shooting and you, like, you get the levels up. The game that doesn't. [00:31:28] Speaker A: Yeah, Last War. [00:31:29] Speaker C: No, the game that does not actually exist. But every single fucking advertisement shows that gameplay. It's so annoying. Imagine if you could click on that and you're playing that game on YouTube. I think that's super compelling. That's not quite the experience that I've had where I, like, see the ad, immediately click on it, and I'm playing in the game. But I thought it was kind of a cool, cool concept. I tried to. I clicked on one that looked like one of those shooter guys. I forget what you called it, Phil. Whatever that, that game is, that doesn't exist. [00:31:56] Speaker A: Class four. [00:31:57] Speaker C: Yeah, but it was like. It was actually. You're a guy, like, running, and you're trying to get big and tall and you're, like, trying to jump far. It's one of the jumping games. I freaking hate those games. That was all they have on Roblox. It's just like a bunch of jumping games and like, oh, see how far you can jump your car. So I played like, literally three rounds of that quit. And then there was a whole game, which was fun, but, you know, it's a whole. [00:32:17] Speaker D: Why are these html5 games popping up everywhere? I feel like they've gotten, like, showed up at all these big tech, super easy to make. [00:32:24] Speaker A: LinkedIn has them Reddit. Basically anywhere you can bolt on a shitty. A shitty HTML5 word, hyper casual game. It gets bolted on. [00:32:33] Speaker D: But is this. Is this new tech or like, is this just a trendy wave? Like, what. Why is it happening now? [00:32:38] Speaker C: I think it's probably a confluence of things. First of all, you have, like, Roblox taking the world by storm, and everyone's like, running around with their heads chopped off. How do we get a part of the Roblox? Oh, let's turn LinkedIn into Roblox, and we'll have a bunch of games that people can jump, you know, into. You also have. The cost of producing these things is plummeted to basically zero. I mean, you could probably. [00:32:59] Speaker D: It's gotten cheaper. [00:33:00] Speaker C: Yeah. In the last three years with AI, I would say so. Like, these are super, super easy to code, super easy to produce. I don't know how hard it is to get them published on these platforms like LinkedIn or YouTube, but I can't imagine it's as hard as it is to get a. An app published on the App Store. [00:33:16] Speaker A: I don't think it's that complicated. I mean, look at casual game numbers, which have insane amounts of dau. Right. But don't monetize very well. I think that's ingredient number one. I think ingredient number two is. Look at how popular wordle is. I think wordle was a very big moment when everyone saw how. Well, the New York Times, I Mean, the New York Times is making a boatload of money. I forget the percentage share of their subscription which is now driven through wordle. But they built a business extremely quickly on that back of a product. Probably the third ingredient is duolingo and gamification just being a hot buzzword trend. And I'd say maybe ingredient number four is that I think all these product managers are basically out of ideas. And essentially at some point product management becomes a forex game and you're basically just trying to like grab whatever you can to throw it into the platform. I mean if you log on to Facebook. Has anyone logged on to Facebook in the last 10 years? [00:34:01] Speaker D: I messaged people on it, yes. [00:34:03] Speaker A: It's a fucking cesspool. [00:34:05] Speaker D: 10 years. [00:34:06] Speaker A: It's a cesspool of product managers basically stapling on ideas. I mean it has frickin. And I think this is like the natural endpoint of all these platforms and games is now just another thing that's easy to plug and play because it has so much mainstream adoption. [00:34:20] Speaker C: An obsessed pool of every single possible mechanic you could imagine. I was scrolling on, I did open Facebook and I realized they had Facebook reels and I was like, what's the difference between Facebook reels and Instagram reels? [00:34:32] Speaker D: It's the same, right? [00:34:33] Speaker B: Yeah, everybody's trying everything except for the. [00:34:35] Speaker C: Experience on Facebook is much worse because you get the regular ads, of course, you get ads in between in the middle of reels and you get ad pop ups on top of non reel. It's like fucking insane. [00:34:47] Speaker D: They just. It's the same product just with extra ads. [00:34:49] Speaker C: Yeah, I was. Which just makes sense, right? It's an older demographic, they have a much lower, you know, they have a much higher elasticity for that kind of bullshit than we do. [00:34:57] Speaker D: But Arto, you mentioned you were working on Telegram games for a little bit, right? Was that the same thing? Like kind of HTML? [00:35:03] Speaker B: It's similar. So there's, there's browser games that are kind of having a comeback for several reasons and I think Telegram is out of all of these platforms and I think there's Reddit and Twitter as well. I think that's maybe most comparable. WeChat, of course, that's one that we haven't mentioned. Telegram's probably most comparable to older Facebook when they used to have games. So how that works in practice, I think there's a, there's a lot of users using it as a communication platform. So in theory you have this ocean spread there. It's not just HTML games. You can have Unity games and anything that functions the browser game can be launched from Telegram. One of the nice things there is that you authenticate users via their social connect automatically basically so you know who they are. They don't need to, you know. [00:35:44] Speaker D: But is, is there much traction there? I, I heard, I know Web3 was buzzing about Telegram games for a bit and I have a friend who is a Web3 guy who like, basically was like it was a, his, his Telegram game was a flop. But I'm just curious if it generally it's been successful. [00:35:58] Speaker B: There's users on the platform, there's not a lot of like successful games because the discovery within the platform and paid channels and all of that is still like very in its infancy. I've seen the tools developed quite a lot within the year, but they're still quite bad. And the other thing is like just monetization in the platform is something they're pushing pretty hard, but I think it's still a pretty poor and concept for a lot of, a lot of users, except for in certain countries. [00:36:22] Speaker A: How do you think the games piece if at all ties into like the core loop of Telegram? Like I get the idea Telegram has a lot of users. Okay, we solve for distribution. But like what. And I guess that's the thesis behind all of these. But I guess is there something that you see deeper that makes sense to make games for Telegram where Telegram and. [00:36:39] Speaker B: WeChat differ a bit from, let's say a LinkedIn, is that they're multipurpose tools for a lot of people. Not necessarily for the four of us, but for people in Eastern Europe, they do spend a lot of time there. So that's one thing. So it's not that weird to think that they would go to games from the panel on the left. One of the things in practice, if you launch a game through Telegram, you use a Telegram bot as kind of a launcher bot can also function as a push notification replacement. Basically it can remind you that the game exists and something's happening. So there are a few things there that make it I think a more interesting gaming platform than a LinkedIn. But yeah, it's still, still developing. [00:37:15] Speaker A: What do you think they need to do to make the platform more compelling for the use case that you guys have at your startup, the game that you're developing? [00:37:21] Speaker B: Um, I'm pretty sure WeChat is doing a lot more. They have a paid ad ecosystem with actual like targeting and I believe they also have like a My Games tab, which is what happens in Telegram is it's a Lot of, lot of clutter. So you have to use folders to remain sane on that platform. And the games, even though they can exist there, they kind of get lost in the shuffle really easily. So I don't know how you would like bookmark games in a more usable way. I think that would be a big thing. [00:37:51] Speaker C: So on Telegram, there's no way to, as a developer, there's no way to like advertise your. Your game. Like they basically have to be natural. [00:38:00] Speaker B: There are ways. So you can advertise within other mini apps is what they call them, basically apps, so that that ecosystem exists. You can advertise with within like the chat channels and things like that. [00:38:11] Speaker C: It seems like a huge opportunity. What, what do your user numbers look like on those platforms versus just like a web browser. [00:38:19] Speaker B: So what we see working for us within Telegram more than in the browser context is just a social spread. So we've gone pretty heavy on incentivized referrals. Copy this link to your chat it is. If that works somewhere, I would think that it works in that context. And that's been created so that there's been people sharing links. [00:38:40] Speaker C: So you're getting noticed, but you're not necessarily getting plays from that versus they are playing primarily through the web browser. The reason I ask is our game is also a web browser. Well, one of our games is also a web browser game. That's the primary product that I work on. [00:38:52] Speaker B: If you go to like the pros and cons, I think it's very similar to WeChat from what I understand. So you have like an instant gaming setting basically. So you have like at least the theoretical advantages on how friction free it is to sample games. So kind of like free to play, but more. But then you lose out on the kind of people come in with very little interest. So day one retention might be 20. [00:39:12] Speaker C: Yeah, that's. That's one thing I've noticed in web three gaming is just retention is much weaker. And part of that's because typically the way that people interact with your product is either through a brief interaction like that, or it's a wallet connecting to the program. And then, you know, maybe, maybe it was somebody, a bot that connected to the program. And so, you know, that shows up as a user. It's a much lower barrier to entry compared to something like downloading an app on the App store. [00:39:38] Speaker B: Yeah, it makes it a pretty interesting challenge for finding benchmarks because they don't exist. [00:39:48] Speaker C: So, so basically what I wanted to. Sometimes I just get these ideas that you know are like, would make fun python experiments or fun python exercises. And in this particular case, I was thinking about the way that algorithms, specifically content a sorting algorithms on platforms like Instagram, YouTube to promote a piece of content to a user or put that content in order from top to bottom. How do the decisions about how we design those algorithms impact the platform both from the consumer's point of view. Somebody scrolling through my enjoyment of my Instagram reels and the producers or the content creators, the people who are putting that content out there. So what I wanted to do was study the consumer welfare of the consumers, the people who are consuming that content, and the wealth and equality of the producers, the content creat under different reinforcement algorithms. And so I'll go through three very, very basic, basically the most, the three most basic forms of algorithms that you could imagine and look at the different the performance of these KPIs over a simulation of 450 periods. So without further ado, we're doing a slideshow. [00:40:57] Speaker A: I love it, Chris. Brings me back to grad school. [00:40:59] Speaker C: Yeah, exactly. I'm like, we're an econ podcast. We should, we should. If I was a better person, this would all be written in latex and this would be a have an ugly blue background. Anyway, we've got end users, K producers and t time periods. Now the time periods here are going to have to be pretty, you know, like, loose with what this means. Every single time somebody scrolls is a time period. So we're just gonna, we're gonna measure one event per time period. So everybody gets served a piece of content in t time period and then we move on to the next time period. The decision process is not really a decision process. It's mostly stochastic. And then there's kind of a decision on whether or not to return to the platform. So in the consumption side of things, a person's consumption is based on some probability that they're going to get a specific piece of content. So the probability of user I receiving content J at time T is pitc is influenced by a weighting algorithm called digt. So dijt this is just the weight, the probability that that specific piece of content gets sent to the consumer is going to be their utility, effectively. So I have some probability of consuming and I have some utility that I get from the thing. Probably P is the probability that they consume the thing, D is the specific weight of the thing based on the algorithm, and U is the actual utility they receive from the thing. Now, utility is distributed across these goods such that some Goods are better than others. So if I get a really good piece of juicy content that I love and the probabilities, you know, you roll the dice and I happen to consume that thing, I'm going to get a lot of utility from that. My utility is high. I could, however, consume something that I don't like. My utility is low. After I've consumed that thing, which is a stochastic process, based on the probability of having encountered it, then I'm going to make a decision on whether or not I want to return to the platform or churn. This is the first true kind of decision, but it's still not really a decision because it's a coin flip. A bias coin is flipped based on the utility that you consumed. So if you consumed a really high utility good, you're going to have a really low probability of churning. [00:43:05] Speaker A: Okay, people, people, people, people, people consume things and it's, it's weighted by the, the algorithm. Okay, sounds good. There's profit. This makes sense. I want to get, get me to the meat of the model though. This is what I want to get. [00:43:18] Speaker C: To the profit for the producer. The person creating the content is just going to be the sum of all of the people who consumed it. They get one, whatever you want to call it, one unit of, of income. For every single unit that's consumed, you sum all cijt, which is the consumption of that good across all the individuals who consumed your good. And that is your profit for, for producer J in time t. So that's PI jt. Now this is where this decision variable or this weighting variable dijt comes into play. Digt is the probability that somebody encounters a piece of content. So and every, at every period of time, there's some probability that I encounter good j at time t for individual I. And dijt is based entirely off of this Omega thing. Omega is just a weight that that thing gets. The algorithm is going to determine Omega and Omega. This function is basically just saying for specific good J, put that specific goods weight and divide it by the sum of all the other weights and that gives you the probability that you encounter it. So if there's a weight of, for each of the goods, there's a 1 over n probability that you encounter. [00:44:23] Speaker A: We did Diablo math. There are weights, they translate to probabilities. Okay. Okay. [00:44:27] Speaker C: Yeah. So now is where we get into the good stuff. We're going to define three distinct algorithms. I think of these algorithms as being three points on a triangle where you have the most kind of each of those points. Represents the most basic type of algorithm you could possibly imagine. The first algorithm, the tip, the top point is going to be a fully random algorithm. So weights are distributed fully, uniformly. You have the same exact probability at any time of encountering any piece of content. It's, it's uniformly random across time and across individual and across good. So even if you have good goods or highly, you know, valuable, a really good good and a really bad good, and you know, you've got all sorts of different, you have the same probability of encountering that thing. So that's what the random algorithm does. The popular algorithm is going to basically just take a look at all the consumption from the previous period, sum it all up, and it's going to give a weight to each of those goods based on their consumption. So basically the most popular good gets the biggest weight, the second most popular gets the second most, and so on and so forth. So what you can imagine in this model is that the more, the more a good is consumed, the higher weight it will have next period, which means the more it will be consumed. So there's kind of this reinforcing. [00:45:40] Speaker A: Hold on, tell me more about this. How does it actually get, what do you say? Content that was more popular yesterday is push day. So you're saying what you do is you go through and you see what people actually consumed based on the random, let's say period one is the random weights. And then what we do is we see which of the random weights actually had the most uptake, maybe most time viewed, whatever it may be. And then what we'll do is we'll then adjust the weights based on that and then the next day we'll serve based on that, that determination. [00:46:06] Speaker C: Yeah, and this is, this is algorithm. It's, it's going through, it's calculating these weights every single day and it's, and it's producing the new probabilities, those dijts and serving them to the consumer who then makes, you know, gets the consumption based on those probabilities. So in this model, the popular stuff gets reinforced over and over and over and over again. And if you were to kind of take this to its extreme, only this good would eventually be consumed. Now, I incorporate a couple of different random components into the simulation. First of all, I assume that there's a 5% probability that a churned user comes back in any period. So basically of all the churned users, 5% of them come back. There's kind of this natural, you can call it some magnetism of the app application. Maybe I churned yes. You know, three days ago, today I come back. The other piece of randomness here is just a slight shock, a small stochastic shock to the probabilities in each of these models. So the weights do get a, a normal shock to them. So you, you would technically have an epsilon here. But I don't put that in the model because it's a little confusing. But basically, if Omega would be, you know, 2.14, we add a random shock that's, that's uniformly distributed. So it gets added 2 point, you know, an additional 0.5. So now it's 2.64. That's the popular model. And then the final model is kind of the ideal model. This is the individual model. This is one where the algorithm perfectly serves an individual the content that maximizes their utility. So basically, if I really like specific good J, then the algorithm is going to give me specific good j in my feed. It's not going to, it seems unrealistic. [00:47:42] Speaker D: How would it know? [00:47:43] Speaker C: Well, exactly, that's it. The point of these is not necessarily to be realistic. It's to show the different extremes and then everything else kind of falls from there. Now if this was like a fully fledged econ paper, I would have a much more, you know, I would have. [00:47:56] Speaker A: Why can't we just do multi arm band? Isn't the multi arm bandits the answer to all of this? Right, there's an exploration arm and your, your, I mean Algorithm 2 is basically. [00:48:04] Speaker D: Like a global multi armed bandit. And this would be like an individual multi armed bandit. [00:48:08] Speaker C: I mean how is algorithm to a multi arm bandit though? [00:48:12] Speaker D: Maybe it's not a multi arm bandit. [00:48:14] Speaker A: So you're serving, you're serving the high, highest payouts, right? You're pulling all the different levers, you're seeing which of the levers has the highest return to player rtp. And then what you're doing is you're adding new pieces of content into the pool based on an exposure variable that you've set based on how wide you want it to go relative to the RTP that you know is winning. And then you're readjusting based on the exploration and the ones that you know are winning. And then you're again trying to find the true RTP of all the different slots. [00:48:42] Speaker C: Yes, my, you know, we don't have the multi armed bandit kings here right now. But my, my understanding is that the algorithms themselves would be the arms on the, the bandit. So you would serve three different algorithms. You would see which one of those rewards, you know, has the Best result. [00:49:01] Speaker A: Meta, Metabandoning, metabandoning, where you're just throwing new models in. [00:49:05] Speaker D: I was saying each piece of content is the. [00:49:07] Speaker C: Well, we don't have a way to control the type of content. [00:49:09] Speaker D: Isn't this, this is controlling the type of content that's displayed. [00:49:13] Speaker C: Yeah, but that's what I'm saying. Like you can't change the content, you can change the model. So what you would do is you would start with these three models, you would use all of them, and then you would find that the individual, as I'll show, you'll find that the individual model is the optimal model. Now, obviously this is an ideal. [00:49:27] Speaker D: Let's, let's see the results. [00:49:29] Speaker C: Okay. All right. So this one just serves, this is an idealistic model. We serve the user the best content. Now I'm going to show you two different worlds. One world is where all goods are homogeneous. This is our baseline. It's not realistic. And then the next world is where all the different goods, I keep using the term goods, but videos, content creators, all the different content creators are heterogeneous. There are some really good ones or some really bad ones. So first, for the baseline case, the heterogeneous or homogeneous goods. So you can see I've got the random model, individual model, and then popular model on the left, center and right, respectively. I've got active users on top, total utility in the middle, and the Gini coefficient, which is how we're going to measure the producer's income inequality. So this is a Gini coefficient of content creators profit. This is content creators profit plugged into the Gini function and calculating a coefficient. You can see the random algorithm sitting there, around 875 users with a total utility in the 620 to 640 range and a really tiny Gini coefficient of 0.02 to 0.04. Now, a small Gini coefficient in this context actually makes a lot of sense. There's no differentiation to quality. If you improve your quality, the algorithm is not going to reward you for it. It's just completely random. So you would actually expect a low Gini coefficient. Low Gini coefficient means that wealth is perfectly distributed. Everyone has exactly the same amount of wealth. And if there's no way for you to differentiate in order to get more wealth, then, well, yeah, it makes sense that it would be a very low Gini coefficient. We go to the individual algorithm, we see something very different in terms of the total users and a similar Gini coefficient. Okay, so the total number of users skyrockets for the individual algorithm, we're sitting around 980 to 985 user, much higher total utility and around the same Gini coefficient. In fact it's a little bit lower. Now the reason for this is are. [00:51:20] Speaker D: User preferences, like totally heterogeneous user preferences. [00:51:24] Speaker C: Are drawn from a distribution. [00:51:27] Speaker D: So they're all random and independent. It's not like there's good content and bad content. [00:51:31] Speaker C: Exactly. Right now all the goods are homogeneous. There's no differential utility for different goods. All goods provide the same utility to consumer homogeneous good. So you could think of this at the individual level. All preferences are, are the same. So we'll, we'll skip through here real quick. Gini coefficient again super low. Because there's no return to having good content. There's no advantage to having good content. It's not rewarded through wealth. And so the GENIE coefficients remains low, but much higher total utility, much more active users. But from a producer standpoint, why the, the hell am I on this platform? I don't get rewarded for having good content. The wealth is perfectly equally distribute distributed. Why would I put out better content? Why would I spend more money in order to produce better content? The popular algorithm is basically the same as the random algorithm. The only difference is that the Gini coefficient is higher. Now what this means is that there are select few content creators that are receiving the majority of the income and that's because they've randomly. At the beginning of the period when GENIE was basically zero, they randomly won the lottery. They just happened to be served to more people and they got reinforced more and more and more and more and more over time. And this would probably converge to one if we were to run this algorithm for quite a while. [00:52:44] Speaker A: So why isn't everyone. Why, why isn't so, So I think we're getting to the kink of the model, right? So you have the popular algorithm. There's a lot of inequality in utility, but I don't understand where the inequality. Why isn't everyone, everyone is better off in the popular algorithm. But why is it only some people are better off? Is it because the algorithm scales for everyone? [00:53:03] Speaker D: The individuals are better off in the individual algorithm? Like the DAO is higher. [00:53:07] Speaker C: Correct. [00:53:08] Speaker A: Correct. Because it's an all knowing algorithm. Right. It can just perfectly pair content. But why is the Gini coefficient rising. [00:53:16] Speaker C: For the popular algorithm? [00:53:17] Speaker A: For the popular algorithm this feels like the crux of the model. [00:53:20] Speaker C: So this is, this is definitely, I mean it kind of depends on GENIE is not the full picture because there's two components to inequality. The first one is, does an additional unit of quality get me better more users? And in this model, because the goods are homogeneous, there's absolutely no return to a better good. Because people's preferences for those goods are homogeneous, no good is better than another. They're basically uniform. So there's no, there's no return to differentiating your content or making better content. Yet Geni is going up. [00:53:51] Speaker D: You say it basically randomly. King makes it like randomly picks. [00:53:55] Speaker C: It's exactly right. It's creating a king randomly. It's saying, okay, you know what? Goods 1, 2 and 3, you guys are the winners. Or goods 1 and 2, you guys are the winners. So you're going to own most of the pie by the time this is over because you're content is being reinforced. But one and two, they could have the crappiest content in the world or, you know, their content value doesn't matter. There's no, there's no differentiation. So this is the baseline. The reason for this, this setting is really just to get us a baseline for those user numbers for all these KPIs across the three models. Once we introduce a much more realistic environment where we have heterogeneous goods, we get quite different results. So starting on the random algorithm, the first thing you'll notice is can you. [00:54:32] Speaker D: Define heterogeneous goods here? [00:54:34] Speaker C: Yes. So now there is a, There's a good. There are five goods in this, in this model, as I mentioned at the very beginning. Now there's a good that is, has low average utility. The average utility that it gives to, to the consumers is quite low. I think it's got an average utility on a scale of 0 to 1 of like 0.1. And then you've got the next good that has an average utility of like 02 and then 0.5 and then.07 and then 0.8 or 8,5. Each one of these goods has a different utility function. And the average utility that the average consumer gets from each one is increasing the more high quality the good is. So heterogeneous goods just means that there's a shitty good, there's a medium good, and then there's a really good good. So there's a bunch of different qualities across those. So if I'm a consumer and I consume good one, I'm going to get a lower utility from that on average than if I'm consuming good five in this case. So there are, there are goods that are higher quality, they return more utility to the consumer. And we're going to see that reflected in the individual algorithm in the middle. So the random algorithm, you'll notice lower retention, far fewer users in this model. And that's because people have much more strict preferences now. There's different goods. Not all the goods are giving an average utility of 0.5. Some goods are giving an average utility of 0.2 0.1. Some goods are giving an average Utility of 0.8 0.9. So what this does, if you're randomly selecting stuff, it's going to land around 50% 50 users because half of users are getting an above average good and half of users are getting a below average good. Total utility is thus also really low. You'll notice though Gini coefficient still very low for the random algorithm because it doesn't differentiate based on quality. It just randomly selects a good to go in front of the user on the popular algorithm. We actually see worse results in the random algorithm, which is kind of interesting. So we see that they're floating between 4 and 500 users. Total utility is around 200, 250 compared to 250300 for the random. Gini coefficient still going up for the exact same reason as before. What's interesting is the individual algorithm now has a much higher Gini coefficient because you can get those differential returns by improving your quality of your content. As you know the best quality good, you're able to get more of the pie because more people want to consume your content. So you have super high active users, super high utility and a solid Gini coefficient in the.05 to 0.6 range, which is what we would expect for like a first world country. [00:56:59] Speaker D: And that's because it's the middle one. It's basically just picking the top tier quality. You mentioned the five quality tiers. [00:57:05] Speaker C: Yeah, this is like this would be pure random. This is purely. If we were to only, only promote the most popular thing. And then this is if we were able to. This is the ideal, this is the ideal model where we're able to perfectly situate the content to the, to the user. Now realistically all models are going to be some formulation of these three. They're going to be something in between. [00:57:24] Speaker D: Either something your popular algorithm sucks. [00:57:28] Speaker C: The popular algorithm sucks. [00:57:29] Speaker D: Something about it isn't working. Right. Like I would expect the popular algorithm to outperform random on heterogeneous goods. But like there's, there's no feedback loop because what's popular is determined by what's shown, not what's engaged with. Right. Like the popular algorithm doesn't get that feedback of what Is the utility the user is getting and reinforcing that. [00:57:47] Speaker C: The critical thing about the popular algorithm is it experiences information cascade. If it feeds a bunch of shitty low quality content in the first period, randomly, purely randomly, because the weights are equal in that first period, then it's going to reinforce those shitty things for the rest of time versus if it randomly selects a bunch of high quality goods, then it's going to get stuck in a high quality state. It's going to information cascade towards those high quality things. [00:58:13] Speaker D: But is, is there any mechanism by which high quality feeds back into the popular algorithm? Like, like repeat views or higher retention or anything? [00:58:20] Speaker C: No, purely on the, and the amount. [00:58:22] Speaker D: Consumed is purely based on what's displayed, it's not affected by the utility or anything? [00:58:27] Speaker C: No, has nothing to do with the utility. Now consumers choose to come return to the platform. If it's low, low quality, then they won't return hence. [00:58:36] Speaker D: But the popular algorithm doesn't know that, right? It's not factoring in user potential. [00:58:40] Speaker C: No, it's not. It's, it's factoring in purely how much was consumed in the previous period. Now in a way it does kind of care about how much people liked because if, if everybody leaves then it's, you know, it's going to receive lower input quantities for some of those goods. But because the people left, it's kind of, it's like a degenerate equilibrium. [00:58:58] Speaker D: So hit us with the takeaways. [00:59:00] Speaker C: So the individual model wins, obviously, there's no question about that. The popular model basically amplifies the initial random successes and so it can actually lead to information cascade. You can end up in a situation where the model is constantly reinforcing popular content even though it's not valuable. I mean, this is like, this is basically the risk, the algorithm risk. This is why you want algorithms that are updated on a regular basis, that have, that have search components to them. So you know, they, they explore rather than constantly exploit. They do some exploring and some exploiting and then real world models are going to combine all the different three components. The idea here was to kind of show potential risks associated with leaning too heavily into a specific type of model. For example, a very, very lots of randomness, lots of exploration can be risky because it lowers overall utility of the platform, increases risk. Churn something that's more on the popular side can get stuck in these edge cases. So you want a nice, this combination of the different types of algorithms, that's, that's the takeaway that's really cool. [01:00:07] Speaker D: You just kind of scrap this together as a side project. [01:00:10] Speaker C: It did feel like a, an econ presentation, like a little nerve wracking. [01:00:16] Speaker B: What was the assumption on Churn again? Was it true that that was directly linked to a consumer utility? Which I guess is not true in the real world setting. [01:00:25] Speaker C: Well, so yes, the nice thing about the model, a simplifying assumption that I make, is that almost everything is basically normalized between 0 and 1. So you consume utility, you get a utility between 0 and 1. So we can translate it to a probability. So let's say you consume something that has a 0.3 utility, you have a 0.3 probability of retaining and a 0.7 probability of churning. So we flip a coin. If it's heads for the 30% you retain, if it's tails for the 70% you churn. Now there's also a 5%. There's another stochastic coin flip that happens for anybody who churns in a, in a period, they have a 5% probability of returning the next period. So we flip a, another coin to see if they come back. [01:01:07] Speaker D: Do you think the algorithm designers care about producer GENIE at all? I think they're just fully. All they care about is maximizing. [01:01:13] Speaker C: But I think they should. And that was kind of one of the reasons that I put this together was like we should, we should probably care about the people who are producing the content on our. [01:01:20] Speaker D: Does it affect their bottom line? [01:01:21] Speaker C: Right, but it could. I mean, but in your model you. [01:01:24] Speaker D: Had two where the bottom line was the same but the genie was wildly different. And like I think that most companies are kind of indifferent between those two. [01:01:31] Speaker C: I see what you're saying. So it doesn't matter if you have shitty content or good content as long as utility is high and consumers. [01:01:37] Speaker D: Yeah, like if, if you showed that high genie causes lower long term things or something. Yeah, like I think they would care. [01:01:43] Speaker C: But yeah, that's a good point. It's kind of dismal. It kind of makes all the Roblox slop games make a little more sense. Millions of different games with varying qualities. [01:01:54] Speaker B: Yeah. [01:01:55] Speaker A: So can, can you tell me something practical for Roblox? What, what should Roblox do based on the simulation? [01:01:59] Speaker C: Ooh, that's a good one. I would say there is a risk in. There's a massive risk and purely going based off of popularity, which I know they're, I think, I think that their algorithms are heavily based on what's most popular and they reinforce content like that, especially on like the homepage. So I would say there's A risk associated with that. I know for a fact anecdotally that producer or, you know, producers of content on Roblox struggle to get their games noticed and it's really difficult to kind of break out. So some randomness could be. Could be. [01:02:34] Speaker A: So what about this, Chris? What about this? So isn't the answer here based on the Netflix UI and the Roblox UI is that we use tiles and rows? So tiles are the games that usually sit within the row. Why can't each row just be one of your algorithms? Why can't I have a popular row, a random row? Why can't I just solve. So each, each row is an algorithm. Isn't the way to kind of multi arm bandit kind of this problem is. [01:02:58] Speaker C: Like, yeah, right, yeah, they do something similar. Well, they probably are using different algorithms. [01:03:03] Speaker D: Netflix, each row is a different algorithm. [01:03:05] Speaker B: Yeah. [01:03:05] Speaker C: And the question is, do you stick with the algorithm that performs best or do you continue to. Because then there's like a meta. Meta multi arm bandit. [01:03:14] Speaker A: I think that's the interesting part here. Right? Let's say popular is really popular. Most users opt in or they make a choice based on that row. Do you start to use that row in the second row rather than it being a random row? Do you replace the second random row with the popular row? Essentially? Do you like basically spread what the user is responding to? Do the other rows or do you think, do you hold them fixed? [01:03:34] Speaker C: That's probably where like this paper should go. I should create parameters that basically allow you to calibrate between random and popular. And it could just be mixing the two. It could be literally just like an alpha that goes from 0 to 1 where 0 is pure random and 1 is pure popular. [01:03:51] Speaker A: Let me, let me ask Ardo a question here because the, the thing I've always thought about is, is, you know, the PMs will love to talk about recommendation engines for like IP SKUs or personalized offers. And I basically have thought that it's a bunch of hogwash. Bullshit. Because the search costs in mobile games are very low. There aren't that many SKUs. It's not that hard of a decision cost. And the idea that you're just going to pop up a single SKU isn't really optimizing that much versus just literally throwing all the different SKUs into a fucking page. As consumers are so highly trained, they're going to pick the one they want. Do you think there's anything to that or do you think we should use algorithm based design like Roblox to Serve that type of content? [01:04:29] Speaker B: Yeah, that's an interesting. I guess depends on the toolkit. And that's one of the ways in which web shops are interesting. Right, because you just have more choice. [01:04:36] Speaker D: There's that. [01:04:37] Speaker B: It's a hard one because there's also like a lot of variants, like build your own offer, all that kind of stuff where you just make it more interesting rather than trying to predict what they would want to be served. Right. [01:04:50] Speaker A: Do you think that's a better strategy? Do you think we should just give users autonomy and how they set up the content themselves? [01:04:55] Speaker B: Too much choice is too much choice. That's. We do know, like, giving them too much options is bad. [01:05:03] Speaker A: Do we know that? Do we know that in free to play games? Because. Because I know. The psychologists tell us that. I don't think I've seen that. I don't think I've seen that from an econ perspective, especially when the consumers are highly trained and they have high knowledge. I don't know. To feel free to disagree. [01:05:14] Speaker D: Why do you think Fortnite does a rotating store instead of just have everything available a la carte? [01:05:18] Speaker B: They've. [01:05:18] Speaker A: They've moved away from rotating store. They've added more and more options. They used to have six items. They've moved to 30, 60. Everything is becoming more permanent. Call of Duty offers all of the SKUs. If you keep scrolling, they offer it in the. If you go into each individual weapon, you can see all of the call of Duty SKUs. [01:05:32] Speaker C: There's definitely some behavior econ papers on it. I don't know how good we feel about those. [01:05:37] Speaker D: Yeah, I think Phil's saying those are sus. [01:05:38] Speaker A: Yeah, the psychologists are always sus. I give a little bit more Crayoler to Thaler than I do. You know, Dana, Ariely, Mr. Mr. Frogtoes over there in Duke, who still is somehow as a tenure track position and got like, no, not even a slap on the wrist for fabricating data. Do you. Do you think that there's such thing as choice overload in mobile free to play games when it comes to SKUs. [01:06:00] Speaker B: I mean, intuitively, yes, There. There was the thing that Eric wanted to talk about with discrete choices as well. So I think that's kind of like going in the opposite direction of forcing a choice between a few limited options. I think that that setup produces different types of information than the whole here's the whole catalog kind of setting, especially if you were able to remove the option of none of the above, which I don't think is very easy to do. [01:06:25] Speaker D: Yeah. Aren't there like discovery and evaluation frictions here if you've got a thousand different things? [01:06:30] Speaker C: Right. [01:06:32] Speaker D: League of Legends had this problem also with cosmetics. They're like highly heterogeneous, so it's like very hard to scan the whole catalog. [01:06:39] Speaker B: Depends on what the things are as well. If it's just like different sizes of hard currency packs, that's a, that's a different setting of like, let me just like read a lot of stuff about these like 20 things you're trying to sell me. [01:06:49] Speaker A: So let me, let me ask you another question that's been bothering me, Arto. It's been bothering me for a really long time. So like when you go to a hard currency store, usually we offer quantity based discounts. So the price per unit of hard currency decreases as you're willing to spend more dollars. We're giving you, we're giving you a bundle saving. Why isn't it the case that we just can't let the user decide how much hard currency they want to purchase and then give them a button that generates an auto IP offer? Why offer discrete choices? Why not just offer a slider? [01:07:16] Speaker B: That's what the European Union wants, right? [01:07:18] Speaker D: I think this is the app store. Shit. Like the Apple iOS, you have to offer specific price points. You're not allowed to offer arbitrary. [01:07:25] Speaker A: That's true, but I think we could expand that range. I mean they increased the cap to like 20,000. Not only that, you have web shops. We could do this in. I mean crypto could do this. [01:07:33] Speaker B: Yeah. [01:07:33] Speaker A: Why not do it on the web? [01:07:35] Speaker B: I don't know what different results you would get, but in practice, I think what a lot of people are trying to do is match the. Match the quantities to things you buy within with the hard currency. Right. So that's what's the up. [01:07:46] Speaker D: What's the upside, Phil, you think more conversions by being able to choose specifically $17. [01:07:51] Speaker A: To me, if we think that there is an uplift from offering quantity based discounts for let's say $100 SKU. Right? So we can imagine one world in, you know, a dollar buys, you know, 10 units of hard currency and then you get that same exchange rate whether you spend $10 or $100. That's not the paradigm in free to play. The paradigm in free to play is that we think it's LTV versus a control based on that exchange, you know, flat exchange rate. I just described. To offer quantity based discounts I think is sound and I think we'd find that in an experiment. My argument is if that's true, the way you break that idea is that you essentially give them even more options. You basically let them design. You basically let them choose what that discount is, going beyond the hundred dollar, letting them say, okay, I may have these fixed preferences. Why not these fixed intervals where I can express my preferences. Why not just make it a continuous rather than discrete distribution? [01:08:41] Speaker C: As you increase the value, you kind of see the, like, you have the 45 degree angle and then you have your value and it like, like keeps, you know, extends above and beyond the 45. [01:08:52] Speaker A: I don't want ketchup packets, Eric. I want a ketchup bottle and I want to squeeze how much ketchup I want out of it. [01:08:58] Speaker B: Yeah, it does maybe, but it does make it feel less of a game. Because if you look at how the hard currency stores work, everything is visualized differently. You've got this like discrete choice there. And I don't know if you've seen this. You must have seen this, this trending or trending mechanic where first purchase of each gives you double. So you kind of did this all the time. You kind of remove those options from the toolkit. If, if you don't have, you know, the six options. [01:09:25] Speaker A: I think we, I think we could, I think we could design around that. I mean we could offer you bonuses. One time bonuses based on reaching certain thresholds. The first time you reach this threshold and the first time you reach this threshold, the first time you reach that. [01:09:35] Speaker C: Threshold, you can emulate it all. You could have a cute little crate of coins that gets bigger and bigger and then ads as you scroll farther and farther with your sku. [01:09:44] Speaker B: Or like hold the button and stop when you reach the quantity you want and that kind of stuff. [01:09:48] Speaker A: Look at man, I played a lot of non stop night and all I'm saying is I would have liked to be able to choose how many animals I wanted in my skus. I didn't want just the dragon. I wanted to be able to say, I want the dragon, I want the frog. I want all the different animals. And I wish, I wish I had that ability. And to your point, Arto, I think the thing that's interesting is like Squad RPGs are letting you design some of your skus now that's kind of interesting. [01:10:07] Speaker D: What does that mean? Like you pick the characters in the bundle. [01:10:10] Speaker A: Yeah. Like, yeah, you pick the resources. [01:10:14] Speaker B: And there's a gacha variant of it as well where you kind of, kind of like tweak the odds. So just keep, give you a bit more agency there Wish listing. [01:10:21] Speaker C: But Phil, I do think, like, probably there was some psychologist who was like, oh, if you offer bundles, then the person has to. They can't get the bundle that's cheaper than the amount that they need. They have to get the bundle that's above what they need. So if I need $18 worth of stuff, I need to go get the $20 bot. No. Skew. [01:10:39] Speaker B: I think. [01:10:40] Speaker A: I think the EU thinks that's the case. I've never seen a monetization designer actually do that. I would say every, every free to play hard currency store is usually pegged at 100 to 1 or something along those lines. It's pegged at an easy number. And the quantity discount unpegs that because it's in the interest of the consumer. It's about things getting cheaper rather than things getting more expensive. It wasn't about some mental math trick. It's because we offer one a quantity discount. And the reason we don't just make everything. [01:11:04] Speaker D: Fuck. [01:11:04] Speaker C: Fuck. [01:11:05] Speaker A: We could spend a lot of time in this. Fuck the eu. They know nothing about games. Those regulations are stupid and we should cover them. Next episode. [01:11:12] Speaker D: I really like what they did for usb. C. Standardization. [01:11:15] Speaker C: Okay. Oh, that's us. Usb. Like on your phone? Yeah. [01:11:19] Speaker D: You know how like a. Every phone and everything has like a different charger? [01:11:22] Speaker A: Nope. That is stupid. Eric. How are they gonna. Okay, let's imagine that it was micro USB2 that was the standard. Would you have been happy? Would you be happy? [01:11:30] Speaker D: No, because, because they, Their, their regulators did their research and were like, what is a good standard going forward? This thing seems. [01:11:36] Speaker A: You don't think they would have micro USB2 when that was a standard. What about USB4? When will USB4 be introduced? [01:11:42] Speaker D: You're inventing hypothetical scenarios where the regulation went bad. [01:11:46] Speaker A: I'm not. I'm saying that there's going to be a new USB standard that will be invented. USB3. USB4 will come around. Who gets to decide when USB4 adoption happens? The EU Commission now gets to decide. What about HDMI? [01:11:58] Speaker D: What about all these other. [01:11:59] Speaker C: Oh, it's pretty convenient to have all of my devices on the same. [01:12:03] Speaker D: All right, well, let's see how this turns out. Maybe in 10 years I'll be eating my words, but I think right now I'm very happy with the situation. [01:12:09] Speaker A: You. [01:12:09] Speaker B: You. [01:12:10] Speaker A: You absolutely will. Let's close the episode here, Arto. Is there anything we can close with for you? Anything we can plug for you? Substack? [01:12:20] Speaker B: Nothing in particular. I. I stopped producing content online. [01:12:24] Speaker A: You have a Twitter. Can we link out to your Twitter. Should we follow you on Twitter? [01:12:28] Speaker B: You can, but you tweet. That's what I meant. I stopped. [01:12:32] Speaker D: Tell us who we should follow then. Who should we follow? Yeah, Is there any YouTubers or people who post on Twitter or whatever that you. You like? That you would recommend we follow? [01:12:42] Speaker B: No. Think of a shill right here. [01:12:47] Speaker C: What about your web3 game? What about your web3 game? What do you have for us to buy? [01:12:52] Speaker B: I guess like my. Oh, for the games. I think like my assessment for the state of Web3 games is probably pretty close to yours. I mean, in terms of cool games to play. It is quiet. [01:13:04] Speaker A: This is going to be an uplifting episode. [01:13:06] Speaker D: Stay off social media. You heard it from Ardo. [01:13:09] Speaker B: Yeah, it was funny because the UGC thing, that discussion is completely different in the context of Roblox versus Twitter. [01:13:15] Speaker A: I feel like our guest today has been Arto. You can follow him on Twitter where he doesn't make new content and he also has a game coming out which we'll link to. His startup is awesome. I know your co founder as well and we're happy to have him here. GIM economist cast episode 44 in the can we should teach this to our children. [01:13:33] Speaker D: Economics is major. Everyone has to major in economics. [01:13:39] Speaker A: Number one for personal survival. [01:13:41] Speaker D: Economics is major.
Previous Episode

Other Episodes

Episode

February 20, 2024 • 01:14:43
Episode Cover

E22: Airdropping The Missing Web3 Palworld Take

No Palword take is too late; something is not lost on the Game Economist Cast crew. Eric beguiles us with Cassette Beasts' analysis, while...

Listen

Episode

May 19, 2024 • 01:05:43
Episode Cover

E26: Price Theorists Battle Behavioralists for Marvel Snap

Phil gets Eric pilled into a trio of roguelikes, where we prime the pump for Squad Buster's eventual shake-up of the genre. Chris ran...

Listen

Episode

March 26, 2024 • 00:49:47
Episode Cover

GEC BONUS EP: What's up at GDC 2024?

Eric Guan on the economics of the video game job marketPhillip Black on why big gaming co. employees are frustratedDr. Julian Runge build a...

Listen