E05: Korv, Deluxe Edition Theory, & Market Power

<cite>phil:</cite> <time>0:00</time> Hotdogs are actually quite big in Sweden, like Cove. Cove is a culture. People think it's all about meatballs. No, man, it's about co cov and Canal. Bula. Cove is just like Swedish for hotdog or like sausage, K O <cite>christopher_kaczmarczyk_smith:</cite> <time>0:12</time> Almost sounds like a bad Polish word. It's if we have any Polish listeners, <cite>phil:</cite> <time>0:15</time> A lot of people on seven 11 will just go out and they'll just be slamming a cove down 7-Eleven. That's normal behavior. That's not weird. That's not weird trucker stuff that's white call activity <cite>christopher_kaczmarczyk_smith:</cite> <time>0:23</time> it's the way people slam big gulps around here. They're just doing it with a sausage instead of a giant cup filled with soda. <cite>phil:</cite> <time>0:30</time> If you need to pick me up I guess I'd slam a cor down. The only weird thing they do is that the bun will be around the dog. It's like a baguette. It's <cite>christopher_kaczmarczyk_smith:</cite> <time>0:38</time> Oh, yeah, yeah. That, that's super popular in Europe. All of Europe. It's the bun is not like this, like V shape. It's like a, it wraps the whole thing up. It's like a giant pig in a blanket. <cite>phil:</cite> <time>0:48</time> but you can't fit ingredients on there. <cite>christopher_kaczmarczyk_smith:</cite> <time>0:50</time> Oh, fair enough. <cite>eric:</cite> <time>0:51</time> Some of these are huge. <cite>christopher_kaczmarczyk_smith:</cite> <time>0:52</time> Eric, don't, don't go down the core rabbit hole. <cite>phil:</cite> <time>0:55</time> If you come to Sweden, I will take you to the finest Cove shop. I know <cite>christopher_kaczmarczyk_smith:</cite> <time>0:58</time> I'll take that as a, take that as a, an exciting opportunity and not a threat. <cite>phil:</cite> <time>1:20</time> Hello and welcome to one of the most inconsistently recorded podcasts It is Game Economist Cast episode five. I've got my usual co-hosts next to me. I got Chris. Chris, how are you <cite>christopher_kaczmarczyk_smith:</cite> <time>1:34</time> Doing well. Been playing a lot of video games. Probably too many video games lately. <cite>phil:</cite> <time>1:39</time> some land in Star Atla? <cite>christopher_kaczmarczyk_smith:</cite> <time>1:41</time> Oh, that was, that's old news. That was last month <cite>phil:</cite> <time>1:44</time> that's about how frequently we record. Eric, how are you? <cite>eric:</cite> <time>1:47</time> Doing all right. Can't complain. <cite>phil:</cite> <time>1:48</time> Have you found the super layer yet? <cite>eric:</cite> <time>1:50</time> We're looking, there's a whole, there's a whole new business model we're trying to try out with our next wave of projects. Maybe we'll talk about it sometime, but <cite>christopher_kaczmarczyk_smith:</cite> <time>1:56</time> ooh. Spicy <cite>eric:</cite> <time>1:57</time> Celebrating a Valentine's Day, a day early to dodge the rush. <cite>phil:</cite> <time>2:01</time> in crypto. <cite>eric:</cite> <time>2:02</time> No, no, no, No. Just in my personal life. <cite>phil:</cite> <time>2:04</time> Oh, okay. <cite>eric:</cite> <time>2:05</time> Yeah. <cite>phil:</cite> <time>2:06</time> And I am Phil out of Game Economist Consulting. We have an awesome episode for you guys. Today. We have two things that we're gonna talk about. I will be talking about Harry Potter and its incredible launch, but more specifically talking about deluxe editions. It seems lame. It seems like retail stuff. It is so important to your bottom line into revenue, and I think there's a lot of really interesting insight when you get into the empirical evidence behind deluxe editions usually forgotten about. And Chris, what are you gonna be talking about today? <cite>christopher_kaczmarczyk_smith:</cite> <time>2:36</time> I am gonna be talking about just an academic paper and Saunders to 1993 article entitled, allocated Efficiency of Markets With Zero Intelligence Traders. But more generally, I'm gonna be talking about the double auction mechanism. Or I guess more particularly I'm gonna be talking about the double auction mechanism and generally market efficiency and the power of markets and constraints and determining, equilibrium. Even if you have a bunch of morons running around, you could still come up with equilibrium results. And this is gonna be, I think this is gonna be an exciting topic for some of our reader. <cite>eric:</cite> <time>3:06</time> Is The implication here that MMO players are morons and just trading <cite>christopher_kaczmarczyk_smith:</cite> <time>3:10</time> Yeah, That's the whole point of game design because you have to design it so that even an SIL can't come in and ruin your game. It's gotta be foolproof. No. That's not the that's not the thesis. <cite>phil:</cite> <time>3:21</time> But before we begin, what have we been playing, guys? <cite>eric:</cite> <time>3:29</time> uh, My wife's been playing Hogwarts legacy. She is a mixed minder. She, likes it. She's I enjoy playing it. I feel compelled to keep playing. She's it's not the best game, but I just feel like I want to keep playing, so I dunno that's praise, <cite>christopher_kaczmarczyk_smith:</cite> <time>3:39</time> It's the magic. There's magic in every corner. <cite>phil:</cite> <time>3:42</time> There's still a lot of life left to milk in this franchise. It seems like there's pent up demand. I just What? What do you do outside the wizarding world? I feel like this is the Star Wars problem. The Luke Skywalker problem of people wanna explore the universe, but showrunners keep bringing us back to these key franchise places and I want to go away from them. I don't want to go to ta. Take me out of here. <cite>christopher_kaczmarczyk_smith:</cite> <time>4:02</time> I have two friends, two close friends who both purchased PS five s just to play this game. I think this is going to be a phenom like It's going to industry moving. <cite>phil:</cite> <time>4:13</time> Have you been co-piloting with her when she plays? <cite>eric:</cite> <time>4:16</time> A little bit, not too much. Personally I've been playing the cycle. An extraction shooter You recommended? Five rounds in. I have not run into another player. So it's just go in mine rocks, which so far has been like I'm paranoid the whole time that someone's gonna pop out and shoot me, but nothing ever happens. So I don't know if I'm just lucky or if this game is actually boring, but I'm gonna keep trying until I kill someone or die trying. <cite>phil:</cite> <time>4:39</time> People keep trying to make this market happen and it's just not happening. You had Call of Duty with DMZ recently. Battlefield took a shot at, when I was at Dice. You have Tarko, which just floats around, Russian web servers. And then you've got this weird little cycle game, which is out of, I believe a company in Germany has made it. And they made a huge pivot from like a PPV E mode into kind of this extraction shooter loop. They even changed the arts. Why isn't anyone getting this right? like you said, it is just bizarre. <cite>eric:</cite> <time>5:09</time> I think Tarof probably got it right for the degree of, how big their studio was, but yeah, no one's been able to replicate it, it seems <cite>phil:</cite> <time>5:17</time> I have been playing a shit ton of Clash Mini, which I can not stop talking about. I just got into Grand Master and if you're unfamiliar with Clash Mini, this is a game that is in soft launch from Supercell. It is in the auto chest genre, and it tries to take the auto chest formula and convert it to mobile. And if you haven't tried an auto chess game before, the setup is that you have a board checkered very similar to chess and that you place units with unique abilities on the board. And traditionally this has come out of mobile games, so they've taken mobile characters and mobile abilities and naturally converted them to this board. And then you'll try to position the characters. You might make a couple choices in terms of the economics of the character. So you can buy some, you can sell some and choose how many you wanna field. You were talking about super auto pets, which again, you got me addicted. I call it the cocaine of gaming. It's easy just to just sit there and just sniff a couple rounds. I've played a lot of super auto pets over the last three to three to four weeks, and I've say this is of my extension. I have super auto pets, and a Clash mini. The difference here is that they expand the board into more of a traditional auto chest. You don't have that two dimensional, one versus one at a time in super auto pets. You have a much more interesting board to think about positioning, which is really the key that Supercell wants players to master, is how you position your characters, how ag agro works, how targeting works. But the problem with a lot of these games is they don't tell you the rule structure. on how these things work. So you have to go into sort of these like Bayesian learning loops where you're trying to test on how pathing and targeting works. So will my unit go left or will it go right? To give you like a very small example, like you could have a unit at the front of your side in auto chest, and you would have two units equalist for on the opposing side to that unit. your unit has to make a choice of whether or not it's gonna hit the character on the lefts or on the right first. There's no clear UI or UX on what the character is going to do. And if you spend enough time, you'll figure out that it always chooses the right side. But if you didn't know that beforehand, it's absolutely baffling. Auto chess is still like just hanging around. It hasn't really been solved yet. It was like this flash in the pan after another mod that came out of Dota. And of course, if you're a valve, you make Dota. Dota itself is a mod. Half of their portfolio is a mod. It's easy to see why they wanted to strike gold and move fast and make a, an auto chess game put on the market. But it never really found success. And it's been about three or four years and it's great to see Super Cell, like trying to take a stab at this genre and see if they can't solve why this thing failed. I don't know if they have a strong game thesis yet. It seems to be screw the drafting mechanics, screw the economy management stuff. Let's just make it about positioning, which is great to see that someone does have a strong thesis and at least they're trying to falsify it. But I think this one still needs a little bit more time in the oven before it's ready for the mainstream. <cite>christopher_kaczmarczyk_smith:</cite> <time>7:58</time> Does this kind of a game get ruined if it's written on a blockchain? Because everybody knows exactly how everything's gonna work and there's no. I feel <cite>phil:</cite> <time>8:05</time> if you, you mean if you like publish the rules? <cite>christopher_kaczmarczyk_smith:</cite> <time>8:08</time> well I, one of the, one of the things about what, writing a game on a blockchain is that, everything is public information, which is cool. There's a lot of advantages to that. You can, as a player you can learn about the meta very quickly. For example as a researcher you can do research as a company, you get a lot of information. But, I'm like trying to I've been recently trying to think of all the genres where this type of mechanic kind of breaks or this type of information actually breaks the game. And it's, this seems like a good example of something where you don't want these hidden rules where you're doing this Bayesian updating in terms of learning the rule set or the, the, I don't know, the strategies set up these players. If you can basically just read the rules off of the blockchain, defeats the purpose of this game or not. <cite>phil:</cite> <time>8:47</time> What do you mean read the rules? Like I, the learning process is important here, but the learning process would be much better if I understood, if I had UI indication. It's the degree of difficulty in learning. It's like learning something in Spanish, versus English. <cite>christopher_kaczmarczyk_smith:</cite> <time>9:04</time> okay. I thought it was like a gameplay mechanic was the fun of figuring out how these things are gonna interact and maybe like these objects having mixed strategies or something like that. But that's just an annoyance. <cite>phil:</cite> <time>9:15</time> I would say that's a part of it. That to me is what a Roguelike game is and why the genre has exploded so much is that you're setting up these really interesting, I think, what are those machines that you sometimes see on YouTube where like you have dominoes and the Dominos will go down to hit a marble and no one hits <cite>christopher_kaczmarczyk_smith:</cite> <time>9:28</time> Rube <cite>phil:</cite> <time>9:29</time> Yeah. You I like to think. Both super auto pets in particular, and Clash Mini and many rogue like games where you're defining strategies on the fly is basically like building one of those machines and then trying to predict what's going to happen. I think the question is how do we make that an enjoyable experience? And I think you want clear and known rule sets to optimize against. So if I just said that there's gonna be a part of your machine when you're letting the marble go, that's going to have like anti-gravity or like the gravity's gonna be on the moon, or it's just gonna pick like a random planet, then that would be a huge problem because I couldn't accurately design machine around that. But if I knew it was going to be Mars every single time, I could actually plan around that and I can build a system that takes that into account. So I think you want to expose this type of rule set to players. I think you'd want them to know how targeting works. I don't think that's something that is useful for them to learn. Even though it's content, it feels like churn. Churn outweighs retention on that one. <cite>christopher_kaczmarczyk_smith:</cite> <time>10:26</time> Nice. Speaking of chopping up meat and turning it into sausages, I've been playing a lot of Monster Hunter Rise lately. Recently came out on, um, it usually came recently, came out on games Game Pass and I'm a huge Monster Hunter fan. I played two of the early games, monster Hunter, freedom Unite and Monster Hunter, like three or something like that, I can't even remember. And Monster Hunter World. And now I've played Monster Hunter Rise, and it just feels, it feels like the classic games. What I love about the Monster Hunter franchise is I think it's the type of game that if you describe to me, I'd be like, that sounds boring. You're fighting the same enemy over and over again and grinding for these items. But it's it's the perfect example of beautiful gameplay, just like addictive gameplay. Like the doing is the fun part. Getting the rewards is fun for sure. Getting, these cool sets of armor is like, is a huge bonus, but for me it's the perfect example of I just want to I'll go fight the same enemy over and over again and I'll be perfectly happy. So that's what I've been spending a lot of my time on. I usually try to be productive, like in the mor in the wee hours, be bef between when I wake up and work. Definitely gone on a couple of hunts hunted a couple <cite>phil:</cite> <time>11:33</time> This game has been enormously successful. They've been doing a lot of DLC packs for it. You just slay monsters, right? Monster Hunter's kind of in the Title I used to watch the anime, but that was a long time ago. <cite>christopher_kaczmarczyk_smith:</cite> <time>11:42</time> The anime was like so minor too but yeah it's literally like there are these major monsters that are the main point of the whole game. They're like these boss things, and you're, the whole purpose is to beat those up. Either slay them or capture them, take, extract their resources, extract their, hide or sh claws or whatever and make armor and weapons out of them. It's like, It's a basic crafting loop. And they've actually really improved on it over the years. They are, they're notorious monster hunters, notorious for just these. Incredibly tiny. There are the, there are these things called jewels that improve your weapons and your armor, and they give you special abilities. So like the attack up jewel is like the most, highly prized jewel in all of monster Hunter Lord or whatever. And this thing is like impossible to get it's impossible to get a full set of them to completely max out your character to the point where people don't put those items in their meta gear. So if they're like, here's the meta Lance build, they won't include some of those jewels because they're so incredibly difficult to acquire. And I always just think it's so funny because like you still don't see churn. Like the people who play this game are like, I'm gonna slay this monster. I slayed the same boring level one monster. or probably, tier two monster. Like 15 times, and this is a monster. I'm not gonna have to fight later. And I was doing it for this one stupid piece of gear just because it was so addictive like that. I just wanted to get it. So Monster Hunter's amazing. It's it's the perfect example of amazing game design. Not a, super economic game. But then again, I don't really play, I'm not a huge economic gamer. <cite>eric:</cite> <time>13:13</time> It almost feels like they took Eldon ring style combat, but put it in a live service loop where you can just fight the same giant boss monsters over and. <cite>christopher_kaczmarczyk_smith:</cite> <time>13:22</time> Eldon ring and souls like games. I I could be wrong on this, but I'm quite certain the first Monster Hunter game out came out before the first the first, uh, souls game. So, Demon Souls came out in 2007, 2006. I'm pretty sure Monster Hunter came out early two thousands, like 2004 2005. So it's very similar style, right? Similar game style, similar game studios like culturally very similar. So the gameplay is very similar and it probably explains why Monster Hunter and and Souls Games are my favorite two types of games period. But yeah, it's like it takes the best thing about Dark Souls and just repeats it over and over again. <cite>phil:</cite> <time>14:03</time> So if you have crafting loops and you're going after gear, but you are fighting bosses, it sounds to get most of the gear, I would assume it's very traditional Japanese boss design, where they're like phases. There's like a trick at each phase. Is there, <cite>christopher_kaczmarczyk_smith:</cite> <time>14:19</time> depends on the <cite>phil:</cite> <time>14:20</time> sharp diminishing returns if you already know the puzzle. You only solved that puzzle once, right? <cite>christopher_kaczmarczyk_smith:</cite> <time>14:26</time> So first of all, the materials that you need you would need to hunt multiple of the monsters to get the full, if you want a full armor set, for example, you have to fight the same monster at least five times. For most armor sets sometimes even more. And the more complicated monsters, yeah, they have these phases where you have to watch out for certain special moves. But in terms of, once you've solved the puzzle, uh, yes, there are certainly move sets and that you need to account for and the monster's gonna be doing stuff that you can preempt and try and react to. But it is non-trivial. You watch some of the crazy YouTubers slice up of the more diffi difficult monsters in the game. It's like an art form. But. <cite>phil:</cite> <time>15:04</time> So are there master, is there mastery just in terms of your ability to execute against a boss? <cite>christopher_kaczmarczyk_smith:</cite> <time>15:10</time> Oh my God. It's why I can't play with randoms online because like I go in with a very specific strategy and plan like I'm going to, so I play Lance a lot. And the lance should, they should target in their, in a group fight, they should target the core of the monster. And then you've got these bludgeoning weapons to should focus on the head and you've got these sharp weapons that should focus on parts of the monster that can be cut off. So you have these different strategies that you should follow. But if I play with randos, they're just go in whack the monster as much as they can dodge its attacks, but it's just, it's so boring. Um, so I actually typically don't play online with people unless it's somebody I know or it's somebody I know is really good. But you could get, you could force your way through the game without any real skill. <cite>phil:</cite> <time>15:54</time> Before we move on, do we want to talk about Yucca Labs and the sewer thing? <cite>christopher_kaczmarczyk_smith:</cite> <time>15:58</time> The game, the like, <cite>phil:</cite> <time>16:00</time> the poop one. <cite>christopher_kaczmarczyk_smith:</cite> <time>16:01</time> I honestly haven't looked much into it, <cite>phil:</cite> <time>16:03</time> All right. That's fine. That's fine. I was, so I would say it's worth looking into, we're talking a little bit on the discord, it's worth looking into because the way they distribute revenue is now like skills-based. It's how they distribute, divide. <cite>christopher_kaczmarczyk_smith:</cite> <time>16:15</time> that's <cite>phil:</cite> <time>16:15</time> it's like exclusive sourcing of NFTs. And I thought that would be interesting to talk about. Let's talk about some other point. It's like a poop game. Oh, one, one Other, one other rant. Just while we're here. So I hate dunking. I think dunking is bad. When you take someone's comment that is like wildly stupid and you point out how stupid it is. But sometimes, like there are bad habits that I have and sometimes people like, do they just put out red meat and they're like, come have a feast. And this one guy on LinkedIn recently posted that reminder, your target as a Web three game is mainstream gamers and not web three gamers. And I like, just blew a fucking basket. And it was just like, dude, dude, you're gonna go after mainstream gamers without Steam mobile console, literally any distribution channel beyond a browser. And you think people are gonna take that seriously? That's fucking suicide. You should not do that. Go after crypto natives. <cite>christopher_kaczmarczyk_smith:</cite> <time>17:10</time> Because there's a market there, right? It's like when mobile games came out, if they were like, we're only aiming for hardcore like World of Warcraft players. World of Warcraft, craft players don't really wanna play your game. So yeah I agree with the sentiment. The games need to be good and they need to focus on a larger audience. But no, no game in the web three space right now, except for maybe Sky Weaver and one or two others would be even considered by like a web two native. <cite>phil:</cite> <time>17:38</time> Yep. Yep. And I keep going back to just like the insane friction every time I wanna play crypto unicorns, is that oh fuck, I have to remember what my password was. I have to transfer money from Coinbase into this third party application, and then I have to go to open C, and then I'm hit with ASVs, and then I gotta export and import these things. It's an unbelievable amount of friction. I know it's gonna be solved, guys. I know Soon. Soon. Hashtag. But let's start, set some deadlines here. Do we meet next year, six months from now? Like where, when are these platform solutions coming? I, I'm sure they're in the works, so I just need to see them deployed. <cite>eric:</cite> <time>18:12</time> are tokens, games all blocked from mobile app stores? I know they're blocked from steam, but Skyway was on iOS. <cite>phil:</cite> <time>18:18</time> So it's the reader app though. It's the reader app exception that you can read digital assets on the app. So if you purchased off chain or not off chain, but if you purchased off platform, off mobile, you could see the cards that you had and you could use them. But if you wanna make a new iOS purchase, it's <cite>eric:</cite> <time>18:34</time> So it has to be a companion. It can't be the only way you, I. <cite>phil:</cite> <time>18:37</time> Correct. And the other thing is that you have to remember that for many of these NFTs there are know your customer laws, which means that you still have to go through a real person identification, which is also <cite>eric:</cite> <time>18:49</time> Not facilitated by the App store, so you have to do it on your own. <cite>phil:</cite> <time>18:53</time> Correct. And if you list the item while on iOS, that's another 30% Apple wants even on action house. It is atrocious. Okay. Shall <cite>christopher_kaczmarczyk_smith:</cite> <time>19:02</time> next phone is gonna be Android based. Sorry. <cite>phil:</cite> <time>19:04</time> That is not gonna, that's not gonna help the situation. What is that gonna do for you? <cite>christopher_kaczmarczyk_smith:</cite> <time>19:07</time> I'm not gonna support Apple. <cite>eric:</cite> <time>19:08</time> Google's got their own monopoly, right? <cite>christopher_kaczmarczyk_smith:</cite> <time>19:11</time> Yeah. But their fee is what? It's 10%. <cite>phil:</cite> <time>19:13</time> No, it's 30 man Google plays 30%. It's the same as iOS. <cite>christopher_kaczmarczyk_smith:</cite> <time>19:19</time> My whole world. <cite>phil:</cite> <time>19:20</time> This always amazes me when people enter this conversation, they assume that if iOS opened up, you would get better results. And it's we've already seen that. It doesn't happen. It's on Android. <cite>christopher_kaczmarczyk_smith:</cite> <time>19:29</time> Oh shit. Okay. Nevermind. It's a duopoly is the problem. <cite>phil:</cite> <time>19:34</time> It is a duo. There was a piece out recently that was basically looking at the Chinese app stores, which I thought was interesting on. There's a lot of like really weird distribution channels in the east. There's this one Korean company that's like trying to scale right now and has US support that. I'm curious to see. It's like Korea one or something. <cite>christopher_kaczmarczyk_smith:</cite> <time>19:51</time> That'd be nice to see <cite>phil:</cite> <time>19:53</time> yeah. No, I agree. <cite>christopher_kaczmarczyk_smith:</cite> <time>19:54</time> smaller transaction numbers. If the platform comes outta the east, it won't be attractive to the west because it probably isn't really bringing in as much revenue as they expect. I don't know how, I guess I don't know how much it would actually impact iOS Android decisions or Play Store and Apple store. <cite>phil:</cite> <time>20:12</time> Do we wanna talk about hog? <cite>eric:</cite> <time>20:14</time> Yeah, let's do <cite>christopher_kaczmarczyk_smith:</cite> <time>20:14</time> do it. <cite>phil:</cite> <time>20:15</time> Avalanche Software and Warner Brother games have launched Hogwarts legacy. This came out February 10th for deluxe edition owners. This is a game that's been in development for about five years out of Avalanche, and this is not the Swedish Avalanche, which is obviously far superior and is across the street from me in Sweden. This is a far inferior avalanche software that is, I believe, out of Utah, but un to their credit, they've just launched an insanely successful game, and they don't really have a. Track record of doing incredible software launches. They go back to 96 with their first title actually being Ultimate Mortal Combat three, which is one of the weaker mortal combats. And then they've done a bunch of licensed stuff. One of the highlights being Rugrats in Paris, the movie for the PlayStation in N 64. And more recently, they had been involved in the Disney Infinity project, which had a very interesting start and a very colorful ending. And so this has been their title. This is their only title really in the last seven years. And so they've released it to incredible critical success. I believe it's sitting around the low to mid eighties and meta critic, which is a pretty successsful outcome for them. And of course, when it comes to the players, it's been unparalleled in terms of their success. With Eldon Ring actually being a really close comparison point. For a game like this. And during the deluxe edition, I think the thing that's really interesting here is that they had an enormous wave of players join. It was about 500,000 CCUs or concurrence on steam. So if we're to translate that into D A U it's around 4 million d a U. So they had around 4 million d a U when they launched with just the deluxe edition, which I believe is about$80. And that has only grown in the day since they are now sitting at 900,000 CCUs, which is probably closer to around 8 million. Dow an incredibly successful launch. Obviously there's pent up demand for this franchise. I think we've seen that with the Fantastic Beast franchise. I think we've seen that with the parks, which now that I'm in la I definitely wanna go back to the Hogwarts Park. It's very magical to go there and see the trains and get into the little what is it, a hippogriff? Is that the thing with the elephant? Not the elephant, the eagle. What is hippogriff? <cite>christopher_kaczmarczyk_smith:</cite> <time>22:37</time> the hippo? Yeah. <cite>eric:</cite> <time>22:38</time> Buck beak. <cite>phil:</cite> <time>22:40</time> What is it combined with an eagle and a lion? <cite>eric:</cite> <time>22:43</time> I think so. <cite>christopher_kaczmarczyk_smith:</cite> <time>22:45</time> And a horse maybe. <cite>phil:</cite> <time>22:46</time> it's fun to get in that hippogriff. But the thing I really wanted to talk about here was the, first of all, the success of the deluxe edition cannot be understated for a product that's$80 or$90. That is impressive that they've been able to sell. It looks like maybe even half of their copies through the deluxe edition. And this is something that's gone really underdeveloped and unnoticed by monetization managers. Usually, we spend so much time on the sexy stuff like mtx, digital stores, real world pricing to hard currency pricing that we forget that for most games. And I, I say this with an asterisk or at least around half of games. The revenue still comes from box revenue. It still comes from brick and mortar retail stores. And when I say box, by the way, let me take that back a moment. When I say box, I just mean the initial purchase price still drives an incredible amount of revenue. And you're getting that from 100% of your players because you need to be able to buy that, that key to be able to have a license to play the core content of the game. And when you think about deluxe editions, we usually put them to the side like, oh, 50 more dollars for a marginal piece of content. That's not something that's really in my interest. I think maybe for a couple of us, that may not have been the case for Halo three, if you remember those master chief helmets that they gave away as a, the deluxe edition. And then there was a, there was some great ones from Call Duty Black Ops, where they gave around an RC car. Now, most of these end up being breakeven deals. They're actually marketing campaigns because the cost of physical packaging is so expensive. But when you think about the role of a deluxe edition, you know these things are still maintaining really healthy unit and revenue share. So we're talking, or maybe around 30% of units can be deluxe editions when they sell. And due to the markup you're looking at even more or at 50% of revenue can come from deluxe editions. You break this down to one other distribution, which is thinking about the upsell on the base price. So if you're selling a deluxe edition for$80 and the base game is$60, that's a$20 upsell. So if you just sum up that marginal revenue, we could be talking about 15 to 20% of total lifetime sales can come from this deluxe edition that is batshit fucking crazy. That is a huge number, and especially if you're a big franchise. And of course we compare this to the cost. What is the cost of putting together a deluxe edition game? It's virtually zero. It's virtually zero to put this stuff together because you usually are already gonna have content lying around or you're gonna be engaging in some sort of live service or some sort of post-release content. So just earmarking some of that content and putting it into the deluxe edition with a couple of marketing assets can be this insane cost benefit win. I think as a game economist though, you say, okay. Why is this the case? What's the theory behind this? And I think we've seen something similar happen in free to play games where a lot of games are going back to pricing things in U S D, or at least pricing some of their assets in the U S D or having some SKUs in U S D. So if you to check out the Fortnite store on the Epic Game store or on the PlayStation. you would see cosmetic bundles that were priced in U S D as well as in their virtual currency. And the reason I would argue that you go after this is that you're actually maximizing your funnel your impression funnel, because everyone who wants to download Fortnite needs to go to that page. So you're looking at a hundred percent impression rate, potentially depending on your ux for a lot of these virtual currency skews. And that's usually not the case with a lot of MTX stuff. You might only have 30% of players check out the store on a daily basis. And of course, that's why you might put in free login rewards or you might put in hard currency on that page, but having DLC on the steam page on those web stores really maximizes your impressions. And I think you can see a lot of that down funnel results when it comes to ultimately conversion rates. But even more so than that point. The argument I would make is that when you are opening your. it's a lot easier to ask for 20, 30, 40 more dollars at the point of sale because the marginal cost for asking for that additional money is less than the fixed cost of having to open up your wallet. So again, when you think about free to play, I'm gonna have to ask you to open your wallet 10, 20, 30 times potentially over the life of the product. But if you're thinking about a deluxe edition, I'm only asking that after I'm already$60 in at the point of sale. So the marginal cost of that additional upsell is really low. The thing I've been starting to play with is how far you can push this. So could I go to$120? Could I go to$200? How much of this game could I wrap up in a single edition? Why not just continually package all of the MTX bundles in some sort of sku? So call duty releases bundles every single week. Why not continually pack the last week's bundle in an accumulation of bundles such that whenever I purchase this edition of Call of Duty, I'm getting life to date bundles and the price reflects that as well. So I'd love to see a lot more creativity around this. If you look at Battlefield 2042, which I may or may not have been involved in, I think you'll see some interesting stuff when it comes to special editions as well. Only did we have a deluxe edition. We also had an ultimate. which was doing a lot of the pre-sale stuff, it was selling a lot of assets that you might sell down the line upfront. And I think we've seen this with crypto stuff too, is that asking for money at one point, rather than asking it for a bunch of different points, just makes economic sense in terms of reducing fixed cost. So I'd love to see developers play a little bit more with taking content they already have and wrapping it up into these additions. It's extremely cost effective and can bring in a significant amount of revenue. And to me, the other thing that I think is fascinating here is how far you push early access windows, because a lot of this is about minimizing marginal cost and re you know, removing the ability for some players to pay, or let's let's frame in perhaps a better light. Adding the ability for players who purchase additional additions to play the game early is just a question of flipping a switch is almost zero marginal cost for players. And honestly, it could be good for scaling technically the game. So if we're gonna talk about 72 hours for Hogwarts Legacy edition, why not push that for a week for an ultimate edition? Why not push that for a month? And not only that and please, Eric please let me know why this is the case. Why the fuck would you not have an addition for each of the houses? Why would you not have a Gryffindor edition, a Hugo Puff edition? Why am I not starting the game? Buying an edition and then getting a headmaster edition that is$200 and gets me like some dumble door Shit. This is the easiest win. Where is this marketing team right now? <cite>eric:</cite> <time>29:27</time> That's brilliant. I feel like, Pokemon was doing that ages ago, but it fits so much better on Harry Potter, <cite>phil:</cite> <time>29:31</time> Thank you. Pokemon is another amazing example of this. Do you remember Soul Caliber did this as well? You'd have, platform exclusive characters. <cite>christopher_kaczmarczyk_smith:</cite> <time>29:38</time> I want it to be known that you said Huggle puff instead of Hufflepuff but, uh, That's besides the point. Oh my God. Trying to talk about Harry Potter. Franchise <cite>eric:</cite> <time>29:47</time> you serious? <cite>christopher_kaczmarczyk_smith:</cite> <time>29:47</time> it for real. <cite>phil:</cite> <time>29:50</time> Are <cite>eric:</cite> <time>29:50</time> Maybe that's how they pronounce it in Sweden. <cite>phil:</cite> <time>29:52</time> Oh, it, Let's see how many people we can offend. <cite>christopher_kaczmarczyk_smith:</cite> <time>29:57</time> I have a hot take about, About why it was so successful this this Deluxe edition. I went into the the store to buy Hogwarts Legacy, and I could not find just like the regular game, I had to click through the different dlc and I was like, oh, there's the game clicked on it. And I was like, yes, I want the 69 point 99 edition. I don't want the 79.99 edition. And I attribute a lot of people just purchasing it to either a, they're uninformed. Like I know anecdotally I just know that there were a lot of non-gamers, or at least disen, not disenfranchised. People people who haven't been in gaming for a while who came in and were like, oh, I guess a game's$79 nowadays, and they just bought it. Or somebody goes in and they just see it for$79 and they click on it. I don't know how legitimate that take is, but I have been thinking this just. I've had multiple in experiences. I asked my friends why they bought the deluxe edition. They didn't even know why because they weren't they got it when the regular edition's com, uh, perfectly <cite>phil:</cite> <time>30:59</time> important about price sensitivity? Doesn't that just tell us that consumers are relatively pricing elastic for games? Like why are we not, why aren't I just setting the base? We've already been doing this, right? The price of a game has been creeping up even at adjustment for inflation is a weird story. There's a lot of weird, there's a lot of weird things that go on in that story. I think economists should definitely study that question, by the way, cuz with mtx revenue. with this two-tier revenue. It's there, there is an optimal price for these games. That is if you were to consider two games, one with a live service revenue strategy and one without a live service, the revenue strategy, I even if the box price still made sense. So if we're moving, we're abstracting away the idea that there's a freeto play world here, I still think the optimal price would be lower for the live service game because there's a revenue tail potentially. And then you wanna increase the gate, right? You're trying to compare the loss in revenue from a decline in price relative to the ARPU that you would gain from the marginal user who is now able to purchase that, if I've done my math correctly. So if you lower the price by$20, all the additional people that you bring in going from let's say 60 to 40, all those additional people who purchase it, they need to bring in more in terms of revenue than that$20 across the entire purchase price though, cuz you can't price a screen. Fuck that is a big task. Maybe I'm wrong on this one. <cite>eric:</cite> <time>32:17</time> there's a mild social virality component, but yeah, I think the bigger thing is the long tail worth 20 more dollars or not? <cite>christopher_kaczmarczyk_smith:</cite> <time>32:23</time> I think that it's the so you asked, are people price in insensitive? They are. When it comes to these like highly valuable, nostalgic franchises. So I think there are price. Think about how expensive it is to go to Harry Potter World. Think about how expensive it is to go to Disneyland or Disney World. One of these, right? Like people are highly price insensitive when there's nostalgia on the line when there's emotions on the line, when their children are on the line when it's their spouse or something like that. I made the decision to buy Hogwarts legacy after my wife said, you know what? I would play that. And she's not a very, she's not a video game player. By typically. And that was why I was like you know what? I would've paid$40 for it. I guess my wife would pay$30 for it. That adds up to$70, so let's buy it. It's like that, that, I don't know if you <cite>phil:</cite> <time>33:08</time> Becker? Didn't Gary Becker have the idea of a joint utility? Joint utility function? I've almost never seen that used that side of that paper, but I like this one. <cite>christopher_kaczmarczyk_smith:</cite> <time>33:15</time> I don't think it was additive So I can guarantee you that. <cite>phil:</cite> <time>33:19</time> a, It's a non, I guess it is a rival risk. Good. So I guess you can add utility, you can add marginal benefits together, right? This is totally fine. All right. Sorry. <cite>eric:</cite> <time>33:27</time> but Chris, what? What you're describing, actually that exact same thing happened to me when I bought Overwatch. I bought it for$60. The game was actually$40 and there was some upsell and I just clicked on the first button on the Blizzard store. Cause I was like, I trust Blizzard. And I later found out, oh, they upsold me for$20 and yeah, sure they got that extra 20 bucks outta me. And you're right, I would've paid$60 for it. But now every time I buy anything from Blizzard, I'm very careful about am I getting upsold or not? <cite>christopher_kaczmarczyk_smith:</cite> <time>33:53</time> Yeah. <cite>eric:</cite> <time>33:53</time> And so It might be pricing elasticity, but I think a big part of it is you just, the customer trusts you. They get, you get them to pay more and they stop trusting you after that. And they're much more price discriminatory. it's the same way. A lot of grocery stores don't raise prices. They ra would rather go out of stock because they don't want the consumers to stop trusting their prices. <cite>christopher_kaczmarczyk_smith:</cite> <time>34:11</time> Interesting. I do think there are two there. So you talk about trust which, puts me instantly into the behavioral economics mind mindset. I think there are two things, right? So that nudge is the one that I that I already brought up, is this idea of presets this ch this option that's given to you. You have, you can opt out of it, um, but it's already chosen for you. It's a nudge, right? Cl we've talked about you, oh God, what is that term? Paternalistic libertarianism or whatever. We've talked about that before. But the other one, the other behavioral econ thing Phil touched on, and it's it's anchoring to a price. So you've already got your wallet out. You've already accepted that you're gonna spend$70 and they say, how about 80? And. Oh, sure. What the hell I've already got. I already accepted the fact that I was gonna lose$70. So what's another 10 <cite>phil:</cite> <time>34:55</time> See, but I don't look at that as a behavioral story. That to me is just a classic Neo, it's a neoclassical story, right? It's about marginal costs and fixed costs. Like there are fixed costs, not literally for me taking out my wallet, just the physical act of doing it. But I think there are cognitive costs to subtracting revenue from, your account balance. So that to me is not a behavioral story, though I'm not anchoring to the price I it's a fixed cost. <cite>christopher_kaczmarczyk_smith:</cite> <time>35:18</time> That dude, no, that is, that's the behavioral thing, right? That's that's prospect theory. That's prospect just described as <cite>phil:</cite> <time>35:25</time> that's absolutely not prospect theory at all. This is simple. This is just shape of marginal cost curves. I don't understand where prospect theory comes into this. Prospect theory is about thinking about tail risks and thinking about the potential inequality between gains and losses. I don't think that has anything to do with deluxe edition pricing. <cite>christopher_kaczmarczyk_smith:</cite> <time>35:42</time> So you said there's this large, there's this large non pecuniary fixed cost associated with getting your wallet out. I just <cite>phil:</cite> <time>35:54</time> Perhaps. That's one <cite>christopher_kaczmarczyk_smith:</cite> <time>35:55</time> Okay. with spending money. I just, is that fixed cost? That expensive though? Is that non <cite>phil:</cite> <time>36:01</time> I think it is. And here's why. So I think there's the cognitive cost argument, which is a little bit hard to specify, right? But there is some cognitive cost. There is the physical cost. And I would say the other piece that I think is bigger here is that things change, right? So I might have expired credit card information. So that to me can be rather complicated. I might not have payment information into the portal. There's actually a lot of complication that can come with this. The PayPal, did I update the right PayPal address? There's a lot of bullshit that can come with payment is certainly not a frictionless process. So I would argue there's a lot of fixed costs. I just came outta Whole Foods cuz I am, I'm in Los Angeles this week. There's no other grocery stores. I'm guilt guilty as charged. They're doing this Amazon Palm thing where you just put your palm on the machine and that pays for it. But I would say, here's the other thing, Chris, do you use N FFC to pay Eric? Eric's gotta use NFC to pay. I hope. Hopefully you do like it. It is a big cost because people keep optimizing it, right? Like nfc. The reason NFC and Apple watches are successful is because payments paying for things sucks. <cite>christopher_kaczmarczyk_smith:</cite> <time>37:00</time> I am content with how easy a credit card is to use is as easy as I want it to get. Huh. <cite>phil:</cite> <time>37:07</time> You don't do the tap. <cite>christopher_kaczmarczyk_smith:</cite> <time>37:09</time> I'll tap every once in a while. I live in Indiana where we don't like we don't acknowledge the newest technological trends until they've been out for 10 years everywhere else. So tap is not even that popular. And I know for most people around the world that's like that, they've had that for five years. I just, I don't, I see where you're coming from. I don't buy that. This non pecuniary cognitive cost is worth like$10. Oh, I've got my wallet out. I can, ex I can explain this with Neoclassical Mar marginal cost <cite>phil:</cite> <time>37:36</time> So here's the thing though, right? Like how much does a behavioral explanation for this differ from like a neoclassical one? And this comes up in a lot of conversations I had about behavioral economics. I don't know if I need a theory like yes, changing defaults is expensive cognitively for people. There's a lot of resources you have to invest to a neoclassical story for why defaults are so effective to me. Is extremely powerful and explanatory. I don't feel like I need prospect theory to get there. I don't think that tells me anything. I think the experimental psychologists love to sell me some weird shit, but I, I don't see what the behavioralists are giving me that I don't get with Neoclassical. <cite>christopher_kaczmarczyk_smith:</cite> <time>38:09</time> It might surprise you to know that I actually completely agree with you. My advisor in my PhD was hardcore neo classs. He was like, there's nothing that a little bit of information asymmetry and neoclassical theory can't explain that behavioral economic theory can. And I would argue, I totally buy that argument that you can pretty much contrive any of these weird theory results with some sort of, asymmetric information or un, decision making under uncertainty. So I definitely agree with that. How many layers of complexity in probabilities do you need to get to, to achieve what prospect theory achieves? And if it does a good enough job, why not? But I'll concede that an anchoring problem. <cite>phil:</cite> <time>38:51</time> I actually, I want to eat a lot of my words on this. Of course, there have been challenges to the neoclassical paradigm. That cannot be explained with a lot of Neoclassical models. That's the whole point of Richard Taylor's 1987 anomalies paper was pointing out that neoclassical theory is really falling short to explain a lot of these things. <cite>christopher_kaczmarczyk_smith:</cite> <time>39:10</time> This will be a perfect transition into into the paper that I was gonna talk about. This will be beautiful. So the paper I'm gonna be talking about today is it's hands down one of my favorite papers, if not my favorite paper of all time. And I think it's the paper that really made the flip switch in terms of what a market is and what the power of a market can do. And it really is the paper that made the whole, Adam Smith invisible hand thing Makes sense to me. It was a paper I read in my PhD. For a computational economics course. And so the whole paper, the paper's title is Allocated Efficiency of Markets with Zero Intelligence Traders Market as a partial substitute for individual rationality. Now, this is a paper by Go Day and Sun out of Carnegie Mellon University 1993. So this is a pretty old paper. This is a computational economics paper. 1993 there, this is much harder to code in 1993 than it is to code in 2020. So 2023. Keep that in mind as we're going through this paper. This paper is. Set in the classic, the like classic mid-century stuff on markets. So you can think Gary Becker's 1960s two paper where he basically says he has a famous line, households may be irrational. And yet markets are quite rational. And it's this idea that markets can elicit seemingly individual rationality, but it's just because the market is making them behave the way that they should, in quotes. So the whole paper, it, it's set in this classic beautiful literature. If you look at the referees and the comments in their in their, uh, in their, what's the fricking word in their thanks section you'll see Colin Camera Robin Das, Stacy Jacobs Palfrey, Charles plot Vernon Smith. Like it is just. It's like this paper is fucking awesome. Yeah. So anyway. What do they do? They do something that part of me wishes. I was, an economist in the 1990s because there were these really sweet, beautiful, short, punchy papers that came out that were like here's this really cool elegance, simple theory. And I personally, I could be wrong, but I think that this is about the time when theoretical stuff started to get reigned back in it, or it was like, or no. This was like right before theory started to just get ridiculous out of hand in the early two thousands mid 2000 tens. But anyway, very simple idea. They take three different types of agents in a double auction. Now, I'm assuming most of the readers or listeners know what a double auction is, but for those who don't, a double auction is exactly how it sounds. You have a double auction. So the first person on in this double auction is the bid or the buyer, and the second person is the ask or the seller. So you have two people, two, a bidder and an asker putting in bids and asks into an order book. And those orders are transacted only when the bid is above the ask or vice versa, or at the ask. So this is how a transaction happens. This is how the stock exchange works. This is how lots of different N F T markets work in web three. So anyway it's how foreign exchange markets work. They want to test this, this, that's how shoe market, nice. They want to test the, this concept of market efficiency and the, the market kind of causing this rational rationality of the individuals in the market. So what they do is they have three different types of agents performing in this market. We don't care who's the bitter or the asker, we just care about the levels of rationality or the levels of intelligence. So they start with two coded AI agents. Now, these are zero intelligence agents, so they're randomly behaving. You would not even call them ai. Today AI is super fun and popular topics. So I'll throw that word out there. But basically little robots or bots. They're just little bits of code that are randomly selecting values to put in as a bid or an ask order. One of those bots, one type of bot. And there are two different, the bots don't interact with one another. There there's two experiments with these robots. The first one is with these robots that are perfectly zero intelligence. So they perfectly randomly choose their. Their bid or their ask, and they don't have any constraint, meaning they could put in a bid that would lose them money. Likewise, they could put it in ask that would lose them money. So they have no budget constraint. And so these are the zero intelligence agents with no budget constraint. And then you have another type of bot, and those are the constrained bots. And they are zero intelligence with a constraint, meaning they can't lose money, they're not allowed to put in a bid that would lose the money. They're not allowed to put in an ask that would lose the money, but they're still randomly behaving. They just have to adhere by the market. The market mechanism the double auction market mechanism and their budget constraint. And then the third group, Are the human traders. So they take these three groups and they run three different simultaneous kind of experiments you could think of. They run one with those zero intelligence unconstrained agents, one with the zero intelligence constrained agents, and a third with the human agents, presumably. And the idea is that if you can measure the difference in performance between the UNC. Zero intelligence agents and the constrained ones, you'll be able to identify what is the, what's ha what is the, what's the additional gain in market efficiency from the constraint. And then by adding in the human layer, you can compare the human performance, market efficiency. So you know how much profits were gained, how much How much did the price the equilibrium price differ from the theoretical equilibrium price? So you can do these measurements and you can try and measure what did, what was the difference in those the human agents and the zero intelligence constrained. So you have these three layer layers of intelligence, three different markets that should have very distinct levels of efficiency. And they're trying to identify what is the gain in efficiency, in market efficiency from someone being a rational agent. Meaning, what's the difference between these zero intelligence agents that have a constraint? And the human intelligent agents who learn they have constraints, they're smart, they, they use strategies. They're not just randomly putting in bids. And miraculously we don't wanna worry about too much about the zero intelligence unconstrained, because those are just literally that's just white noise. Their bids are all over the place because they have no constraints. So there's really no market there's no mechanism there. Once we get to the zero intelligence constrained agents and compare those to the humans, uh, the humans are able to they have about like 99% market efficiency, meaning they're extracting all the profits and all the consumers and producers surplus that's available in the marketplace. And they're finding the price, the equilibrium price very quickly. The equilibrium price is of course determined by the available budget constraint of the individuals. There's a range of values that they're allowed to put a bid in at, and they have a certain amount of money, and there's an equilibrium price because you basically, from that, you can plot the supply and demand curves and you get the equilibrium, the theoretical equilibrium price the bots with the constraint. Performed within 90% market efficiency, and the humans performed with a 99% market efficiency. So the bots were not far away from the humans. Humans were extremely efficient sorry the bots were ex extremely efficient miraculously. And they attribute this to just this beautiful, the beauty of the double auction. So that's the main finding. They run through different types of markets with different budget constraints. So you can see like what is a more, so the humans do better in markets where the price is harder to find meaning that the budget constraints are a little bit more complicated and there's less wiggle room in terms of the, where the buyers can put their bids to find the equilibrium price. But it's just a really cool paper, simple, elegant, took me longer to explain it than it would take to just read the paper. So anybody who can get their hands on this paper, certainly do it. It's it's absolutely wonderful and it emphasizes the power of a market. It emphasizes the usefulness of running simulations, and it emphasizes the importance of designing an economic system, especially in the context of games, right? That is not really, that just has these natural market consequences that elicit this rational behavior. You almost can't game a double action. <cite>phil:</cite> <time>47:08</time> So if I'm understanding you correctly, Why isn't the lesson here really about the ability for humans to optimize against their own goals? Because they were given reservation prices, I would assume, in the experiment. So they were incentivized to bid in certain ways, and so they just had to solve the problem of how do I bid in such a way that maximizes my surplus, right? Like I assume that they were, if they got the correct answer, if they got the right bid that was given to them, they'd be rewarded financially. Usually that's how you incentivize participants in these lab experiments is that you make sure that if they do the right thing, if they invest more cognitive resources and solving the right thing, they'll have more to stand. The more to gain. But wouldn't the lesson here be that, relative to to your point, rule-based agents? You said it was 90% for the simple rule-based agent efficiency, and 99% for humans. mean, That to me is a huge gap. Thousand nine percentage points. <cite>christopher_kaczmarczyk_smith:</cite> <time>48:00</time> Sure, but that's nine percentage points. Like from going from literally zero into if you were to take a market and you were to put a bunch of, if I were to go and, have a, an agent randomly assign, in the stock market. The idea here is that a bunch of randomly behaving this, maybe the stock market's a bit extreme, but a bunch of randomly behaving agents would come to a similar equilibrium price and have similar market efficiency as a bunch of super, super hard thinking rational agents. I think the gap, you're talking about like a 99%, gap in intelligence and a 9% gap in efficiency. And it's not, it's not to, the point of the article is not to, in my opinion, not to ar emphasize, the stupidity of focusing too much on rationality. I think it's the, it's to focus on. What a good mechanism is. And I'm of the belief that the double auction is one of the most profound discoveries that has come out of economics or really came out of mathematics. But that has been studied in economics in modern times. <cite>phil:</cite> <time>49:00</time> So this really boils down to the effectiveness of bidding and asking for items and how that reaches equilibrium and how important that is for MA maximizing the preferences of all the participants. <cite>christopher_kaczmarczyk_smith:</cite> <time>49:09</time> Yeah, I, if you read the paper you'll see some really cool graphs that kind of display both the budget constraints or in other words they'll actually display the demand and supply curves right next to the actual price discovery over a 30 period experimental time period. So there are like these 30 periods that the traders are executing a trade on in each period. And you can see the market price achieve that equilibrium price that you can see in the supply and demand graph. So it really It's all about the fact that a constraint and a double auction, you combine the, that market mechanism and you end up with this just pretty robust results. That is the equilibrium price. And this is a finding that's been, that had been discussed in other papers in particular, the 1962 Becker model or paper. This was just the from what I could tell, this was one of the first times where, They teased out how much of that is the rationality of the agent and how much of that is the, the robustness of the model. Because people had studied the model and people had studied the agents, but they hadn't combined the two. So these guys, they went ahead and they said we have this experimental data from human agents. Run another experiment except for with zero intelligence agents. It's a really fun experi. It's a really fun like coding assignment. Actually, because you get to kinda recreate this thing. It's a lot easier to do nowadays than it was back then. <cite>phil:</cite> <time>50:27</time> And benchmarking against simple agents with clear rules to me also provides a wonderful baseline to assess how the humans performed because it's very easy to understand how the agents performed. It's not a black box at all. It's a simple heuristic they have to follow, and it's a very simple game to play. <cite>christopher_kaczmarczyk_smith:</cite> <time>50:45</time> This is a technique I use a lot in my daily life. In my job I'm constantly, I would say typically, especially, if I'm crunched for time, I'm usually just using zero intelligence agents or, very low intelligence agents because half the time it gets the job done right? Like you said. Also, you can use it later on as a heuristic measure to say, oh, this is a little bit above, this is a little bit below. Why is that <cite>phil:</cite> <time>51:06</time> So walk me through just like a purchase decision that you would make with a zero intelligence assumption. <cite>christopher_kaczmarczyk_smith:</cite> <time>51:10</time> So in this so maybe walk through like an example of what this model would look like. So you've got I think it really helps to visualize the order book. If you've got a zero intelligence agent, you, as long as they have the budget, they might throw out, let's say the equilibrium price is 50 and they can put a bid from zero to one. But they have a budget constraint that says you only have$50 to spend. So they have to put a bid in that's less than or let's say that they have$70 to spend, they have to put a bid in that's less than$70. If they, let's say they throw out a bid that's that's$20, so that, that's a bid. So that's really low. Most of the sellers, most of the sell orders are not gonna be that low. What the age, what the designers have done is they've created, they've designed the budget constraints such that, uh, they tend towards the equilibrium. So they say you're not allowed to put in a bid at,$1. You're not allowed to put in an ask at$1 as a seller because you'll lose money and you're not allowed to lose money. It's, I guess you could get into a conversation about whether this is rationality or, no person is gonna lose money. So that's more rationality than it is a budget constraint. <cite>phil:</cite> <time>52:20</time> So how does the agent though work? Like you were talking about sometimes you, you assume that zero knowledge agent. How does that help you? What would that outcome be like? What would be a situation where you're like, okay, I'm going to assume a zero knowledge agent to help me make this decision. What does that decision, and what does that agent do for you? <cite>christopher_kaczmarczyk_smith:</cite> <time>52:36</time> So typically this is like outside of a market. For me personally, I, any sort of process that I wanna repeat over and over again I'll have some sort of stochastic process, right? They're they're going out and they're mining for this thing. They, every once in a while they get, they're 5% above the mean. Sometimes they're 10% below the mean. There's this stochastic way that they acquire these resources. And that's not the part that I'm interested in. The part that I'm interested in is I cut them off from certain parts of this this game map, what happens? What does their behavior look like? So it's more looking at, this like ideal world where their behavior is limited. Yeah, I, that's how I would use it. I, it, I've run the, I've run the pricing thing. I think it's interesting. The problem with using this practically is so the way I use it is much more, is a much more you could think of it like as a coding version of machinations where you're just, you're trying to maybe with some slightly more intelligent agents where you're trying to see how does your, lottery system play out over time or something like that. The problem with using something like the golden Sonder model. In practice is that they make assumptions about the budget constraint. You can't make assumptions really about the budget constraint of your players because you don't really know what those look like. So that's the one issue. You don't have control over the budget constraints in a non-experimental setting, so you couldn't really use this in practice. Like nobody, nobody can predict what their players are willing to spend on microtransactions or what the secondary market value of something is gonna be because they're, they ran a simulation with zero intelligence traders. I guess you could make some assumptions about the the budget constraints of the consumers. I think the powerful thing though is that like you could get this, you could achieve the results without the utility function. Technically, according to this paper, if you did know the budget constraint, you could achieve the results. You could figure out what the equilibrium price is without the utility function. And at this point, I'm just talking off the cuff. <cite>phil:</cite> <time>54:28</time> Okay. I think I know what you mean, right? But if I have, so if I have the budget constraint and I have, what do I have? So what would you give the agent? You give them a budget constraint. What else do you give them? <cite>christopher_kaczmarczyk_smith:</cite> <time>54:41</time> Given them a budget constraint and the ability to put a bid in, in this paper. <cite>phil:</cite> <time>54:46</time> a utility curve, why do they bid? <cite>christopher_kaczmarczyk_smith:</cite> <time>54:47</time> so, so you mean like without the des, without the desire to, to earn. So in this context, the the agents are putting in bids. They're bots. the, the <cite>phil:</cite> <time>54:57</time> they've been given reservation prices that they're trying to maximize against. like they're like, they're submitting towards a purpose, which is that they wanna win. Like they might value it at$70. They have a$50 budget constraint. I need to maximize that gap between <cite>christopher_kaczmarczyk_smith:</cite> <time>55:12</time> so so you could think of it like this. Let's suppose that they do have this let's suppose that they do have some sort of optimization behavior, and that is to put in a bid because obviously they wouldn't put in a bid otherwise. This paper is saying if you're gonna put in a bid and you do it randomly and you're a zero intelligence trader, the equilibrium price in the market efficiency would be very similar if you do that randomly, then as if you were to be, optimizing some sort of budget constraint or using market information or something like that. So I think the I see what your point is. I think it's, point. I'm just not <cite>phil:</cite> <time>55:48</time> Like I, I love when they benchmark against the simple heuristic agents. I always feel like you get something interesting outta that. that's what I've been trying to do in a lot of my recent work, my, game econ work is okay, what are we benchmarking against? Have you ever seen this wins above replacement stuff like that's come out of baseball? <cite>christopher_kaczmarczyk_smith:</cite> <time>56:03</time> Wind's above replacement. <cite>phil:</cite> <time>56:05</time> So this is something that I'm actually, and it's there, there's a football equivalent called expected expected points. And you can take the, this has been the workhorse of football analytics and it helps get at the value measurement problem, which is really the key, I think, to all saver metrics is how do we assign value to things? And so they've developed one for football, which is more interesting than the one for baseball, in my opinion. Baseball, very much a one-on-one. despite the fact they're outfield, there's very low coefficient on the success or failure of a batter. But when you think about football, everything's moving at once. Very hard to measure things. Expected points per attempt is a really clean way to do this. And they do it at such an aggregate level. And I think you can take this model, and I think what you can do with it is you can apply this to almost any game where there's a lot of simultaneous action. And I think we can get some really interesting results. And I wanna write about it very soon. <cite>christopher_kaczmarczyk_smith:</cite> <time>56:53</time> Interesting. So you think there's a so this is almost an alternative to some of the skills based measurements that they have to, <cite>phil:</cite> <time>57:00</time> the problem I'm trying, so if we go back to Clash Mini, for example, So there's a board, there's all these units, they're moving simultaneously. Once the round starts, it is, One thing to go ahead and measure, like damage taken and damage received and to assign that, value. But of course, there's so many different variables that determine the success or failure of a particular match. So to give you one example, like there's this thing called rotation speed, which is rather important for a lot of these units because when they're pathing or trying to find a new target, they have to twist left and right. They have to be facing the direction that they're headed. And so each unit will have a different rotation speed or the speed at which it takes them to go from center to let's say, right? So if they're doing a lot of rotation, it can be rather complicated, which means they're not gonna be in combat for a really long period of time. So there's like all of these variables beyond just like damage Dell and damage received. There's also stuns. So there's like all this opportunity cost stuff. So I think you could take this expected points per attempt, and I think what you could do is you could back into a value for each of the units and use it for balancing, like I think this is good for like card games. <cite>christopher_kaczmarczyk_smith:</cite> <time>57:58</time> Interesting. Okay, so balancing the actual cards <cite>phil:</cite> <time>58:01</time> So to give you an example, yeah. So if you think about magic at the gathering, you're trying to reduce your opponent's life total. And one of the ways I could measure the card is by like the expected points lost expected points I've inflicted on the opponent expected points that this card does for it. And like that would be a use of this. So basically you just fast forward, this is how the expect the football one works, is that you just fast forward to the end of the play and you say, okay, how many points do they score in a similar situation? That's where all of this is really derived. And then you control for a bunch of other. <cite>christopher_kaczmarczyk_smith:</cite> <time>58:28</time> Interesting. So what I thought was you would take two decks that were identical except one didn't have that card. You would play a thousand games with each and the expected, whatever the difference is between the outcome is the, is war or is this wins above replacement? Is, that's <cite>phil:</cite> <time>58:42</time> Yeah, so sorry, I confused you. So there's wins above replacement, which is a baseball metric. And so that one is trying to measure how many wins this player will bring to your team above an average player for that given year. And it's extremely predictive in terms of the number of wins that a baseball team will have is that you just sum up all the war for all of your players and then you get your total wins, right? So it's been extremely predictive. So it's okay, why don't we have anything for football cuz of the simultaneous problem. Okay, let's do expected points per play. That'll be much better. So it's an expected, it's expected value function. Once you take a given down distance team score, like all these different factors, here's how likely it is on a discounted basis that you'll go ahead and score X amount of points. So like the expected points for a given play at an opponent's five yard line versus an expected points per, an attempt when you are at your own five, five yard line are radically different because the likelihood of you getting those five, those seven points is significant when you get to their five yard line. And so you just try to fast forward. Based on all these things, how many points did you score on an expected value basis? <cite>christopher_kaczmarczyk_smith:</cite> <time>59:47</time> So you're saying to bring it back to the zero intelligence thing, you're trying to use that zero intelligence to me. There's, there, there's a counterfactual here somewhere, right? There's a, there has to be in order for you to be able to measure this. Correct. So you're saying the counterfactual is the zero intelligence agent? <cite>phil:</cite> <time>1:00:03</time> We mean connecting the two things we've just been talking. <cite>christopher_kaczmarczyk_smith:</cite> <time>1:00:06</time> Yeah. Or is there no connection <cite>phil:</cite> <time>1:00:07</time> I'm just, I'm basically just thinking about like simple models. No, there isn't, there's no benchmark. You're right. I fucked that one out. The wins above replacement is much more apt for that cuz that's like a simple, like what is the replacement player? How can we think about that? How can we measure that? And like that to me is really consistent cuz like that you're always trying, to me, you're trying to measure something away from some sort of like base choice. Like it, it's an opportunity cost point, but I think it's valid. How valuable is this relative to this next best option? Not just how valuable is this thing in an abstract. <cite>christopher_kaczmarczyk_smith:</cite> <time>1:00:36</time> Yeah. Interesting. I'm still, yeah maybe reading the paper would make a little bit more sense, but I'm still fighting with this idea that I, so you're saying that because the programmer programmed the agents to put in bids that was inherently making them. Like forcing them to have preferences forcing they, they have a preference because they're putting in bids in the first place. So you're almost saying you're like, debunking the entire paper <cite>phil:</cite> <time>1:01:02</time> No. So I guess what I was confused by, and this will all clear up, I'm sure once I read the paper, is that what you said was that you can get these results without knowing utility curves of the agents. And my counterpoint to that would be if you don't know the utility curves or they don't have utility curves, that would be actually two different things. But if they didn't have utility curves, then why are they submitting a bid at all? There's no reason to submit a bid. Like they have to have some sort of reservation price for the, for them to act. <cite>christopher_kaczmarczyk_smith:</cite> <time>1:01:33</time> Interesting. So <cite>phil:</cite> <time>1:01:34</time> you have to like why else would you act? You're trying to maximize benefits relative to costs. <cite>christopher_kaczmarczyk_smith:</cite> <time>1:01:38</time> for sure. For sure. And I think that there's a fundamental issue, there's a fundamental disconnect between an experiment and real life. I think like the practical or the application is I guess that it doesn't work because the application, in application, you're actually just ob observing weak axiom or reveal preference, and you're not actually, you're not like really doing this experiment about, budget constraint versus utility. So perhaps it doesn't really extend outside of this, because when we look at the behavioral of people in a market and we try to make assumptions about that we're really, we're not observing their utility function directly or their or their budget constraint, which is re we're observing kind of the result of the combination of those two things which is warp. Yeah. That's interesting. So this is, this paper is what Phillip is saying is that this paper is not useful outside of a purely theoretical. <cite>phil:</cite> <time>1:02:29</time> no, let me just read <cite>christopher_kaczmarczyk_smith:</cite> <time>1:02:31</time> and it's my favorite paper. of all <cite>phil:</cite> <time>1:02:32</time> me, lemme just read the damn paper and make sure I understand it. This is completely on me and my lack of alone understanding on this one. This is, I just need to read it and get into it without <cite>christopher_kaczmarczyk_smith:</cite> <time>1:02:42</time> I I think it's, no I think it's a valid point. Like I don't, there, there's perhaps, there are no lessons to be learned from this paper and I certainly wouldn't suggest I, I certainly wouldn't suggest. Someone trying to take lessons from this paper and apply them and say oh, I can make predictions about, that's of what my point was. You, you can't take this type of framework and say, oh, I'm gonna go use this and I'm gonna make assumptions about my players' pre my players' co budget constraints and make predictions. Let's take a web three company for example. They've got their little, their crypto asset and there's a market for that crypto asset. And they say I know my average player has, this, uh, willingness to spend on my market. I know they face this is what their budget constraint looks like. I'm gonna run a golden sumder 1993 experiment. I'm gonna just put in those budget constraints. I'm gonna put in the bid in the sell the buy and the sell side budget constraints to create the demand the demand and supply functions. And I'm gonna run an ex a zero intelligence experiment to try to find the equilibrium price. That doesn't work. It doesn't work like that. So it's not, it's, you can't apply this outside of, the context of this paper. Um, so when I said I, I used this in my daily life, that was perhaps a very overly strong statement. What I meant to say was, I love this paper. I f I find it super fascinating and I do like to run, I like to run my, my if I have any sort of if I have any sort of simulation, I always start off with zero intelligence randomly behaving agents. For example, <cite>phil:</cite> <time>1:04:11</time> I would be interested in learning more about how you're setting these up, by the way, in a separate time and how you're setting these up in Python. <cite>christopher_kaczmarczyk_smith:</cite> <time>1:04:17</time> yeah. We sh we, we should talk about that because I think it's a really interesting It was some of the stuff that I started to get into when I was at Meno, but obviously I didn't have the opportunity to, to finish that out. But, you start off, you've got a, you've got a player who has 10 resources and five items they can craft with those resources. know, It's really fascinating because the first thing is you say, okay, they go out and they randomly find one of these things and they have some probability of en encountering them, and that's the rarity of the resource. Okay. Simple enough, they get that resource and they mine for 10 hours. In Python, you just don't, you go like 10 times, 60 times, whatever. And, got this resources now they have to make a decision about what they want to, what they wanna build from a pure machination standpoint. And, you're just trying. You're just trying to do this process a million times. You don't care what the preferences are. You're just trying to do this thing a million times. From a pure machination standpoint, you're gonna have some distribution of those crafted assets that people are gonna have crafted. But those agents were randomly behaving In practice. Some of those items are gonna be better, more desirable, more important for one reason or another. Even if you as a game designer, were like, these things are just as valuable as one another. There are gonna be reasons why one of, one of them is one or two of them are more important than other ones. And that's gonna cause this kind of breakdown. What I was suggesting or what I was getting at is I love to have those, like those, almost time series simulations saying, if you, if everybody just randomly behaved, this is what it would look like. And then I can compare from that. That's my baseline, and then I can see, oh this item is being like overly produced in quotes. Why to me that's that's the value add. <cite>phil:</cite> <time>1:05:51</time> Game Economist Cast episode five in the can talk soon. Good luck, man. <cite>christopher_kaczmarczyk_smith:</cite> <time>1:05:56</time> Good See ya. Bye.

Show Notes

Episode Transcript

Other Episodes

Episode

E33: Halo's Economist & Player Price Experiment Complaints? (w/Dr.Jason Arentz)

Episode

E02: Fork & Knife

Episode

E22: Airdropping The Missing Web3 Palworld Take