Episode Transcript
[00:00:00] Speaker A: Of it's not a utility calculation for God. Yeah, exactly. Like, I mean, I mean he would tell, he would tell Anne Frank. You know, you would tell the family holding Anne Frank. Like you got to tell them where they are. Like that's what philosophy professors will also bring up as the example because it's the most extreme and illustrates. Illustrates his philosophy.
[00:00:15] Speaker B: Yeah, well, you know, I guess I will, I will not evaluate game design decisions and content ethics if he doesn't like magicians.
[00:00:22] Speaker A: Let's start with utility. I don't understand what it even means.
[00:00:27] Speaker B: Everybody has some kind of utils in their head that they're calibrating.
[00:00:31] Speaker A: There's hardly anything that hasn't been used for money. In fact, there may be a fundamental.
[00:00:35] Speaker C: Problem in modeling what I want to model.
[00:00:41] Speaker A: Game economist cast episode 45 if the notes are finally correct and I haven't gotten it wrong for the first time in Game Economist Cast history we are back.
Three stooges are here.
[00:00:54] Speaker C: Musketeers.
[00:00:55] Speaker A: Musketeers. Much, much more professional, much more upstanding. And we have an exciting roster of topics. We will be talking about skill based matchmaking, but not in the way we did previously as a Call of Duty white paper, but as an economics problem and how economics can solve. I would argue compensating differentials. Eric will be talking about.
[00:01:16] Speaker B: I'll be talking about surveys. Just, you know, economists famously dislike survey data and you know, it's got a bunch of problems, but I want to do a deep dive into like what it really means and like how to how use it and how to think about surveys because you definitely cannot use them at face value.
[00:01:29] Speaker A: One of the most popular tools in games. Did you enjoy the match? Match of the Gathering arena loves to ask me that question.
[00:01:35] Speaker B: A little thumbs up, thumbs down after.
[00:01:37] Speaker A: Yeah.
UXR surveys often get tossed around gaming studios. But before we do, let's talk about what we've been playing.
[00:01:45] Speaker B: Got a selection of good things on sale. Stranger. We have been playing wordle.
We've all been playing wordle. Everyone but Phil. Phil hates wordle for some reason I think is a great game because he doesn't know words.
[00:01:58] Speaker A: Well, first of all, I do. I do fucking hate word games. I think they're easy. I think they're easily gamed. You can basically put them all into chat gdp. There is no sense of progression in any of the word games. Your skill level is always constant. You rarely get better. So it's a game without progression. It's literally just solving the same puzzle over and over again.
[00:02:18] Speaker B: All right, I'm going to hit some of those points. One is, you said it's easy to cheat, but like, okay, but you're not playing the game to win. You're playing the. You're playing to play the game. You're not. Like, what are you fucking. Do you cheat when you play chess? Like, I don't understand why that's a problem.
[00:02:32] Speaker A: Yeah, but chess, you're trying to solve an opponent. It's a moving part.
[00:02:35] Speaker B: It's super easy to cheat at chess. Just play everything into a chess bot. You win every time.
[00:02:38] Speaker A: Yeah, but like, I don't. But dominating, Dominating another player is where I get the joy in chess. Playing against a bot and winning against a bot would not be.
[00:02:46] Speaker B: So you never, like, just solve puzzles on your own or like riddles that don't have a human opponent? It's just like a abstract puzzle you're solving.
[00:02:52] Speaker A: Only if it's an economics puzzle.
[00:02:54] Speaker C: The incentive to cheat is much higher in chess, I would think, than something.
[00:02:58] Speaker A: Like wordle, like an economics puzzle has.
[00:03:00] Speaker B: So it sounds like your problem is that it's not competitive.
[00:03:04] Speaker A: My problem with it is that the intrinsic problem solving process of a word game I find to be cognitively draining with little marginal benefit at the end of the. I don't get a sense of satisfaction once I've solved it. Okay.
[00:03:18] Speaker C: You don't like crosswords?
[00:03:19] Speaker A: No, I fucking. I fucking hate crosswords. I just want to throw crossword across the table.
[00:03:23] Speaker B: Yeah, I'm not a huge fan of crosswords either.
[00:03:26] Speaker A: I hate all word games.
[00:03:27] Speaker B: What was the other point? Skill progression.
[00:03:30] Speaker A: Okay.
[00:03:30] Speaker B: You just. If you do them more, you get better at it. Is it. Is this not true for any.
[00:03:36] Speaker A: I think the relative gains to the N +1 attempt are relatively small on your X axis relative to your Y axis being your skill. So as you, as you play more and more games, I think it's relatively flat. I agree with you. There's probably some slope, but I would imagine that slope is very small.
[00:03:54] Speaker B: Okay.
[00:03:55] Speaker C: Yeah, but I mean, that's true of almost any game.
[00:03:57] Speaker A: Yep.
[00:03:58] Speaker C: I was going to say my progression on a lot of games is similar. Like, it's logarithmic. Like, I get that first early bump in the first couple of games, even with chess. Like, when I first started playing chess, 400 grinded, you know, for the first month or two, then I got up to 700 and then, you know, never, you know, never got over a thousand in all of my years of playing.
[00:04:18] Speaker B: But, like, you'll play like roguelikes, like, like Super Auto Pets. It's not exactly roguelike, but, you know, it's. There's no goal, right?
[00:04:25] Speaker A: There's no.
[00:04:26] Speaker B: You just play the game over and over and over and you get a little bit better, but you don't get that much better. Like, it doesn't have a lot of the same problems you're describing.
[00:04:33] Speaker A: So to me, I go back when it comes to roguelikes, I go back to the. The Goldberg machines. You know, when you have the domino down and hit the hot wheels, which goes down and hits the toilet bowl, which makes the penny drop. A roguelike, to me is running that simulation in your head. And that, to me is intrinsically satisfying. Figuring out how all these different pieces are going to fit together and trying to play that simulation out in my head. Exactly. Like, why I think all of us on this podcast, especially like every economist I've met that does play games, they almost always will have at least played Magic of the Gathering for a serious period of time. Because that puzzle solving simulation process of. Okay, I have a theory of the deck. I have a theory of this Goldberg machine. Now I want to see how that Goldberg machine works when it faces friction and whether or not I can solve that problem. And that isn't the case when you play any word games. I don't think there's that fun machine you're figuring about in your head.
[00:05:26] Speaker B: No, it's a different type of puzzle. I think you just don't like word puzzles, it sounds like, which is a fine stance to have.
[00:05:34] Speaker A: I don't like them and I don't think they've solved the game design piece, particularly wordle. So there isn't a lot of artificial boosters that you can get there.
[00:05:42] Speaker B: There's not a lot of meta progression, that's for sure.
[00:05:44] Speaker C: Right.
[00:05:45] Speaker B: There's. There's not a lot of systems where you can increase numbers, which I guess from your perspective, right, as like a, like a mobile game forever game kind of guru, right? There's not much to work with there.
[00:05:54] Speaker A: But I think there could be a lot to work with. And don't get me wrong, there are popular word games that are free to play, which I'm sure writers or listeners of the PODC we'll come back and slam in our face, which is completely fine. But at least when we take word of one of the most popular word games in the world, they don't have any. There's no, there's no vertical progression. There's no boosters. There's nothing. There's nothing there. That's Assistance and alley oop.
[00:06:16] Speaker B: I don't think it needs it. I think that's part of the appeal for a lot of the New York Times audience is that it doesn't have all those bells and whistles.
[00:06:22] Speaker C: I did, however, see I saw on LinkedIn a couple of weeks ago, and I'm. I'm trying to. I'm racking my brain to remember what it was, but it was like a word game with boosters. So if you hit a word with two. Two vowels or two consonants in a row, two of the same consonants, you would get like a 2x boost or you would get an additional guess or something like that. So they did actually kind of have that in round progression. It was somebody's like, Covid project that they just coded up really quick. But, you know, so there, I think there are ways that you can do it. Certainly within round progression for word games is possible. I mean, I've got two word games on my board game shelf that are very interesting. There's multiple mechanics. It's not just guess and check of the word, but yeah, I mean, there's cost benefit in it.
[00:07:06] Speaker A: Scrabble has abilities, right? I mean, that's. That would be an example. It's a word game that you can throw away your tiles, you can get new tiles. It's very possible to build a game economy around word games. That I think that. That becomes. That's the interesting puzzle is how you utilize those mechanics. How do you ration them?
[00:07:21] Speaker B: Yeah, can I, Can I give a. Can I give a defensive wordle, please? I think it's a beautifully simple and elegant game. I agree. It does not have a lot of these progression systems or like as much skill, expression or whatever, but it takes that simple logic puzzle of Mastermind. If you're familiar with that game where you're trying to guess the hidden thing and you get partial information on each guess and you, you know, iteratively improve your guess. But Mastermind is. It just turns into like a sorting algorithm at some point.
Whereas because they're words, you all of a sudden have the texture of the English language. You know that C and H often go together. So like, if, you know, you look for com of words that use specific letters and it makes it. So the puzzle isn't just this, like, you know, all permutations of all possible code words. It's like actually you're using your understanding of language and sounds to add text to the puzzle, which I think is great.
[00:08:11] Speaker C: Yeah, I like that.
[00:08:12] Speaker B: It's not much more than that, but, you know, it does what it does really well.
[00:08:15] Speaker A: The thing I think I would probably change the puzzle experience most for me is if there was more personalization when it came to difficulty. Like one of the things that Match three is really good with is optimizing seeds and having adaptive drop. So one of the things that will happen is if they see that your likelihood of completing the game is low, they will change the weights of the colors. So they might say, oh, you're more likely to get reds versus other colors, which makes there be more likely matches to happen, which makes the board more likely to clear, which makes you more likely to complete your goal. The thing I've struggled with is that the difficulty always feels very flat also in word games. And I'd love to see how they can scale and make it more intuitive and give you those satisfaction moments.
[00:08:58] Speaker B: Yeah, I mean, to your point, to your point, there, there's not numbers to tune.
[00:09:02] Speaker A: Right.
[00:09:02] Speaker B: Like the design space is very limited. You know, you don't have a difficulty slider or you know, stat points you can adjust.
[00:09:07] Speaker C: I mean, you have similarity in indexes you could put together. So if somebody's, you could build a wordle that has a maybe variable length even and it's working from a bank of a hundred thousand words or I don't, I have no IDEA how many five letter words there are. Maybe 50,000 words. And you adapt the word.
You know, this Joe is always guessing. You know, other is the word that, that he always starts with. How can we adjust it so that he gets a better first guess? Because if you have a terrible first guess, you're not going to get a lot of hits and you're less likely to complete the word.
I suppose there's ways to hold the.
[00:09:42] Speaker B: Word after the first guess. Like after someone types in their first guess, you could like pick a word that makes it easier or harder.
[00:09:48] Speaker C: Interesting. So every time you miss it, it's a different word. I like this idea that you hit on Eric where it's like, this is a elegant, short, sweet game. It's not supposed to be some highly complex, you know, thing. I mean, chess is actually really simple when you think about it. It's a small grid. It's much simpler than a game like League of Legends or far fewer.
[00:10:08] Speaker B: There's no progression systems in chess. You can't increase the stats of your night.
[00:10:12] Speaker C: No out of round, no in round progression. It's, you know, pure strategy. But I think it's a perfect game. I think it's the most, probably the most perfect game that was ever that's ever been invented. And it is that elegance. A lot of favorite board games are simple and elegant. They're. They're complex in the ways that are. That are interesting, like with words. For example, the ways that A, C and H can move together. And they're simple in the ways that. That I also appreciate, which is there's not a bazillion pieces of garbage running around which gets. Gets into my game that I've been playing, which is Total War, which is an overwhelming hodgepodge of shapes and colors and dings and booms and it's.
[00:10:50] Speaker B: Which. Which iteration are they?
[00:10:51] Speaker C: Are.
[00:10:51] Speaker B: Are you playing?
[00:10:52] Speaker C: I have no idea. I went to the app store because Phil finally spilled the beans on what the actual game is called that all of those advertisements use where you've got the guy and you're side scrolling from side to side and shooting Last War survival. Oh, is that what it's called? Last. What did I say? Totally.
[00:11:09] Speaker B: I think you're talking about like Total war Warhammer.
[00:11:11] Speaker C: Sorry. No, no, you're. I'm totally wrong. It's last. Last war. That's a generic name that I completely.
[00:11:17] Speaker A: Oh, you're playing fucking X. Oh my God.
[00:11:20] Speaker C: This is. You consider this 4x.
[00:11:22] Speaker A: Yeah, this is. Oh, it's. It's the holy. The genre that's part of. Very, very defined. Yeah, but a fake. That's a fake ad. It eventually becomes a 4x game.
[00:11:29] Speaker C: But I played that for 10 hours and I was shooting.
[00:11:32] Speaker B: Guys, wait. It doesn't. It doesn't transition to like a base management game.
[00:11:36] Speaker C: I don't have my. Well, it is a base management game, but that part sucks. I don't care about that stuff. I just try.
[00:11:40] Speaker B: That's the part that gets people to spend money, baby.
[00:11:42] Speaker C: I click it through it as fast as possible so I can get to shooting bad guys again. And they have this infinite horde that you can basically just. You can go down this track forever and just play that version of the game with no progression. It's freaking amazing. I don't know why they have that in the game if they want to get people to spend money. But to me, that's an example of a game that has too many systems that get in the way of the core gameplay loop. And I'm just confused.
[00:12:03] Speaker B: Well, the. The systems are there to get you to spend money.
[00:12:05] Speaker C: No, I know.
[00:12:06] Speaker B: It's like a trend where, like, it's an arcadey game that tricks you into playing a 4x game.
[00:12:10] Speaker C: Yeah. And. And like I uninstalled it after 10 or 12 hours. Because I was like, okay, this is like. This is getting to the point where. Come in. I've got this base that needs to upgrade in two hours. I don't have two hours. So then I was playing just the.
The infinite kind of pathway version of it, and that was. That was a little too fun. I was just sitting there mindlessly shooting these guys, which is not really a challenge. Wasn't challenging me at all. Very, very low cost, low cognitive burden, Very, very high marginal benefit because I've got these flashes and colors and, you know, the satisfaction of mowing over these enemies. So it's like the opposite of wordle. And I'm still playing wordle every morning. And I'm not playing the last war. This is not for me, Phil.
[00:12:52] Speaker A: I'm sorry. I'm sorry. You abandoned it. You missed out on a great high LTV experience.
[00:12:57] Speaker C: I did miss out on. They missed out. Thank you. They missed out on a high LTV experience. But I have recently gone into watches, which I think is going to fill.
[00:13:05] Speaker B: No. Oh, that's very high.
[00:13:07] Speaker A: The ultimate velvet. Good.
[00:13:09] Speaker C: Yeah. Dude, I gotta. I gotta not. Also, you know, I've been playing Dark Souls 2. Return to a classic. No monetization in that.
Not overwhelming. No overstimulation.
[00:13:18] Speaker B: Was it a remake or a remaster?
[00:13:20] Speaker C: It's the.
Whatever the digital version you get on the Xbox store is. I think it's the Scholar, the first SIN edition. So it comes with the downloadable content, which I'll probably get to after I've completed the game. Phil, what have you been playing?
[00:13:35] Speaker A: So one of the growing parts of gaming in the United States is something called sweeps or sweepstakes. And this exists on the app Store. You can see a game like pulls, but this has been like a large proportion of my consulting business for the last six months. And it's this really weird part of US Legal law where it's essentially gambling. It's not called gambling. And the idea is that you buy virtual currency packs and you get what you might consider normal. We call it. We call it soft currency. And then you get these things called sweeps coins as a part of a bonus for buying the virtual coins. And what you can do is you can essentially play random casino games. And you can either play them with the soft currency, the gold coins, or you can play them with the sweeps coins. Usually there's literally just a button that lets you.
Between the two. It's literally the exact same games. And what you're able to do if you play with the sweeps coins is that you can actually cash them out for real money. And so the reason this has grown in popularity is essentially, it's a legally compliant way to do gambling.
[00:14:40] Speaker B: Wait, how are they getting around the gambling laws?
[00:14:43] Speaker A: Yeah, so let's. Let's talk about that. So there's actually three things you need to be considered gambling in the United States. The first is a prize, chance, and consideration. So the way they're able to get away these things is that, first of all, you are never paying directly for a chance to win. You're only paying for virtual entertainment. And the sweepstakes currency is granted for.
[00:15:06] Speaker C: Free, which is a bonus for you're buying the soft currency and you're getting the sweepstake for free.
[00:15:12] Speaker A: Correct. That is exactly what's happening.
[00:15:14] Speaker B: But then what about the cash out?
[00:15:15] Speaker A: So the cash out can only happen after the sweeps coins have been played through one of the games. So they're considered a virtual prize.
[00:15:25] Speaker B: But the prize is. The prize is just a direct ach transfer. Is it a gift card?
[00:15:29] Speaker A: It can be a direct ach transfer. Wow. But it has to have been circulated through the game. The other thing you need is you need to have an alternative method of entry. So there's no. There's no purchase required. You often see that in sweepstakes, like the monopoly, you know, McDonald's prize. You see that all the time. And so all of these. All of these sweepstakes games provide you the opportunity to mail in an entry form to. To an address. And so there are people who will just do thousands of these alternative methods of entries and will basically just buy bulk stamps and send them out because the benefit of the prize is greater than the cost of the stamp. So they win on expected value basis. That's mostly been bought down to zero. But there's an army of people that this.
[00:16:16] Speaker C: So you can. You can mail in some sort of form in order to earn these in game sweepstake tokens.
[00:16:22] Speaker A: Yep. It's essentially like you're. You have an entry point.
[00:16:25] Speaker B: Okay. So then technically you can get it for free. So technically it's not. Wow.
And is this. This is a recent development. Like, I feel like these loopholes probably existed the whole time.
[00:16:36] Speaker A: It's been more and more popularized in the last year. I'm sure this will be closed within the next 18 months, maybe 24 months. Like, it's clear that it's moving in that direction. There's a bunch of cases in various states which are moving through the court system. Basically trying to say this is a loophole and get this stuff banned. But right now it's been compliance in many states and it's been growing at an incredible rate.
[00:17:02] Speaker C: So what is incredible is it like a lottery? Like everybody pays in and then they take some of that pie and they.
[00:17:08] Speaker A: Just know you're playing slots winners, you're playing slots. I mean, it could be anything. Like if you're sports. So think about all the things that are gambling are associated with. You can put this through a sweeps model, so you can basically run a sports book in which you use sweeps coins. So are the Patriots going to win next Sunday? You could use sweeps coins and do normal betting essentially with that.
[00:17:26] Speaker C: Most of these platforms for betting or are they like, are they slot machines.
[00:17:30] Speaker A: Like you anything you can imagine gambling. So some of them are taking the approach of doing slots machines, other ones are taking the approach of doing sports books. Again, it all comes down to crazy, whatever is associated.
[00:17:43] Speaker C: Isn't it interesting that this is such a heavily regulated thing, but clearly people want to spend so much money on it. You know, it's like, it's like alcohol. It's like, okay, you could ban alcohol, but there's some negative consequences of people consuming alcohol, so we should ban it. But people really, really want to spend money on alcohol. It's the same thing with gambling. Like we try so hard to ban it and it's like, Jesus Christ, if somebody wants to blow a thousand dollars on sports betting, I, I'm insensitive in this way. You're spending your own goddamn money. You're not on heroin, you're spending money on a betting game.
[00:18:18] Speaker B: I think you kind of have to meet someone who can't control themselves and throw away their life savings.
[00:18:23] Speaker A: So this, this is often the defense I give of loot boxes and why loot boxes are not gambling, which is you can't fall into the gambler's fallacy. And I don't think that's a minor point. But you can't chase your losses in a loot box system because this, the money is always trapped in the system. And this was EA's defense. This has always been the defensive loot boxes from a legal perspective. Not only do I think it's a legal perspective, I think it's a functional behavioral difference. If you can't chase losses, that is where you end up in the deep is, oh man, one more game, I can get back all I lost. That's never the case.
[00:18:50] Speaker B: So then do you think tradable ones like counter strike loot boxes are, are gambling?
[00:18:54] Speaker A: Well, hold on. That's still trapped in the csgo. System still trapped in Steam credits. You can't back that out into real USD.
[00:19:00] Speaker B: You can use it to buy games and other. True, you can launder it, you can extract it from the system by trading it to someone who gives you money.
[00:19:08] Speaker A: Yeah, but that's a violation of the terms of service which Valve has often pointed out.
I mean, but you still. I think it's depending on how aggressively they try to enforce it. But it still is trapped within the system.
[00:19:19] Speaker B: They don't try to enforce that at all. Are you kidding me?
[00:19:22] Speaker A: It's still not technically sanctioned on the platform and they don't provide an easy way to do it.
It's still, I mean look it, we can, we can debate about their level of enforcement but at least that is technically something that's not possible.
[00:19:33] Speaker B: Okay, so, so you're going to say that those are not gambling even though.
[00:19:36] Speaker A: They are transferable and it's still within the circuit of Steam. And even if you got a reward inside of a loot box it is still something that you could redeem within the game. So it's just a question of how we're defining what system. To me it's, it's, it's obviously not USD, you know and if you're doing something outside of that, that is a transfer that's against the terms of service. That's something they don't sanction. There is no back out into PayPal gift cards. If there were a back out into PayPal gift cards that would be a problem. Now the interesting piece is actually which it violates all of this and I think actually is gambling and you can chase losses.
[00:20:09] Speaker B: Well what if they just put something in the terms of service that says you're not going to trade it for usdc?
[00:20:13] Speaker A: I think, I think that would be, that would be a great attempt. I think, I think the problem with.
[00:20:17] Speaker B: The blockchain though, let me be clear, with zero intent to enforce it. Just like they just write it down.
[00:20:22] Speaker A: In there would be interesting. I think the problem is that technically it depends. So if we think about Draper Labs, think about NBA Top Shot. The question is you could sell those things for usdc. You could always take it off the blockchain. Like literally you could not keep it on the Draper. I mean I guess it was a private chain.
[00:20:39] Speaker C: Dapper Labs.
[00:20:40] Speaker A: Dapper Labs. So I think the question is whether or not technically they create their chain in such a way that you cannot do that. I don't think just putting in the terms of service would be enough if that technically you have that possibility. To be able to take it away. Like you can't do it.
[00:20:53] Speaker C: But it's only gambling if, if there's randomness, right. It's like, I mean, I'm trying to think of games that, yeah, you're opening.
[00:21:01] Speaker A: You'Re opening packs, man, for NBA Top.
[00:21:03] Speaker C: Shot in that specific instance. But there's a lot of games that aren't like that. But I think a lot of web3 people think that they're getting around all the gambling stuff because I'm seeing a lot more gambling platforms. It's the fastest growing app in the Web3 ecosystem.
[00:21:17] Speaker A: Definitely illegal. I mean, don't be right. Definitely legal.
[00:21:21] Speaker B: The only I think normally use case for crypto I've seen is people trying to play online poker. That's the only time any of my friends has ever asked me about crypto.
[00:21:29] Speaker A: So anyways, putting, putting that stuff aside, I've been working in sweeps and I would say the other thing, it's gotten me into like a lot more gambling economics and I do find it fascinating.
[00:21:38] Speaker C: What's your day to day look like when you're working on one of those projects?
[00:21:41] Speaker A: Like to me it's just like, just gamble all day. Gamble online?
[00:21:44] Speaker C: No.
[00:21:44] Speaker B: Like, are you a b. Testing payout tables for like optimal engagement or something or.
[00:21:48] Speaker A: Well, so I think on one hand setting up the model is really interesting. Like coming up with the design mechanics is very interesting. So to give you an idea of something, that social casino has come up with Jackpot Party, which is a social casino game. So it's not under the sweeps model, but they have something called a parachute. And what happens is when your balance hits zero, they immediately refill it with 10 million coins and it's on cooldown for five minutes. Yeah, it's, it's a fucking brilliant game design mechanic. Right. So there's a lot of ways we can fuse game design and monetization design into these, these games. And the other part is that we also need to run it through a model. So for instance, with sweeps, they also try to cap the amount of payouts that they're giving to users at any one point because they don't want to end up in a liquidity crisis. And sometimes what I've talked to these companies about is buying insurance. There could be a situation in which you get hit with a huge payout and you need to have insurance. Or what you can do is you can limit the amount that you allow globally or individually from these sweeps coins.
[00:22:42] Speaker C: So you're just a monetarist. You've. You've you're basically doing crypto, but in Web two, you're balancing their books to make sure they don't lend out.
[00:22:52] Speaker A: It's exactly that. But it's like, okay, could we put game design mechanics around it? Could we change your payout maximum based on your VIP status? Okay, how can we decide mechanics around that? So there's the core. There's a core slots, which is really interesting. I've started to do a lot more simulations around that. The interesting thing that you end up popping out of that model is that ultimately what drives a lot of casino and slots is velocity. So every time you take a bet because you have negative expected value on the bet, it's okay, first of all, what is that expected value decrease? And by the way, Most of the RTP in Social Casino, even sweeps, is 90% plus. The key question is, like, how many times can I get you to do the spin? Because you take a hundred dollars and then you spin it down. Okay, I'm down on 92. And you spin it again, you spin again, you spin it again. So how many times can I get you to spin it down before you cash out? And most people don't cash out. That's the thing I think people need to really understand that's why social casino works without even having the sweeps component, is that it's entertainment. It's the joy of the poll that matters.
[00:23:48] Speaker C: I'm curious. I'm assuming 50% of your distribution are. Well, actually, probably more like 65% of your distribution are losers, and 35% of the distribution are winners. What does the retention rate look like for the winners versus the losers? I'd imagine it's like, dramatically higher for the winners compared to the losers.
Crazy to me.
[00:24:07] Speaker A: No, because, remember, in the long run. But see, it's. It's not here. It's already the wrong framework. It's not about winning or losing. It's about the experience. Just like every other form of entertainment, it's the. It's the cost per hour. And this is just the type of entertainment that people are finding compelling.
[00:24:21] Speaker C: So I actually did a retention analysis on our players based on whether or not they were positive ROI on the day in our game. So basically, given what you spent in crafting fees in the material that you burnt to craft, items in the fuel that you use to move your ship around, all that stuff, all the costs. Compare that to the amount of atlas that you made, whether you were positive or negative. And I found that the negative people actually had a higher probability of returning tomorrow, which Is I guess, kind of leans into this chasing, chasing mechanic where you're trying to chase a win coming off of a loss.
[00:25:00] Speaker A: The thing that's become really interesting is we talk about the velocity and so you start to think about like other randomization mechanics. One of them is also popular scratch cards. If you remember scratch card you can get at like any gas station. But you also remember the velocity of each scratch card is so much lower. Okay, are there UX things I could get you to do to. Can I auto reveal the scratch? How long does it take you to do like, how can I increase velocity? Becomes this really interesting question.
[00:25:24] Speaker C: So that's what you've been playing?
[00:25:25] Speaker A: Yeah, well, it's. I will say I love this job because I've been doing some of that. And then I'm also working with some children's education apps at the same time.
[00:25:33] Speaker B: There was just.
[00:25:34] Speaker A: There was, it was actually earlier today. I walked from like a. I'm working on an MMO for 5 to 9 year olds and then I walked into like a street gambling.
[00:25:42] Speaker C: Dude, that's sick. That's why you gotta keep your, your consulting on the DL so that the children's apps aren't like, this guy goes gambling, he's gonna put gambling.
[00:25:51] Speaker A: I think it'd be the inverse. I think they'd be like, oh, this guy does gambling. And the gambling people be like, oh shit. This guy does children's apps and it's a positive.
[00:26:00] Speaker C: Yeah, why not for the gamblers maybe. I don't think the children's apps see a gambling app positive.
[00:26:06] Speaker A: Well, I. There's one company I worked with and the thing with gambling is that it's so inaccessible. If you go to some of these websites and you are not a gambler, it looks like a fucking Excel spreadsheet. And what's interesting is no one has really casualized gambling.
And I think that remains to be seen.
[00:26:21] Speaker B: What do you mean? What is a slot machine?
[00:26:23] Speaker A: I think it's exactly that. I mean, this is why I think Monopoly go was so successful and people have missed that lesson is that I think it casualized social casino and it made that understandable in a way that hasn't been made understandable beforehand. One knows how to read odds by default. It's just like a fucking series of numbers. There's no fatui trip. There's nothing intuitive about it. Like go to all these prediction markets. One of the things about prediction markets is that they translate all of the odds into probabilities, which makes it very understandable. And that's Never the case in any of these betting websites.
[00:26:50] Speaker B: Yeah, I guess I, I have always thought odds were a weird way to display the data.
All right, Speedrun.
[00:27:03] Speaker C: Cool.
[00:27:03] Speaker B: So you guys, how do you guys feel about surveys?
[00:27:05] Speaker C: I'm a five out of five on surveys.
[00:27:09] Speaker B: And tell me why, tell me why.
[00:27:10] Speaker C: Strongly agree. I have no reason.
[00:27:12] Speaker A: Lizardsman's constant, which is what Scott Alexander once wrote about, which is survey, where they, you know, offer the option to see whether or not you are a lizard. And like 3% of people said yes to it. And so he's just like, you can instantly throw out the 3%. Which I actually thought was an interesting lesson about surveys is like, put in questions to see whether or not people are paying attention. And for the people who opted into it, immediately remove them from the survey data. Which I've almost seen no one in UXR do in games.
[00:27:37] Speaker B: Oh, yeah, no, no, that I use, I use all the time. You put like red herring answers. But yeah, no, that's exactly the problem, right, Is that like when you ask someone a question, I don't know, how many, how many calories did you eat yesterday?
[00:27:48] Speaker A: Right.
[00:27:49] Speaker B: You don't actually get that information. What you get is the user read the text, thought about something in their brain, and then typed something in. And you get the output of this behavioral process. You're not actually collecting the information that the question says at face value in that loop from, you know, you have to, you have to perceive the question, you have to cognitively think about what you're going to respond. And then you actually have to execute a response. All sorts of other shit go on. You know, the lizard man example, you know, they're just, their brain is lizard. That's funny. I'm just going to say something that's false, right? And this is just fundamental to surveys. And all surveys have this problem and, you know, with and every step of it, there's the perception bias. Did they understand the question? When they cognitively thinking about it, how hard are they really trying to answer it? You know, there's all sorts of cognitive biases around halo effects and shit, people trying to present themselves as better than they are, that kind of stuff. And then there's also the effort of trying to respond. At some point people get fatigued in a survey and are just like it. I don't, I don't care anymore. Five out of five. You know, obviously if you just straight line a survey, we just start clicking 5 at some point. But I think the way to use surveys is actually to embrace this bias they have these problems and you just, rather than being like, they have these problems and so then they suck. You just gotta work with them and work around them. So for example, if you ask people like, hey, how do you like this game? On a scale of 1 to 5, that doesn't mean anything, like what, what is, what is it? What does a 4 out of 5 even mean? But what you can do is you can ask that question because there's all sorts of biases, like, what does the scale actually mean? How is the player feeling when they.
[00:29:12] Speaker A: Respond to all that?
[00:29:13] Speaker B: But if you just ask the question over time and track how it changes over time, if you assume all the biases are static, then then you're actually observing like a diff and diff. Basically you're, you're saying there was actually some system change that caused the level to change. So that's a simple example of basically working around the biases. Like, yes, there's biases in the data. We're going to measure something else on top of that.
[00:29:34] Speaker C: Con papers that take that approach, they like, basically create a diff and diff and they categorize people who, for example, let's say you wanted to, like, are you good at, at, you know, X, Y, Z activity? And you have two different categories. You have the people who say that they are good at it, and you have the people who say that they aren't good at it. Then you challenge them at that activity and you see if there's a difference between the people who said that they were good at it and suck and the people who said they weren't good at it and suck. So it's kind of an interesting framework.
[00:30:01] Speaker A: So I, I immediately agree with your, I, I want to agree with your proposition, Eric. So that immediately throws out all these happiness surveys we have to read every single year about how happy Finland is, the country that has the highest suicide rate as well, one of the countries with the highest suicide rate. Like, you cannot compare happiness between countries because the Israelis are culturally frowned upon for expressing happiness, to say, especially with the Russians too.
[00:30:23] Speaker B: But you can't compare happiness within a country over time.
[00:30:26] Speaker A: I don't think you can either. If we were to take games as one example, the population that plays a given game at, you know, survey time t changes as well. Like, let's say you do a UA push and now a bunch of gen zers come in. You're still not measuring within a given sample, within a given person.
[00:30:43] Speaker B: So, okay, so if the gen, if the younger generation is using the scale differently, Their definition of a 3 out of 5 happiness is different than the millennial's definition of 3 out of 5 happiness. Then yes, the scale use changes and then it's invalid. But if you, but if they're using the scale the same way, you know, assuming the language and perception hasn't changed that much, then I think you, you can make the claim that younger Finns are more depressed than older Finns.
[00:31:06] Speaker C: There's also, there's also external factors that can manipulate, I mean, I think about consumer sentiment indexes, indices, it feels like those are becoming less valuable. If you look at the U. Michigan Consumerist Sentiment Index, like following the great financial crisis, I'm pretty sure that it's rating, it's, its level is higher in the two to three years following the financial crisis than it is now. And we're not in like the dip, the depths of a recession. So there are like these macro problems. And then just anecdotally from, from having worked with a lot of survey data past, there were a lot of. It's very tempting to use a survey question as a, as an, as a treatment, which obviously you're pointing out you should not use it as a treatment. It's not experimental. But typically what I found when I was studying, for example, racial disparities in wage is that people who said you were going to be a higher earning earner, that typically there was no causal relationship between their responses when they were, let's say 15 and their actual career outcomes when they were 25. So there's also empirically, surveys are just very weak now. Does that mean that they're invalid?
[00:32:22] Speaker B: I think that gets to the point of like, people are better at answering certain questions than other questions.
[00:32:28] Speaker A: Right.
[00:32:28] Speaker B: So I'm gonna run you guys through some questions and I want you to just respond as soon as a question comes to your mind.
[00:32:34] Speaker A: Oh, boy. Oh, this is happening live. Okay. Okay.
[00:32:38] Speaker B: No, no, they're easy. Okay, guys, ready?
[00:32:39] Speaker A: Do I have ink blots?
[00:32:40] Speaker B: Yep.
[00:32:41] Speaker A: Okay. Yeah, I'm ready.
[00:32:42] Speaker B: What did you have for lunch yesterday?
[00:32:44] Speaker A: Guacamole, beef wrap, shit, chili.
[00:32:47] Speaker B: What do you think you'll have for lunch tomorrow?
[00:32:49] Speaker A: I think I'm probably going to have a halloumi pie because I live in Cyprus now and that's the only thing this country produces.
[00:32:55] Speaker C: Yogurt bowl.
[00:32:56] Speaker B: Would you rather have pizza or sushi for lunch tomorrow?
[00:32:59] Speaker A: Sushi in this country.
[00:33:00] Speaker B: Do you want to eat pizza for lunch tomorrow?
[00:33:03] Speaker A: No.
[00:33:04] Speaker C: Do I want to eat sushi tomorrow?
[00:33:06] Speaker B: Do you want to eat pizza for lunch tomorrow?
[00:33:08] Speaker C: Pizza for lunch tomorrow?
Not particularly.
[00:33:12] Speaker B: Why?
[00:33:12] Speaker C: Because I only can have pizza every once in a while or I'll get.
[00:33:15] Speaker A: Fat I don't like. I think I know what Eric's doing and I like it. I like what he's doing. I love it because it's very similar to how we do.
[00:33:21] Speaker C: Let him do it.
[00:33:24] Speaker A: So why don't I want pizza tomorrow? Because the pizza in this country sucks. I don't want to call it pizza. So she's better.
[00:33:29] Speaker B: All of those questions are about food preferences, right?
And all of them tell you some information about people's food preferences, but they're actually very different cognitive tasks. So for example, the first one, what did you eat for lunch yesterday? That's recall. You just think back in your memory and then recall effect. That's very easy to do compared to the second one, what do you think you will eat for lunch tomorrow? You have to do anticipation. You're like, thinking ahead. You're like, you know, trying to make a bunch of considerations there. That's a lot. That's significantly harder than recall. And then for the others I mentioned, like the comparison, it's a lot easier to compare sushi and pizza than it is to evaluate pizza in a vacuum. Like, do you want to eat pizza versus all other options or do you want to eat pizza or sushi? Right. Like, the binary comparison's much easier. And then finally, the last question on rationalization was like, much, much more involved. Right.
And I think if we went back to the recording, you would see the speed of your response reflects how difficult of a cognitive task each one of these was.
[00:34:23] Speaker C: So is the thesis here good survey design?
[00:34:26] Speaker B: Yeah, exactly, Survey design. You should think of it like UX design, where you want to minimize the cognitive load of the user. Right. All of these are different ways to extract food preferences. Figure out which one is the lightest, simplest one for the user to respond to. And yeah, that's the one you should use. And just like other good UX design, right? It should be seamless, smooth, simple. Like, not if, if people are getting tired or confused partway through your survey. That's really bad design.
[00:34:51] Speaker A: I think there's an even stronger econ lens here, which is like you. I mean, you know, you're. I guess it's not. That doesn't, that doesn't help us. You know your preferences best. But there's, there's a lot of evidence on structured interview. And this is something Amazon did really well. You know, if you've ever interviewed on Amazon, I worked at Amazon Game Studios beforehand. They generally always Position their questions as tell me a time you did. Tell me a time you did. And that forces you into first of all evidence based reasoning, right? So you can't, you can't imagine these scenarios. Like it removes a lot of bias because you can't project all of these things that we usually criticize surveys for. It's, it's about memory recall, recollection. Exactly the things you pointed out, Eric. And I think that makes a ton of sense in an interview. Tell me about a time when. Tell me about a time when. And I think the other thing that's really interesting is when you ask, hey, if I were to do a reference call with your, with the people on your reference list, what would they tell me about you? Like what are other ways you can basically get people to confess the truth? And I would love to see UXR surveys do more of that. You know, even just the threat of like checking your data and obviously GDPR makes this a little bit more complicated. But like some ways that you can kind of like get people to, to conf. I think is interesting. The lizard's been constant. We talked about like I actually think incentives might be destructive here, right? Because people just want to finish the survey as quick as possible. If you offer them reward and you get a bunch of sampling bias, I.
[00:36:16] Speaker B: Don'T know, I would say the longer the survey is, the worse that is. If it's just a single question incentivized survey, it's not as bad, but you still get a bunch of opt in bias. But that actually dovetails into my next point which is that like surveys are kind of like ads for mobile games where you are putting this thing that disrupts the game for the user in order to extract a little bit of value, right? Like you're 1/10 of a cent from your ad view or your, you know, like your one data point from the survey. And so like when you use surveys in games, you have to think of them like ads. You have to think of them like this is disruptive. I can only put it in certain parts. Maybe I have to incentivize people to participate. Yeah, they're not free for sure. And I think a lot of companies, once they get a survey team up and running, they're just like, yeah, let's just blast our users with surveys.
[00:36:58] Speaker C: How do you effectively, you know, I'm thinking about like when you run an experiment and you get, get cohort contamination, you get the treatment group and the control group are contaminate or mingling. I think about this with surveys Too. It's like the first survey that you send out that might have a huge impact. Everybody might take it. It's like, oh cool, interesting. Like League of legends wants my opinion, I'm going to fill it out. But the hundredth survey or the 20th time somebody you know, you start to get terrible feedback. I don't know, you seem like you have a little bit of experience with this stuff. Like how do you think about that? What's the optimal cadence for a survey?
[00:37:33] Speaker B: Yeah, no, that's absolutely effect. And the thing you mentioned about the U. Michigan survey responses probably is a result of this. There's our modern world's over surveyed. You get way more surveys than we ever have and people will just stop responding to them. You just can't do it too much. You can't make the rate too much. There's a lot of survey tools. You can lock out people who have had a survey in the last two weeks or last month. So you don't sample them, but you just got to make sure that you're continuing to sample from a similarly represented of pool.
[00:37:58] Speaker A: Right.
[00:37:59] Speaker B: Yeah, but yeah, it is, it is a problem. And I think email and phone surveys are dead. That shit is garbage. Now the only people responding to those are kooks. Whereas like in an in game pop up, that's a single yes or no question. People will still respond to it.
[00:38:11] Speaker C: Yeah, I like this idea of like the red herring question or the, you know, just kind of the testing question.
[00:38:17] Speaker A: That's just the confederate as they call it in behavioral. Behavioral economics, the plant. Do you want to continue? Do you want to do last topic, Eric?
[00:38:24] Speaker B: Yeah, yeah, let's close it out. Okay, I do. I wanted to add one thing to this, which is you touched on earlier, which is like people are good at understanding their own preferences. So it's a lot easier to ask people, do you prefer X or Y? Right.
People are really good at telling you how they feel. Like I feel happy, I feel angry, I feel sad. They are much worse at telling you why they feel that way.
[00:38:45] Speaker A: Right.
[00:38:45] Speaker B: And so when you ask people like, oh, what should we change about the game? Their like explanation for why they're frustrated is usually really bad. But if they feel frustrated, you can trust that they say I'm mad. They're probably mad. If they say I'm mad because your balance is and because you know, you're like, you know, this character is too strong, blah, blah. Maybe that's not actually why they're mad.
[00:39:03] Speaker A: Yeah.
[00:39:03] Speaker B: And so, so I think good survey takes into account like make questions very simple and like, know what information people give is reliable and what's not.
[00:39:14] Speaker A: That makes sense to me.
[00:39:22] Speaker C: Yeah.
[00:39:22] Speaker B: You want to talk about matchmaking?
[00:39:24] Speaker A: Yeah, let's talk about matchmaking, if that's fine.
[00:39:26] Speaker B: Yeah, yeah. I, I thought this was a great article you wrote.
[00:39:29] Speaker A: Yes, I.
You remember a while back we covered Call of Duty's white paper which they released, showing the uplift in skill based matchmaking when it was implemented versus a control group. And for those who are unfamiliar, all skill based matchmaking does is include skill as a factor when pairing you with other players. So the idea is that if I pair you with other like skilled player, you might have other. You might have a higher retention rate than you might otherwise have. And that was exactly what that paper showed. And not only that, they showed it was a very dramatic effect. It was a huge, huge increase in retention. And thank Activision for releasing papers like this. We usually don't get data like this. The other thing that paper showed is that you can actually do GDPR requests and you can get a lot of your own personal data that Activision has about you, which is fascinating. And I can't believe more gamers don't do that and actually think I'm just going to start to do that to see what shit I can get on other companies. But the piece I wanted to write about was based on something that happened when I played Command and Conquer mobile, which by the way is one of the greatest mobile games of all time that was never successful because EA doesn't know fucking jack shit about mobile gaming. But one of the things they did that's always stuck with me is they offer you challenge matches so you'd be matchmade against an opponent that was wildly better than you. And what would happen is that you couldn't go down a rank and if you won, you got 2x the amount of points that you would normally get. Get. And so if you're watching the YouTube version right now, I'm actually sharing an image of what this looked like. You can't lose medals in the match and if you won you would get, you know, higher, higher experience points.
[00:40:58] Speaker B: And so the higher skill players know that it's a challenge match.
[00:41:01] Speaker A: No, they do not know that they're playing down against someone. So the, the idea that I've basically had is that skill based matchmaking, we spend so much time trying to get 50% win rates, which is not fun necessarily. Like ultimately we're trying to optimize for fun, not for fairness. And you think about things like ELO that was not designed for fun. It was designed for trying to make competitive matches. And then I thought about golf, which has handicaps so people of different skill levels can play against one another. I find it kind of amazing that we haven't applied economy design to matchmaking. So why isn't it the case that I can actually have purposeful matches in which I'm the underdog and then I might be able to come out ahead even if there's a slim probability? That's a really exciting possibility.
Or what if I just play a match and because the servers were not as full and I had to play someone who was higher or lower skill than I, then we do compensating differentials like we show, we know this in economics all the time, is that we can pay you to take on additional risk. We can pay you risk premiums. These are all things that seem like really compelling levers on economy and progression design that feel like they haven't been maximized. It feels like command and conquer rivals was just the surface. And I'd love to see more economy designers go after stuff like this.
[00:42:13] Speaker B: I'm totally with your thesis, which is that matchmaking is there to make the game fun, not just to make it fair. And there's a whole suite of things you could be doing around the matchmaking experience, like presenting this is a challenging match, right? It primes the player. If you lose, don't feel bad. If you win, you should feel great. And then the beautiful thing is the, the high skill player in your command and conquer example, they think it's just a regular match. So they think they just played a 5050 opponent and beat them, right? So it's, it's almost win, win. And I think we do see this to some extent, like with the skill differential, you know, people know that if you play against a higher skill opponent, if you win, you move up more than you move down if you lose. So that is happening somewhat. But I do think that's just an incidental effect of the statistical algorithm of elo. And maybe we should be intentionally creating these experiences. There's often a complaint that these live service competitive games are samey. You just hit play, play an opponent, hit play, play another opponent. And creating these moments where it's a particularly challenging opponent periodically can do a lot of things well.
[00:43:09] Speaker A: I think one of the biggest criticisms of skill based matchmaking, which is again, not just say that Reddit bitches about, because the other interesting part about skill based matchmaking and the pushback against it, particularly in Black Ops 7, which by the way they seeded to get two different ranked playlists. Which again ends up with a self selection problem that if you are a bad player, you want to be in the skill based matchmaking playlist. Because if you go into the other one you're going to have this crowding out bias where all of the skilled players have gone to the one without skill based matchmaking because they prey on the weak. That's the problem here is that there are winners and losers in skill based matchmaking.
Which is one reason that explains a lot of the Internet controversy. But the other piece of this is that skill based matchmaking is always pulling you down to a 50% win rate so it never feels like you have progression. And I think that's a huge problem with skill based matchmaking.
[00:43:56] Speaker B: Well, most games design around this by having the ranked tiers, you know, you went from silver to gold to diamond or whatever. That's supposed to be your reference point for I'm improving in skill even if my matches are still 50, 50 and hopefully the game is rich enough for like you can tell, oh my opponent's using more powerful techniques and strategies.
[00:44:12] Speaker A: And I think this is a really great crystallization of the difference between systems and economy design. It feels like the systems designers have won here, but the economy designers need to step up.
[00:44:21] Speaker B: Well, what's, what's your distinction there?
[00:44:23] Speaker A: Well, exactly this. Adding those progression tiers like gold, Platinum, you play overwatch, any of these, you know, most, most skill based systems, skill based, you know, competitive first person shooters will give you this symbol that makes it feel like you're progressing and it does feel like you're progressing. It's, it's what you show to other people. Oh my God, I'm in gold, I'm diamond. But again there isn't a compensating differentials around like XP or, or, or additional currency. You don't see that as much. It's mostly kind of ux.
[00:44:48] Speaker B: Yeah, there is something clever you mentioned Apex Legends let you like wager your rank points.
[00:44:54] Speaker A: My God, I thought this was brilliant. I mean so you couldn't, you couldn't, you didn't have autonomy over wagering your rank points. But what they did let you do, or at least what was one of the components of that game, is that because you have Battle Royale, which is 99 players, you would have mixed games where you would have some players who are very skilled and some players who are not very skilled. And if you were in a high level of ranked progression, if you were in a high Tier then the bar for you to get more ranked points to get to the next rank would be higher even though you were in the same match. So you might need to come in the top five if you were to be able to continue to progress. Whereas other players might have only needed to come, let's say in the top 10 or the top 20 or the top five 50, which I thought was another brilliant way to handle this.
[00:45:38] Speaker B: So this is like the common setting differentials you're talking about, right?
[00:45:41] Speaker A: This is, this is a great example of it. I do wish. That's very interesting, Eric. Like I do wish you could wager more. I wish there was more autonomy.
[00:45:47] Speaker B: Like I want this to be a high stakes match versus I want this to be a chill match.
[00:45:51] Speaker A: Yes.
[00:45:51] Speaker C: Yeah.
[00:45:52] Speaker B: This is something we deal with a lot of games where people will. Some people want to play sweaty and some people want to just kind of mess around and like being. Most games separate the cube cues, you know, casual matchmaking and competitive matchmaking. But it's the same game like everything once you're in the lobby is the same and this creates a Q split issue and stuff. And like if there was a way to merge this by somehow creating the some systems around the matchmaking wagering.
[00:46:16] Speaker A: Right.
[00:46:16] Speaker B: To affect whether players take it seriously or not. I, I will say I'm gonna give the defense of skill based matchmaking the 5050 version because. Or at least I'll. I'll present the theoretical model behind it and then can point out if the assumptions are true or not. So one of the core assumptions there is that essentially loss aversion. If you imagine skill based matchmaking world, everyone's 50, 50. No skill based matchmaking. Let's say, you know, you win 60% of your games and I win 40% of my games. Now the assumption of skill based matchmaking is that the 5050 world is better that we'll have the. Your increased happiness from winning 60% is not as big as my decrease least you know, from winning 40% and that if you just balance the two.
[00:46:58] Speaker A: Right.
[00:46:59] Speaker C: That you know.
[00:47:00] Speaker B: Yeah. As in other words, the shape utility curve is concave down. So like you want to take the average instead of getting extremes. Yeah. So that's pretty much the assumption there. Is that true? Maybe, maybe. You know, to your point about one of the most common reasons cited for wanting no skill based matchmaking is you want to feel like you're getting better. When you start, you expect to lose 60, 70% of the time and then you can feel your win rate improving and Skill based matchmaking kind of removes that sense of reference point.
[00:47:28] Speaker A: So, so I think maybe, maybe we should just before the Internet commenters come, come at us. I mean one of the big innovations in mobile gaming and even some HD games over the last like 10 years is having bots. I mean this is president company excluded. This has been a huge boost for, at least we can say Clash Royale has implemented bots very frequently. We try to do this at dice. I mean, I mean the other thing in like Battlefields, people got to remember is that when you have like even 64, 128 player matches, you actually start talking about the Genie coefficient of kills, which is something that I pulled a very long time ago. And there's very few players who end up soaking up quite a bit of kills. Because you can't do skill based matchmaking when you have that level of server population. It's just, it is not feasible. And so that, that becomes another dimension to this. The other piece that I thought was very interesting you mentioned is you can have people with, you can still, even though it's a zero sum game, like one person wins and the other person loses, you can still have people constantly coming into the funnel, losing a bunch of matches and then churning. You can basically have wind farms, right, which can keep your population stable. You can basically just feed them, feed them new way. That's true, that's true. But I guess at least on paper, like they're, they're, you know, it can work out that way because people can leave the population and they can take their losing records with them. So you can end up with, with a very interesting sample over time.
[00:48:45] Speaker B: The use of bots I think is a really big one, especially in Battle Royales they've like massively popularized.
And there's this tricky thing where like these bots are always masquerading as players, which you know, I know you've got moral objections to, but the point of that is to like make the player feel like they beat an opponent of equal skill. So they won their 50, 50 skill check. But really it's a bot that was intended to lose. And so these bots have to do this tricky thing where they want to appear human, but they also still want to lose.
And it's, it's kind of like this bot Turing test that.
[00:49:16] Speaker A: See this is the thing that to me is a violation of content ethics, which is that when you have a lot of these bot based games, they actually, they're trying to fake there's an intent to deceive. Well, that's what bothers me.
[00:49:27] Speaker B: I mean this is the guy working for the gambling company.
[00:49:29] Speaker A: Dude, I, I don't think there's an intent to deceive. No one's under any illusion about what this is.
[00:49:34] Speaker B: I mean no, no one's. It's very obvious. Your first match in Marvel Snap is against a bot because it, it's like.
[00:49:41] Speaker A: Why not leave it all a bot then do you think? Let's put this way. Do you imagine two worlds, One in which the UX is clear that it's a bot and one in which it's not. What do you think the difference in retention would be between those two things?
[00:49:50] Speaker B: I think it would be negative creative in the labeled world. But if you took your slot machines and labeled their explicit payout expected averages versus like I disagree with you.
[00:49:59] Speaker A: I think it would be almost undetectable. And not only that, remember there was, there was rules that were implemented on the app store that forced loot box disclosure odds and we saw almost no difference in user behavior.
[00:50:10] Speaker B: Well, it's because they disclose it on a website. Yep. Or like it's like buried in the menus.
[00:50:14] Speaker A: I agree. That's fair. I think that's a fair criticism. My point, but I would still hold by assumption that even if you were more explicit about those odds, which by the way now you can click into to many of them, if you play a lot of gotcha based game, you click a question mark, it will tell you the drop rates at least by rarity and sometimes by specific character. No change in behavior I would argue on an A B test versus not having those. And again if you look at some of the more modern like Chinese loot box designs like Delta Force is using probability. They're, they're sampling without replacement. So essentially you have a pool of items and then you're pulling one item from the list and the lods are being recalculated based on the relative weights. They're telling you explicitly what the drop rates are. When it's doing the recalculation, it's a very explicit percentage.
And that doesn't seem to have been a problem at all. Like people know it's a 0.001% drop rate and of course that increases in this model. Again I, I think people, this is another case where people, I would say.
[00:51:03] Speaker B: That for the bot example, right? People know there's bots in there, right? When you click play, you know it's a blended between. There's a chance you hit a human, there's a chance you hit a Bot.
[00:51:11] Speaker A: But you don't know. It's the market for lemons that you're introducing and there's still the intent to deceive. You don't know who's a bot and who's not.
[00:51:17] Speaker B: It's part of the illusion, right, Is that, oh, I beat up the, the, the enemy. You know, it's like when, when single player games throw you against a, like a boss monster in the tutorial, but like the boss monster always loses to you.
[00:51:29] Speaker A: I think here, here's where I concede to you is if people, if we can imagine people behind the veil, if we were to use the Rawlsian veil of ignorance, if you could say, like, look it, they're going to be bots in this game, we're not going to tell you which ones are the bots, which one of the bots and which ones aren't. It's almost like sometimes people don't like taking pills and so they, or like when people get executed, you don't know which. Who's the guy with the actual bullet.
Like, I think some people will, sometimes people will opt into those models. I don't think that's the case here, but if you could show that, then.
[00:51:57] Speaker B: Yeah, I think it's a, it's a similar claim to the matchmaking thing where like, okay, so imagine you get X utility from playing against a bot and Y utility for playing as a human, right?
[00:52:08] Speaker A: Yep.
[00:52:09] Speaker B: And in a deterministic world, let's say it's half and a half.
[00:52:12] Speaker A: Right.
[00:52:12] Speaker B: But how much utility do you get playing against an opponent that you think has a half chance of being human, half chance of being bot? If that utility curve again is concave down, then you're better off blending. If the utility curve is concave up, then you're better off differentiating. Right. So it's, it's again, it's kind of like the assumption made by Snap and Fortnite and everything is that people are happier with a blend, an ambiguous blend, than with, you know, two distinct offerings.
[00:52:36] Speaker A: I think that's interesting. What do you think people would say on a survey?
[00:52:39] Speaker B: I think they would say they want. They would all ask for more information. But again, this goes back to the survey thing. Like people don't know what they want. They know how they feel. They don't know what they want.
[00:52:47] Speaker A: I agree. And people often respond in contradictory ways that violate elementary forms of logic. I mean, we could go on and on about how bad people are at their own reasoning.
[00:52:55] Speaker B: And you know, empirically people play more so, you know, I don't know if you want to go with the, through.
[00:53:00] Speaker A: The roof when you have bots. There's no doubt that boosting win rates.
[00:53:03] Speaker B: With bots, specifically like blended bots versus explicit bots. I'm pretty sure if you ran that a B test, you would get lower. Engage.
[00:53:09] Speaker A: Yeah, I agree. So we're in consensus on this though.
[00:53:11] Speaker B: So then if people, if the consumers are happier with it.
[00:53:14] Speaker A: Right. And they're happier with it, they're happier under the lie. The question is whether or not they know the lie, whether or not they, they, they, they willingly accept the lie. Lie or whether or not they, if, if the truth had been more explicit to them, they would reject it.
[00:53:29] Speaker B: I think, yeah, that's, that's my perspective is that it's like going to a magic show where you are opting into being lied to and like that's part of the experience.
[00:53:39] Speaker A: Interesting. I think that's a, that's a very good defense as well because games lie.
[00:53:42] Speaker B: To you all the time. Like they tell you you're amazing and great at stuff. They tell you that no matter what, all your investments make money. You know.
[00:53:49] Speaker A: But I would say I, I don't see the problem. The problem with the bot situation, again is that there's an intent to deceive.
[00:53:54] Speaker B: Yeah. Just like the magician has an intent to deceive you, but it's like a fun illusion.
[00:53:58] Speaker A: It's more fraud though. It's misrepresentation. I think this is an interesting defense, Eric. I don't know if I'm willing to concede yet, but I think this is probably the best defense I've heard of this, of this rejecting of Kantian ethics. Because to me the Kantian rejection of this is if you were to universalize the bot situation, it would harm all games because then you never know. You never know who's a bot, who's not. Which reduces, which reduces like pure, pure plays. Pure plays for, for PvP.
[00:54:23] Speaker B: Yeah.
[00:54:24] Speaker A: Like where people don't want to, do, want to reject the bots now, now they never know. They don't know who's being truthful and who's not.
[00:54:29] Speaker B: Do you, do you think Kant would dislike magicians? Do you think his categorical imperative rejects magicians?
[00:54:36] Speaker A: Absolutely. I mean he absolutely.
[00:54:38] Speaker B: Oh, then, then, then, then you're at least self consistent.
[00:54:40] Speaker C: Yeah.
[00:54:40] Speaker B: Okay, then I, then I reject Kantian ethics.
[00:54:43] Speaker A: I mean this is, it's not a utility calculation for Kant. Yeah, exactly. Like, I mean, I mean he would tell, he would tell Anne Frank. You know, you would tell the family holding and Frank like you got to tell them where they are like that's what philosophy professors will also bring up as the example because it's the most extreme and illustrates illustrates his philosophy.
[00:55:01] Speaker B: Yeah well I you know I guess I will I will not evaluate game design decisions based on Kantian ethics. If he doesn't like magicians man that's.
[00:55:08] Speaker A: Going to be an amazing hold open quote.
[00:55:10] Speaker B: I definitely thought a lot about it after that you said it the first time and I was like huh? Like what is like why do you lie to people sometimes and like why is it okay to lie to I mean also I have a I have a kid who I lie to of people a lot the time so it's like why do I lie to her? Like when is it okay to lie to her?
[00:55:25] Speaker A: I'll talk to you soon man Night. Sorry.
[00:55:28] Speaker B: We should teach this to our children. Economics is major.
Everyone has to major in economics number one for personal survival Economics is major.
