Beyond Solitaire Podcast 167: Jason Carr on Solo Modes at GMT
Transcript
it's adbreak time the Beyond solitary podcast is proudly sponsored by Central Michigan University's Center for Learning through games and simulations and as usual they are up to some amazing things they recently concluded a successful campaign on backer kit for eyeball to eyeball a lar about the Cuban Missile Crisis late pledges should be available soon also just a quick plug for myself if you want to support the show and help keep it going uh please do go to patreon.com beond soliter which pays for all of my hosting and subscriptions associated with the podcast and channel uh thank you so much for listening and let's get on with the show hey Gamers this is Liz Davis from beon soliter and this week on the PO I have a repeat guest I'm very excited to have this is Jason Carr uh what is your current position at GMT now I'm director of game development and head of GMT 1 excellent uh but we are here to talk today about my favorite thing to talk about and perhaps the bane of your existence uh which is solo gaming GMT it's not the bane of my existence I I it's it's actually how I got started so like I can't call it the bane of my existence so how did you get started with Sol have we have we talked about this before you know I think we talked more about how I got started at GM M and less about how I got started with solo gaming so that's that's a related story but it's different yeah let's have it yeah so I worked on tank doel um which is Mike berto's game about um about tankto tank combat during World War II um little bit upfront little bit Combat Commander and he uh he taught me how to play the game and this was back when I
didn't have a gaming group I didn't have friends I don't I still don't have friends um and I really wanted to play it some more and it's a game so playing multi-hand was just awkward and I only had vassel I should say that like I was playing on vassel by myself so every time I wanted to uh change to a different tank or the other side I had to like um sign out of the game and sign in as the other side and it just got really annoying which like I know lots of players can do that and especially it's especially good for hex encounter games where it's I go you go and you have to move like 60 units and then switching sides who cares that's not a big deal but for tank duel where it's very fast turns changing sides all the time with annoying so I said you know I bet that I could make like a little robot that could play the other side it didn't have to play well because I didn't know how to play well it just had to be able to like mechanize what was going on and like follow some rules so I modified the vassel module actually no I take that back I think the vassel module was already modified so that you didn't have to always sign out and sign back in um I think that had been done like right right after I got annoyed and um I just kind of put together a little little flowchart that it would follow and I had no idea I didn't play solo games at that time like i' had never really been into them I had no idea what I was doing um and I played it for a couple days and then got up the nerve to show Mike and say like hey I did this thing what do you think and he immediately jumped on it as as like really a good idea um and so we kind of designed that together
over the course of a year and what is in the box is shockingly similar to what I put together at the very beginning oh that's awesome so you were a solo gamer and you didn't even know it yet that's I also think it's very funny that somebody adjusted vassel so that you didn't have to sign in and sign out because honestly I feel like I kind of feel like War gamers are like the OG solo Gamers um but that it starts with that wanting to play Both Sides yeah uh I mean I think all of Joel Joel toppins modules have like a player where all the information just gets unid and so he did the the tank tool module so there has to have been a solo player and it I maybe I didn't know how to use it I mean that's entirely possible because I didn't know anything about vassel either um so it was either that I got the module and I didn't know how to use it or it wasn't there yet and so I worked this solo thing out instead so necessity is the mother of invention right indeed indeed um so your first like big design project that it sounds like was in fact a solo mode yeah I mean up to that point the only the only real game workk I had ever done before that was doing a little bit of development um and play testing and the development stuff I had done was real arms length at a distance stuff where where the designer or the design team or the developer would say you know we have
a we have a two-player variant that's not quite working right would you be willing to test it and give us feedback and I took that to like oh I fixed it like I figured out how it should work um and in some cases that was welcomed and in some cases it was not as I'm sure you are well aware uh but this was new this was like the designer and I were on the phone we were just talking about it we were uh working in close proximity and so ideas were very welcome and so there's a lot of stuff that in tank to that's just started from me asking Mike a question and then him saying well I don't know how would you solve that and so I would throw something out and then he would throw something back and we'd go back and forth till we landed where we wanted and that kind of stuff um is really rich when you're trying to understand how a system works which is obviously critical for solo design so that that kind of access made solo design accessible to me in a way that it wouldn't have been before oh that's really awesome um and then you know I think about GMT there's so many different I guess expressions of solo systems now so I've talked with you and with J do hear some about coin solo uh and I know that there are actually protocols for that at this point like a huge document about how to do a coin solo uh I I would actually really like to see this document at some point U if you're willing uh but you know what how do you get from you know how do how do you have a a document that's like that for a whole system of games and then I
want to use that as a Cornerstone to talk about like other types of work that that is happening at GMT yeah I think that's that's a great question so let's start by talking about what GMT 1 is because I don't know that anybody actually um outside of GMT really has a clear picture of that so um GMT 1 is not just like an internal Design Studio when when it started the idea was me and a couple other people like Bruce would be available to designers to do consult or to do design for them so designers who like doing soliter systems would be able to reach out to us and ask us questions and ask for resources and designers who don't would be able to um ask us to do it for them I guess was kind of the idea and um one of the things that one of the things that came up very early on was that uh we would constantly ask each other me and Bruce really would constantly ask each other while we were working on Gandhi um well why is this better than that like why is it good that we remove bookkeeping from the system why is it good that we separate the um action the action selection of the bot from the the space selection of a specific action and so we would we would get down the rabbit hole on that and we would get into things you know okay well it's good to separate out the different types of decision the bot's making because it
creates for cleaner ux and an interface design um when you're thinking about a player interacting with it but it also creates reusability if a if a bot needs to select a space over and over for all different kinds of actions then you can unify all that into one space and that's great but we felt like we were having the same conversations over and over again and so as part of that we came up with a set of principles that underly um that underly the design that we do and I published um articles about this on gmts uh blog inside GMT um and these are things like we value um Simplicity over cleverness we value uh usability over thrownness things like that um and the reason that they are phrased that way is that those are always trade-offs that you make right so um it's possible to have a bot that covers every single interaction that could happen in a game like it it's hypothetically possible these games are closed systems there's no undefined inputs in them um so you could theoretically with enough work create a bot that would handle everything that could happen in a game but that wouldn't be worth it because if you did then you would have something so complex the complexity of that would far outstrip the complexity of the game that it was automated so take shortcuts right we we we we use heuristics we come up with overarching principles we say things
like um we we eliminate aspects of the game so in Gandhi the the Bots don't track resources uh that's to remove an entire series of edge cases around how it should behave with forward-looking to try to figure out if it's going to Corner case itself right so out of that grew an ethos for for soliter game design and that's present in um all the games that gm1 has worked on so all of the coin Bots that they've done that we've done all the autama that we've built for Flash Point South China Sea Twilight Struggle Red Sea um for red flag over Paris for um uh in the fields of fire work that happened we've we've used these sets of principles to to guide the work that we do so for coin we wanted to make those Bots as accessible as possible for people that were not me to build them and not Bruce and so I sat down after we finished all the work on fire in the lake the the Trung bot and I wrote a document that is just a step-by-step process of how you would go through a coin game and build a set of bots for it um and Joe used that document to build the Bots for um for Sovereign of Discord and for people power and and it's been an approach that just kind of works and it talks about how to apply these different trade-offs with the different things that the coin series generally presents to players and
then there's all always a novel space so um in in people power the novel space was the new sequence of play which had to be figured out yeah and then um there was some novelty around the way that the NPA should behave with having a peace like a peaceful protest option around striking or a more direct violent option through Terror um and then obviously the big the big change was just that the uh government faction was criminals so you know you had you had a different approach and style of play that had to be baked in but all of that was accommodated within the original system and the framework that I wrote enabled for fairly easy iteration into the solutions of those problems so Kevin krooks who's doing the bot for uh Cub Libre using the same document using the same approach um and so if somebody were to come to me tomorrow and say man I really wish there was coin Bots for a distant plane I could hand them this and say give it a swing see what happens uh you might you might find that this isn't um as easy as it seems but it's also not creating from scratch oh that's really interesting um how has that documentation changed or did I guess it came you see you mentioned working on Gandhi is kind of where y'all got started with it um so was this also tied to sort of the emergence of the card system as as part of coin bot lore and action yeah so
prior to um prior to the the Gandhi coin bot system the cards um most of the Bots had been done by the same person um I mean I think I did I think vco did the very original charts and um vesar ponin did the chart for Colonial Twilight but everything else I think was done by Iron and um when when we decided that okay we're going to use the card Bots for games and we're going to start retrofitting some of them to use the card Bots to be clear that wasn't because aran's Bots were bad in fact I think they probably play better in some cases if all you care about is I want a more competitive bot um he he has different trade-offs than we did in his design and so there's people on the forums who have run the Bots against each other and have compared how they behave and uh for fire in the lake at least I think his are probably more competitive um but my theory and the working theory of GMT 1 is that most people who play solo games do not do it just to challenge themselves if what they really wanted to do was just challenge themselves there are puzzles out there that are way more effective challenges than playing a board game solo um go do Sudoku like there there's other options but if what you want is to experience a narrative and as part of that narrative to be challenged either to compete within the framework of the narrative or to question
assumptions about your function as a player within that narrative whether historically or systematically then a bot can be a way for you to I guess in essence bounce off yourself um or bounce off the designer's intent in the case of coin and um that opens up really interesting narrative possibilities and so we felt like the card Bots better showcased that opportunity they were easier for the players to remain in the mindset of the faction that they had taken and they were less removed when they used them now your mileage may vary you may say Jason you're an idiot that that's not how I feel at all pay more power to you um but for me at least when I play with the coin Bots with the cards I find that I'm much more immersed than if I try to use the the flowcharts so we started looking at using these for all the games going forward and I didn't want to put all of that on one person I didn't want to say hey Bruce uh better quit your job you're making coin Bots for the rest of your life slave away in the coin mines um make coin Bots the rest of your life like I I wasn't going to do that and so making uh making it possible to build a team and have multiple people work on them was really important so now we have four people who can do that um and Joe really likes it so he tends to do the bulk of the work but like just today he told me that he's done with the Bots for China's war and
so those now go to the other three of us to all play and just do a final check like we've all been involved in the process but now we can get final feedback and sign off on them together so it's a lot more of a team effort than one person you know going off on the going up the mountain and coming back and saying here I've I've done it so this actually leads to um right where I was going actually how convenient so uh there are four people on the team now is way more work than one person can do what has there been a change in sort of the the design load of creating solo modes in GMT over time and what do those Trends look like that is a really good question um so I think there's there's two parts to that answer the first is how much more we are trying to do it like what is the market asking for and how are we responding to that and then there's the question of how solo design has become more sophisticated or complex and the ways that we are um incorporating that complexity into our design process so from the market side I think that's an easy thing to answer I think that in the last 10 years it's become a market expectation that games will ship with solo modes um especially more complex games simpler games I think you can get away a little bit more with not having a solo mode because it's easier to find opponents um some niches in the hobby it's easier um some games aren't solo suitable and those games the markets for
those games tend to not want them um yeah but on the whole historical games especially where there is uh unpleasant history being portrayed have a high demand for solo modes I think Mainline Euro games like the main market for Euro games also has a very high demand for solo modes um it's really just in High interactivity games um the classic example here would be something like uh diplomacy or uh you know any game that's negotiation based um or you might also see that a little bit in games with like extensive amounts of hidden information so given that the Market's going to demand them and that game sales are better when you have them uh we dedicate resources to helping our designers do that we do not mandate it if if our designers don't want to do a solo mode they don't have to um but based on what series the game is in customer expectations around that um sales of the game like sales of games in that genre we will make recommendations to designers and say we really think you should think about adding a solo mode and usually our designers are happy to take a crack at that because they're curious people but sometimes they say yeah this design space really doesn't appeal to me so we support that and that creates workload for GMT one so at any given time I'll have between two and six projects that are gmt-1 projects that I'm supporting and um that are specifically Bots we also do support for gamers that
for game designers who are doing solo only games um GMT 1 did support for Mr President gmt-1 has taken on fields of fire and systematizing that for the future and providing a team for Ben Hull so he has support and so we can do 10 more fields of fire games um and then you know there's solo games like chunga um that that we are working on but for the purposes of this conversation I think the solo bot side is what's really relevant because if you're a designer and you design a solo game we can help you but you're getting regular game development at that point so um yeah so that's the market side right the the other side of this is the part where I think the real workload comes in which is related to the market but not not the same and that is as games get more complex or as we start to put Solo modes in more complex games the the solo Mode's complexity is super linear with the complexity of the game and what I mean by that is every increment of complexity if you can imagine complexities like blocks and every time you add another block to the game that is an increase of complexity well for every block you add to the main game you have to add like four blocks to the solo mode because every component of a game interacts with every other component of a game so when you add one unit of complexity to a game you're
adding n minus one new interactions to that game so it's just this super linear thing that complexity can take off as you add complexity to a game that can blow up the complexity of a solo bot really fast and so there's a lot of techniques that solo designers use to try Tamp that down um some of them would be only designing solo for a part of the game um so for in the shadows um which is a very simple game by the way in the shadows is a bad example of managing complexity here but there there's only a solo bot for one side yeah and and the reason for that by the way is um you can play solo as the French Resistance but you can't play solo as the Nazis and that's a design intent thing like the the game is not intended to illustrate what the Nazis were doing that's well understood um the game is intended to illustrate things about how the French Resistance worked and their interactions require a Nazi player and uh we think that it's that it's enlightening and interesting to play as that player and to take on that historical perspective but we weren't going to put that in there as a solo option like that seems to incentivize the wrong kind of interaction with the game so the designers made the decision that they were only going to offer solo play as the friend resistance that's cool it's also easier to design that because that's half the game that you don't have to think about and so what what Joe got to do Joe Schmidt who's one of the designers who designed the Solo System was poor all that additional
effort that he would have spent in making a German bot and he just made a really really really I'm sorry all that additional effort you would had to put into making a French bot so you could play the German solo and instead he just poured it all into making the solo play as the French as interesting as possible so uh we could talk about that game more in a minute because I would really love to come back to that but that is exactly the problem across all solo mode design is where are you going to put your effort where is the time spent what's your complexity budget what are players willing to take on to run this thing and to interact with the game and how much time are they spending running it instead of making choices on their own and enjoying the narrative the game presents so yeah managing that is I think the biggest challenge and biggest workload and so it's not number of projects I get a 50 games that I'm working on solo Bots where if they were all simple ones but when you get anything that's even that's even a little bit more complicated than than a card game the solo modes can become a lot more challenging to design it is so interesting that you mention that because I feel like there's a fasinating tension in that answer in that people who want to play more complex games often end up being the people who demand solo modes but then we also don't want to make solo modes overly complex because they should at least according to Modern design philosophy be easy to run it's totally true and it's why that's why we don't try to build solo modes for our hex
encounter games like those games are easy to play Both Sides because they rarely have a lot of hidden information um and the ones that do have hidden information often it's easier to just make a set of rules that accommodate solo play rather than try to make a bot and the the amount of work to create a bot is just insane because the decision space in those games is so open yeah um so the few hex encounter games that are playable solo typically those games the solo modes were so much work I mean I think about um enemy action Arden and enemy action kov by John Butterfield great games phenomenal games uh they are three separate games in a box they are a game playing two player and then a game where you play each side with different rules to accommodate each side and so there's a lot of commonality between those three games but they are clearly three separate games that you can play um and that was part of how John coped with the complexity of all of that was he just didn't try to make one system that could play everything that's really smart but as a player it means that there's an investment involved the payoff is absolutely worth it if that's your kind of game um but I think there's a good reason why most of the games that are hex en counter that are Solitaire are games like DJ Omaha Beach where there's effectively a static side and a maneuvering side and you control the maneuvering side
um and I think that pattern is in a lot of historical games um that the bot takes when it's a solo game that the bot takes the static side but when you have to do um when you have to automate what a second player would do if the game's designed well both players have agency and both players have interesting choices and so you don't get the cop out and that's where that's where all that complexity comes from is really if the game is interesting to player and you try to make a solo bot for it the interest the stuff that's interesting is usually the hard part in bot design yeah so you know we talked about you know some games are just not soliter suitable if they have too much hidden information or you know if you know there there are a lot of things that just don't work with a bot um although it's really amazing what somebody like say Ricky Royal can pull out of a negotiation game like John company like we we're pushing boundaries as a field but what is the line for you um other than sort of the more obvious like you know too much hidden information or this is a this is diplomacy why would you do that solo um the SEC yeah so the second one matters way more to me than the first so this is where I completely agree um with this is where I completely agree with amabel Holland um yeah with when NAA came out she got asked a lot well could you do a solo bot I don't know if I'm ever going to get six people around a table and her response was well the whole point is to get six people around a table and have them argue about this like yeah that they you need the players the game is in
the interaction between the players and if you're not interacting with a player then you're not playing the game and I think that that's true a lot um and it's even been true in situations where I have said I will still design a bot for this but I think that when that is the message of the game when the point of the game when the thesis of the game is tied to the way that players relationships at the table change over time and John company is like that then I think that designing a bot becomes a bigger Challenge and often the payoff isn't there sometimes it can be but but a lot of times it's not and so um you know I haven't played the Solo System for John company I really I'm eager to but I think that that experience for me of playing John company was all about the way that my interactions with players changed through the course of the game and and the way that our incentives aligned and then unaligned and it's hard for me to imagine a bot being able to do that but it's not hard for me to be able to M imagine a bot that could create something else interesting that is also in the game and emphasize those things which is I think how the bot in PA Premiere works like when I play Pax Premiere I don't have the same experience solo and two play and two or well not two player but three plus player um but it's still very enjoyable as a solo game it's just a different game and that's okay too as long as you're honest about that in your game
design that that this the thing that I designed solo may not be the same it might just be similar or it might emphasize different things about the game yeah I mean a solo bot I think I mean obviously I love to play against any number of different types right but they're never actually going to replace a person I do think that when you go into a solo game you have to accept that it's always a slightly different experience yeah well and one of our one of our principles at for GMT 1 is we want the player turn to be as similar as possible but that's mechanically um the decision you're making is in is inherently going to be different because you don't have an opponent so you are um you're not going to be outfought right there's nothing to read nobody can it's you're not going to have the oh but if I swap the cups you know sort of thing yeah exactly do you have a complexity cut off as well like how what what bleeding edge of complexity do you think is too complex for a solom mode at this point because I feel like I feel like we're all a little bit crazy if we're playing game solo and will really go far in our play but as a designer where do you have to like just no more for you tonight go home like yeah it's hard so let's talk about coin some more and then we can go further field than that but for coin one of the really interesting things is it's actually simpler to just play all four sides and to try to keep what all four sides are doing in your head so just
imagine that for a second that like it's actually easier complexity wise just in terms of the sheer amount of information you have to process to play for different people than it is to play one and have the bots so so that right there makes you say like okay the person this is for is someone who is looking for a very particular kind of experience and I I say that knowing full well how popular those are I mean people play them they use them I know because I answer questions about them and so I know how many people are using them and I know I get reports from people I mean I get geek mail from people monthly if not more often it used to be a lot more frequently Thanking us for making them and and telling me how about their experiences using them so you know I know there's always this open question of like these solo designs get get built but you know if a person plays a game in their basement and nobody sees them that it really happen um and and like yes it does actually get used um so for me that's sort of one of the complexity thresholds is this more complicated than the coin Bots because I know the coin Bots get used I also know a lot of people think that the coin Bots are too complicated and would rather just play forehanded yes and so my goal is to always push that down to a place where I would say the majority of people who are playing this would choose to use the bot and I don't think we're there for coin um but I think I was there for flasho South China C so relative to the complexity of the game I don't think my
bot for for flasho South China see is that hard to run um if you can play yeah if you can play the game you can you can run the bot um so I think that's the the tradeoff and I would love to have a more formulaic way to look at a game and say oh this is going to be too complicated but but the only way to know that is to get stuck into the project and then realize that you've made a horrible mistake so oops yep and there's no way out of it at that point because you committed to doing it so so then again necessity becomes the mother of invention and you have to figure out how to solve your way out of it which isn't a bad thing at all like sometimes saying yes I can do this and then finding out you don't know how is the best way to stretch yourself and grow to be able to do things that you couldn't do it before fair enough although also terrifying absolutely terrifying utterly terrifying uh so GMT one you know we the coin system has a Solo System almost at this point so you can kind of do a little bit of Team working on it and you know um there's documentation that can give you a place to start but what who is managing the non-coin solo games is that also a GMT one thing and then how are those projects handled without that kind of Base material to work with yeah so I manage those that's what I do leading GMT one um and I work with our solo designers um who like working on these projects to farm them out and get
them done so um you know for Yar Republic um wear Republic is a game that has sensitive um Topics in it like you know the rise of Nazism in wart Germany is the PO is what the game is about and um the designer Gunner really wanted to to show how this was a matter of agency and so he felt that a four-player treatment where one player actually plays as the National Socialist was an appropriate treatment of the topic and we agree but we also recognize that there are players who while appreciating what he is trying to do and maybe studying what he's doing and understanding his point would also not play a game like that and so the only way we would publish a game like that is to give players the option to experience it where there was a non-player recognizing and giving the disclaimer that the non-player experience doesn't fully tell the story that the designer intends to tell so when we got that game and we got into the design of it I didn't have time to do the solo bot for it and so it was a choice of like let it sit forever and and not you know I get to it when I get to it which wasn't a good answer or find someone who was passionate about it and wanted to do it and Schmid happened to be passionate about it and wanted to do it and I think the B that he turned out is really interesting it's a very narrative experience um while still being competitive it will still beat you um but it's really more than that about exploring the motivations of these factions and uh the bot can play all of
the factions so um you can take any seat and focus on trying to do what you're doing and run the other factions with a minimal amount of overhead um so my job there was just checking in with him making sure that he had the the resources he needed to solve the problems helping him get testers we had some good testers who really ran those things into the ground um and and made sure that he got all the feedback he needed and then you know he's he's handed that off and now it's just doing the final development on it and and I mean that's long done so that's how I function in that role is much more as a manager which is how I function for the development work as well um if I'm on the critical path I'm probably doing something wrong that's fair unless you know I also do some design so um I just did the bot for Bell of treason which was um this is the other thing that I thought was interesting about what you said so um Bell of treason is a Series game it's the sequel to red flag over Paris and um Fort Sumpter and that series the final crisis series um has mechanical coherence throughout the whole series like there's the mechanisms that are used over and over again so the bot that I wrote for a built for for red flag over Paris which I had backported to Fort Sumpter um was the starting point for the bot for Bell of treason and I felt like I was going crazy when I was working on it because the things that worked for red flag of repairs didn't work at all and the reason is that the
game playay in Bell of treason is fundamentally different the way adjacency is important is very different the victory conditions are tremendously different and and working all of those things into the existing bot uh was a big Challenge and I'm lucky that again I work with really smart people like I always try to be the dumbest person in the room um and that's easy when I share the room with people like Joe like he is so smart and he is so good at looking at systems and taking them apart and he spotted right away some things I were doing I was doing that were dumb and helped me fix them so I think that the the the importance of having a system to work from and having documentation to work from and H and having all of that is only important in so far as it helps you to build a good team um around you and and to share those ideas with other people the the other interesting part about that is as we build systems like the jard Bots which is what we call the system for coin um named after the jard loom which was driven by cards so it's a horrible pun um it's it's the best it's the best kind of pun um we took all of the different concepts and work that we had done in there and we just stole the best ideas and then ported them to other games so one of the things about that system that's so cool is the best part about it is the space
selection charts so we have these charts that have conditions um on this on the left and then actions on the top and so you you pick the column that corresponds to the action you're doing and you check down the conditions and for that particular action let's say Terror um there's dots in all of the conditions that apply and if a condition doesn't apply for a specific space on the board then you don't select that space and you narrow the selection down until you only have one space well that we didn't come up with that um Bruce didn't come up with that that comes from the Labyrinth Bots that Adam Z did for awakening Labyrinth Awakening and when I emailed him and asked him where he got it he said oh I have no idea but I know I stole it from somewhere and so we are basically like this is some kind of information design it's a Civ is what it is that's the data structure that's that's the thing that it actually is is a Civ but why that particular implementation of it where that came from is lost his time like nobody knows who the person was who said oh you can Implement a CIB this way by doing the conditions like this um but we use it everywhere the bot for red flag over Paris is the same it just only has one action per table so you don't have the things across the top you just have a list of conditions um and that principle can be applied in a ton of different ways and so really what I realized along the way I'm a software engineer for for my career was that bot design is all
data structures it's it's all data design like the same kinds of problems that I'm solving in my day job are the problems I'm solving for for Bots and so um you know for Twilight Struggle red SE as an example um the way that the action selection chart is structured is a visualization of a multi-dimensional array which is a data structure in in a a data structure type um and once I figured that out then the analogies all became clear and it was very easy to implement the challenge isn't necessarily the logic it's the visualization and the interface that players use to track it so I can tell you that's a multi-dimensional array and you go great Jason what am I supposed to do with that um and and that's how every player who picked that up would behave if that's how we presented it but we didn't we presented it as a table and you navigate through the table according to a set of rules and that lands you in a place where it selects an action and if we've done our job in presenting that information you shouldn't even be aware of the data structures that under it unless you're trying to dissect it so as we continue to build more systems for more kinds of games we are growing a toolbox that we can use to solve problems in other games and that's the other part of this is like we have a team of people and we have a toolbox of things that we can use so that we're not Reinventing the wheel I'm gonna come back to that um
but uh because there's a whole team of people working on it and so so we've got like two things going on right there's a bunch of different people who are working in the system and then there's also a set of principles and you can talk about this in like a very computer deterministic sort of way how much variance is there in the Bots that are being produced by people on the team and how often do you have little differences in what somebody chose to prioritize and how do you work through those so I love this question because this is really at the heart of everything that we we do at GMT um I have said this a bunch Jean says this a bunch we like games but we love people so we are a people driven company and it is weird like people when they really encounter the rubber meets the road of where that is it confuses them um I might have a way to design a bot that I think is great and I think it's really fun but if the designer doesn't and we can't come to an agreement after talking to him I would happily in fact I would much prefer to redo it to be something that they like and that they can get behind um and a lot of times the designer doesn't care you know I don't think I'm telling anybody any secrets to say that you know Brian train is not a big fan of coin solo Bots he just doesn't play them he's said that many many times so he's very happy to have people who love them design them because he knows that that's going to turn out a better result and so whatever we come up with he says great do you guys like this and if we say yes
this is for us he's happy um but there are other designers are more opinionated about how their bot should work and when we work with them we want to honor that all the time so right off the bat there's your variance um I'm never going to tell the designer how it has to be I'm going to engage them in a conversation about what it could be right but like also if you and Joe start with the same documentation for example are you gonna come up with the same bot or with a similar one uh probably wildly different if we're starting from Scrat match like it so so here's the thing is if we start from the same document if we start from that the jakar document that says this is how the system works we'll produce something similar because that document's prescripted but if we start from the same set of principles we're GNA come up with something different and the reason for that it that's not a surprise that's intentional the reason for that is that that set of trade-offs those principles are modeled after something called the agile Manifesto um and the agile Manifesto is a a software development Philosophy from about 30 years ago that is Scrum people are into yeah yeah and I you know there there's a dogmatic way to do that I don't subscribe to the dogmatism what I subscribe to is the original idea that like we are discovering new ways of making software and learning that we value A over B and so when Bruce and I sat down and we wrote out the principles of what GMT 1 was going to be we wanted
that idea we are discovering new ways to make solo Bots people have not done this kind of work before I mean in abstract they have it's problem solving right it's automation somewhere someone has solved this problem before but they haven't applied it in this way and so we want to learn from manufacturing we want to learn from information design we want to learn from user research we want to learn from other game designers we want to learn from the medical field one of my favorite things to apply here is this paper called how complex systems fail by Richard I cook Richard cook was a medical researcher so when he's writing about complex systems he also works from the military but uh when when when he writes about complex system failure he's talking about things like how does gauze end up sewn inside a patient but I use that to talk about how does AATA end up in games and it's completely applicable because we're talking about complex systems in both worlds so um circling back around because this was kind of born out of agile and agile really valued conversations with customers we also really value that conversational project practice that that that grows the requirements naturally out of a set of conversations and so we want to talk to testers we want to talk to the designer our game design is a conversation between the solo designer the designer and The Gamers and that conversation continues well after the game is published not just in terms of like on Board Game Geek asking questions but also in the game
itself we are engaging with dialogue as you play it there is a dialogue happening and so we work that dialogue all the way from the beginning to the end and so that's why Joe and I would have really different outcomes because Joe and I would have really different conversations and that would lead to different ideas that would lead to different implementations that would lead to a different solo experience oh that is super interesting to think about and actually it leads right back around uh to what I wanted to come back to uh which is you know we talked about a toolkit in a team and how you know every solo design problem hopefully gets easier as you continue to do them because you have something from the problem before even if it's new so if you have a game where you don't have a system for it you're going to try to make a solo bot where do you start I'm just like curious about this now because I have to think about this in my own life dayto day and never had before so I like knowing what other people are doing so the first thing that I do whenever I have to sit down with a game and design a solo system for it is try to break it down into its component parts so um all our games are made up of systems and those systems interact with each other and for me that's very familiar because in software everything is a system that's inside another system that's inside another system and it's just you know Turtles all the way down so the first thing that I would do is say okay here's the discrete systems in the game so um to use a specific example that might help with some of this that maybe some people are familiar with with
Twilight Struggle Red Sea the first thing I did was I looked at it and I said okay there's the hand of cards and the way that people are going to interact with that or the Bots can interact with that there's the map and the way that all of the things on the map work there's the Space Race which I have to think about and then there's scoring right so those are sort of like the big pieces and the map has systems built into it that are more granular the countries more granular so we had Flash countries Battleground countries and regular countries so the bot had to be aware of all of those things there's an influence value there's adjacency sorry there's a um a stability value there's adjacency and then you have the things that the players can actually do to those countries so placing influence coups in realignment roles and so you start to develop an idea of here's the concepts here's how they're interrelated that's a system and here's how those bundles of concepts are related to each other that's how the systems interact with each other and you start to form a mental model of how all of that stuff works together so it should go without saying that this requires understanding the game at like a relatively deep level you can't just parachute in and say well I played it once I can take a crack at this like you have to dissect it and so for me there's usually um a two-part process there there's my play where I might have sit with my wife or my gaming group um or my friends that I don't have and um
play through the game and just enjoy it just try to play it experience it understand what it's doing then there's the intentional play so this is a little bit like the difference between a musician listening and active listening right when they're listening they're just listening to music they might notice things that are cool because they're musicians but when they're active listening they're trying to remember the structure of the song they are actively thinking about the Dynamics of how that song flows um they are they are building a mental model that the song fits into so they can play song and that's what I'm doing when I am actively listening to a game as I'm playing it and saying oh that's interesting it does this thing I wonder how many cards in the deck do that so I just grab the deck and start flipping through oh okay 13 of the cards do that out of 52 so there's a one in four chance that this is going to happen base but then the game's manipulating the deck so that's how this is going to change and I look for the feedback loops I look for the places where the game is accelerating itself and the places where it's decelerating itself so in pandemic every time you get one of those epidemic cards and you flip the deck over back on top of the draw pile for the infection Stack The Game's accelerating itself it's it's reinforcing something that's already happened so that's a positive feedback loop in the game so I'm looking for those kinds of things which exist in two-player games for sure where where you know it's a self-reinforcing feedback loop so I'm pulling these things apart and then I start saying okay where are the inputs so now I understand the system and I might make diagrams I might make spreadsheets I'm trying to quantify and qualify what's
going on in the game now I'm going to say okay where do I start to modify the system where are my levers what are players actually doing so again back to Twilight Struggle inputs action selection card play that's it right like you pick a card and then you do something with the card oh so easy so you're either gonna play it for an action you're gonna do a influence placement you're gonna place a c you're going to perform a coup you're going to do a realignment rule or you're going to space it and so when I designed the bot one of the questions was well when should the bot space send a card to space and I just said I don't want to solve that problem right now because it's too tied in the card selection part and so I knew that that was a problem but I set it aside I said maybe a solution for that will come later it did when I just decided that the bot wouldn't have a hand of cards at all right and so I eliminated an entire problem space they don't need to manage their cards they don't need to sequence them they don't need to send one to space because they don't have a hand of cards and the way that I enabled them to not have a hand of cards was I put a new thing in the game the card line which holds cards for them but is not a hand and then I have sometimes you have to make a concession to limit complexity so in Twilight Struggle the biggest piece of complexity the thing that brain Burns that people love about the game that is so crazy integral to Twilight Struggle
is when you play a card and it's your opponent's card the event happens and so from the beginning I had to figure out how do I manage that right the bot not only has to pick a card the bot has to forward look at the effect of that card if I get an event which means that they have to kind of figure out what I'm going to do with it and then that just exploded the problem space it was impossible and so I remember sitting one time taking a shower and one of the coolest things is like okay so your brain the way that your brain works is that it creates a ton of memories and then when you have rest it synthesizes those memories and decides what's important and that's what you remember so that's how you make your highlights of your day if you don't rest if you don't have downtime if you're not on if you don't have time when you're not actively stimulating your brain then you will not get those insights you need that that's why shower thoughts happen that's why I like to go for long drives because my autonomic system is driving the car but my brain is free to just do whatever and so I have these insights I have these realizations I solve my hardest problems whether software or board game design driving or in the shower or right after I wake up um and so I was thinking about this problem that way I just put it in the back burner and let it simmer and I realized the player needs to have that experience but the bot doesn't the bot doesn't experience anything so when you
think about Twilight Struggle you think about the experience of pulling that hand of cards and going I got nine cards and I have no idea I got to figure this out what order do I play these in what it's important to me as long as I preserve that it doesn't matter if the bot does it or not because it is not a person and so when I had that realization and I just said oh when the bot plays a card it just plays the card it doesn't trigger your event the complexity went down by like 60% and that also solved the Space Race because now they don't have to get rid of cards to the Space Race so they just auto advance on the Space Race that creates pressure for the player the player has to respond and so those kinds of solutions to problems that's how I approach game design is I try to focus on like well what's the hard problem solve that and then let the rest of the problems kind of flow out and sometimes one hard problem begets another but a lot of times figuring out the core problem for a game in Twilight shet was obviously hand management figuring out the core problem solves everything else because your your approach to the hard problem kind of dictates or guides your solution to the other problems and then it's just a matter of stitching it all together so you already understand the system you understand the inputs you understand how they interact now you're laying your different bot systems on top of that and you're creating the senu that ties it together and if you do it well you have something fun that's a really interesting way to think about that I uh I I love the the fun
like surprise realization that a bot is not a person because like I do think that we forget sometimes yeah we're like oh well but a player wouldn't do that who but I think that when players think about well when they talk about playing solo games like a lot of comments or kind of requests or thoughts that I get are like oh well I want to feel like I'm playing a real person but yeah one I'm not really sure B I can never fully do that uh but two in order to make you have the best possible experience for you the bot in many ways cannot behave like a person at all right right yeah and so this is where we get caught I think that when we expose the inter as bot designers tell tell me where you disagree with this I guess but but I'm curious your take um as bot designers when we expose too much of the internals of how the bot is making a decision we fall into this pitfall because people can see the mechanism and so they can go ah but a player wouldn't do that and what they do is they start to focus on the individual decisions the bot is making instead of the overall effect and almost 100% of the time I want the overall effect and so the bot makes a bunch of suboptimal decisions because those are easy for me to come up with the logic for but the because it because it's either getting an advantage or it's cheating or it doesn't have to do something the overall effect if you just look in between the bot turns looks like a player maybe you have to squint a little bit but it looks like a player
yeah it's just that the mechanisms used to get it there don't resemble a player yeah no that actually sounds a lot like conversations that Ricky and I have about Oath like we talk about kind of strong and a little dumb yeah because if you make a bot that's too smart you fall into the complexity trap Absolut but you have to make a bot that has has enough kind of Brute Force about it that it could win so there has to be a real threat from it and it has to make you as the player play out what you have uh as much like a normal player as possible so like you know the servant bought for oath right now needs work obviously um that's what I do all day um but um you know I've been testing it and it's able to either beat or it's able to beat human players or make them work really hard using everything they know about Oath so yeah to me that's the important part is making sure that the the bot turns are as non-intrusive as possible and I think we can shave down some stuff right now as it is um but letting the player get back to that normal turn I don't actually you know I kind of don't want the player to think about the bot too much at all yeah yeah I think having the player turn be the player turn yeah is that that's what makes it the same game that's what makes it a solo mode for a game when a player is taking a different turn you've created a different game um you said something interesting there and
I do want to call out one other so that's I think the majority of Bot players are what you just said they want to experience The Narrative of the game they want to feel like they have an opponent um yeah games are when a game is competitive having the opponent compete with them is an important part of them feeling like they're experiencing The Narrative of the game so that makes sense and so strong and a little dumb is great but where that doesn't work is when the game has a very specific model and the decisions that are being made are intrinsic to that model and so what I've noticed is we we Tri building autom style Bots for coin and players did not respond favorably to it they didn't take long to play but the players felt like they had they were playing a a a an abstract game like the historical flavor was lost and so one of the interesting things is lot of gmts games where they are very explicitly modeled um the Bots need to honor that and there's two reasons for that the first is a lot of times people because the games are more complex people use the Bots to help them learn the strategy for the game and so if the Bots don't follow the strategy then they're useless as an educational tool um and so in like in coin the Bots are a little bit of like almost a strategy guide um even the card Bots still pay attention to the same key things that a player would want to pay attention to where do I have enough gorillas to place a base is this a space where I could um gain patronage
by siphoning off support those kinds of decisions it's paying attention to those but the other part of it is coin players really value feeling like the bot is an intelligent opponent and the intelligent part is what matters there that the a little dumb they that's a problem and so I think identifying where that's true and where it's not is one of the really big design challenges which games can get away Flash Point South China Sea can get away with that because it's a lower complexity game and so having an opponent that's not thinking and making decisions that way is fine yeah um but a game that's more complex where people are investing more time and energy into playing the game in the first place they generally are willing to invest a little bit more time into the Bots if that bot gives them more insight or more narrative or more competition or in the coin case you know more model uh to hang their hat on and that is I think at the end of the day the reason why not every bot is a fit for every player right uh because because there is a design ethos behind that Solo System which is usually you know you and me um and is and is trying to portray something and in some cases that is an extension of the game design itself yeah oh that's a really interesting way to put it especially because again this is also a thing where I feel like as a designer you have to make decisions about the game where you take feedback or Express desire from players and you have to hear it but you
can't necessarily give them what they said they wanted if that makes sense because somebody who says I want this to be really easy and to feel just like another player and this and it's like okay so let me identify like what it was you wanted to feel and then try to make you feel that yeah but like yeah I mean how to get there is quite variable it's it's super challenging and it's like at the end of the day you have to listen to the feedback you're being given and and take that feedback and think about it to figure out what it means because a lot of times what people say isn't ultimately the feedback they're trying to give you and when you when you look at feedback and aggregate that can be easier because if a lot of people say the same thing than you know they're saying the same thing but especially it's especially tempting to go for that feedback that's Broad and common and you get a whole bunch of it and miss the things that people are telling you in the narrative that they're giving you about how they felt or how they interacted with the game and so I think um even earlier today I was talking to a game designer and he said I just don't want the player to feel bad here and I was like whoa whoa whoa hold on did the player do something that they deserve to feel bad about do you need to provide a negative reinforcement here and and or or is the game just randomly saddling them with the penalty for nothing that they did and and then the other question is can they change the board state do they feel bad and that incentivizes them to do something and and so what you
might get from a player is they say I don't like this it feels bad but what you have to dig into is well does it feel bad because I need it to feel bad because it's gonna incentivize you to do something or does it feel bad because you feel like you lack agency or does it feel bad because you feel like you lost options and those are very different things and as a player you might say this feels bad but as a designer I have to interpret that feedback right and figure out what to do with it so um I tell I tell my designers all the time rough edges are fine in our games at GMT we like rough edges we want we want people to feel a little bit like they're throwing elbows when they when they play like some sharpness to the game is okay I feel like far too often in the hobby nowadays we're and this isn't all like the opposite of this is also true we're seeing a wider variety of games than ever so don't take this as too much of a generalization but I feel like a lot of games that are intended for the broader Market have all the rough edges sanded off them till they're just kind of featureless characterless balls and so I say you know like we don't want to make spherical games where we've just sanded everything down to being perfectly round and harmless like we want games that have some sharp edges on them yeah no it makes some sense you gota it's got to have a little teeth little bit yeah little bit little bit and you know if if the game can make you feel things it might be doing something that I want to pay attention to so yeah well I think that's especially true in Solo gaming because if you don't have
other players the game itself has to help you produce the feelings you're trying to get from gaming that is absolutely true and that's why solo gaming can do some of what it can do because you don't have other players so yeah so we can reprise this another time but sounds like we have more to talk about sometime always but I've got I've had you on the line for a while this time around so um speaking of just games and things that have teeth what what is like a toothy game recently you played that you thought was really good so the thing that I've been into a ton lately which is if you if you're watching the video of this will be no surprise to you is Star Wars unlimited um so I am I am just in love with that game um I it's going to easily be the thing that I have spent the most time playing this year and it has been a lot of design inspiration for me and also a great way to reconnect with my local game store so I decided to buy in I've been going to events I've been paying money for the the collectible card game hit um but but honestly I think the game is just so smartly designed um and I really enjoy the the deck building in it the way that they they chose to do deck building instead of instead of having um well people who haven't played it like it it's very similar to magical gathering in terms of like the concept of the game and how it works but um couple key changes are all of the resources are colorless so in Magic your your lands are associated with different colors in
in Star Wars unlimited there's one type of resource um and instead what you can put in your deck is determined by your leader that you choose for your deck and then the base that you choose to go along with them and that makes seal and draft The Limited play options where you have to build a deck on the spot super super super fun because it's easy to think about what you should do and you get to focus your mental time on like actually which cards do I actually want to have in here but you're not sitting there thinking about like oh how am I gonna avoid avoid Mana flood like um and that is a more interesting space for me so that has brought that genre to life I never wanted to play these kinds of games until this and it's um it's been a revelation for me so yeah that's been that's been it for me this year it has been yeah full of sharp edges super fun really enjoying it oh man even I'm tempted now that would never be get some starter decks smash them they're great it is it is um it is it's getting a lot of hype and I think it's worth every bit of it so for me that's the game fantastic and then if people want to follow your work online where can you be found yeah I'm on Twitter um or I'm sorry x uh and uh you can find me or um I post on insid GMT insid gt.com that's that's the place to follow kind of the more uh thinky part of what I do because I typically post my design thoughts there fantastic uh and then for
everybody else out there who's listening um I'm everywhere is beyond solitire uh Jason thank you so much for coming on this is awesome yeah thanks for having me all right so hopefully we'll talk again soon but for now everybody uh please like subscribe comment ask questions and most of all happy gaming