While there are a few board games designed specifically for single players, most assume two or more. It’s becoming more common for those to offer some form of single player experience, ranging from what amounts to an afterthought to a complex automated opponent, known as an automa. I’ve taken a look at various options elsewhere on the Zatu blog, but recently I had the opportunity to talk to Morten Monrad Pedersen about automas and solo play.
Morten designs games (e.g., Skoventyr, blogs about gaming and design, and of most relevance here, creates automas—his company, Automa Factory, is responsible for the solo experience of many well-known games including Wingspan, Patchwork and Terra Mystica, to name but a few.
Welcome, Morten, and thank you for taking the time to answer our questions.
I’m always happy to talk to people who’re willing to put up with me yammering on about my work 🙂
I gather your first commercial automa was for Viticulture. How did that come about, and afterwards, how did your working relationship with Stonemaier Games develop?
In 2012 I travelled to Tuscany with my family and some friends. While there we went for dinner and wine tasting on a small vineyard. When we got back home, I saw the Kickstarter for Viticulture, which is a game about running a vineyard in Tuscany.
It was a fun coincidence, so I took a look at the Kickstarter and asked Jamey Stegmaier whether he would share the rulebook with me (that wasn’t common on Kickstarter back then). He did so and being a nitpicker used to doing proofreading I can’t read a text without trying to improve it and spot typos, so I emailed some corrections and suggestions for the rulebook to Jamey.
We exchanged many emails with Jamey updating and me nitpicking. After a while I went beyond nitpicking and created a prototype of Viticulture that I played with my wife quite a bit. That led me to post one of the first reviews of the game.
Jamey and I went on to exchange emails about game design, books and other geekiness at a rapid pace after that.
As Stonemaier Games started work on other games I began to help Jamey when he needed to discuss his designs, Kickstarters, and various other stuff. I also joined an advisory board that Jamey and his partner Alan Stone set up for the company.
In the meantime, I started a blog about solo gaming, which was one of the things that made Jamey come to me when he was asked for a Viticulture solo mode. After that I became the de facto “solo guy” of Stonemaier Games.
In my article on automas, I suggested—somewhat tongue-in-cheek—that the Sanctuary solo play ‘opponent’ (a randomised pile of numbered tokens indicating which tile to remove from the display per turn) was the simplest possibly automa, as it does have pretty much the same effect on a player as an opponent would (though it can’t win by its own actions). Do you think something as simple as that could be called an automa? What is it that defines an automa?
To me it fails to be an Automa if it doesn’t create win/lose criteria such as a number of points you need to beat. While I’m keen to try out Sanctuary because I love Ark Nova, I haven’t done so yet and don’t know whether it has a win/lose criteria.
I have a fairly specific definition of what an Automa is:
- It’s an artificial opponent that takes the place of a human player in a multiplayer game.
- The human player must play by the same rules as in the multiplayer game.
- The important player interactions must be simulated. This includes keeping the win/lose criteria.
- The player must face the same decisions as in multiplayer.
- The player must not make choices on behalf of the Automa except in rare cases where it makes sense because of a cooperative element or for thematic reasons.
- The Automa rules must be as streamlined as possible while achieving the above. This for example includes removing the internal state of the opponent that does not directly affect the player, e.g. the Scythe Automa doesn’t have a player mat.
On an ironic note, my own very first Automa, the one that brought the word onto the solo scene, fails to fully meet these criteria and had I designed it today, I likely wouldn’t call it an Automa.
For me, coming from the computer industry (as I know you do to), the appeal of complex automas goes beyond having an interesting solo game; I like the idea of these little ‘cardboard computers’ doing clever-seeming actions with a small, lean implementation. How do you think about the algorithmic nature of automas?
I definitely also think about them in terms of algorithms and I think that my programming background has helped me a lot in designing Automas and in boiling them down to procedures that are simple to follow.
Sometimes I do go too far and my frequent co-designer Lines J. Hutter has more than once had to explain to me that not all players think like programmers and that I should take this into account.
How do you go about designing, implementing and testing an automa?
I start with the 6 principles I mentioned earlier and focus on the idea that to replicate the experience of the game, the core interactions must be simulated. So, one of the first things I do is to identify the points of interaction between players.
Next, I look at what parts of the game don’t have a direct impact on the interactions. Those I will typically have the Automa ignore, so that we can make it simpler and faster to run. E.g. in Tapestry, the Automa doesn’t have a capital city mat (a mat where a player places buildings) because what happens on that mat doesn’t impact other players directly.
A key fact that I keep very firmly in mind is that the Automa isn’t playing the same game as the human player. It’s playing a simpler one that intersects with the one played by the human at the points of interaction.
Once I have a rough system up and running, I’ll ask 1 or 2 of my Automa Factory co-designers to join the project. They help me get it to a state where it makes sense to invite external playtesters.
We then add a few playtesters at a time to get feedback during a process of fairly rapid iteration. By adding them in small groups spread out over the development cycle we make sure that we regularly get fresh eyes to evaluate the Automa and that the final rulebook is sufficient to teach the rules.
As the process goes along, we’ll try to decrease the magnitude of the changes we make and, in the end, we’re just making tweaks for balance and rule clarity.
Should you be interested in diving in further, then I have a series of blog posts that covers my approach to designing Automas.
I imagine there are overlaps between automa design and game design, but also that they’re quite distinct. Are there ever cases where a game’s design hampers the creation of automas, or where you’ve suggested changes to a game in order to better support solo play? What were the problems and—if any—solutions?
Both game design and Automa design are tasks where you use game mechanisms to make a fun and smooth game experience, in both you have to consider complexity, decisions space, user interface, etc. The overall development process with playtesting and refinement is also similar.
On the other hand, when I make solo modes, I feel like a craftsperson rather than as a game designer. It’s not an option for me to change the game to overcome a challenge like a game designer can. I work within tight constraints and am expected to deliver something much more specific than what a game designer is usually asked to deliver.
Are there any automa you’d design differently after seeing their reception among gamers?
One of the elements of the Viticulture Automa for the Tuscany expansion got some flack for being too static and lopsided in the Automa’s favor. I would definitely consider changing that subsystem if we were to make a new edition, I’m not quite sure how I would change it, though. It’s the handling of the influence map, I’m talking about here, in case you know the game.
For Scythe, I’d love to design an expansion where the Automa plays more differently when it changes faction because that’s something some players have missed. I think it was the right decision not to do this in the core game because the Automa is fairly complex as it is, but more asymmetry would have worked well in an expansion.
On a non-design level, our original rulebook for the Tapestry Automa was criticized by some for being too succinct and while I believe that the rulebook covers everything, I can see where the criticism is coming from, in particular I can see how it would be challenging for people who learn best from reading examples because it only has two small examples. So, for the game’s last expansion, Fantasies & Futures, we made a new and expanded rulebook that included more examples and addressed other questions we had gotten over the years. It also integrated the core Automa and all expansion rulebooks into one book. It’s available as a free download now.
Automa are probably not suited to every type of game; can you say where they work well and where they’re impossible?
They work horribly for games that rely on humans as social animals. While I have made an Automa for The Mind just for the sheer silliness of the idea, the end result unsurprisingly sucked. Games like Werewolf, Charades, and Telestrations are some obvious examples where an Automa won’t cut it.
Generally, Automas work best in low interaction games because it’s typically the simulation of interactions that adds complexity. E.g. a worker placement mechanism can work well with an Automa system and you can often get away with the Automa just semi-randomly placing workers.
If there’s a need for one player to deeply analyze what the other player is doing, then you can get into trouble, because it’s hard to make a stack of cardboard intelligently analyze what the human player wants and needs. This for example means that a game that relies on hate drafting poses a challenge.
Hidden information can be both a blessing and a curse:
- It’s a blessing because it means that it’s OK that the bot doesn’t analyze the player’s situation much, because the hidden information also makes it impossible for humans to analyze their opponents.
- It’s a curse because it’s hard to make the bot act based on the hidden stuff it has, given that the player would need to know what the bot has in order to carry out the procedure for running it. This was a major roadblock in the development of our solo mode for the trick taking game Origin Story.
The breadth of the decision space in the interactions and the level of intelligence required for them are major factors. For our Scythe and Terra Mystica solo modes the jostling for position on a large board were tough nuts to crack.
Chess is an example of a game that would be completely impossible for these reasons.
Ironically, the interaction can get so low that it’s hard to make good Automa solo modes.
When we were making a solo mode for the game Rolling Realms, which doesn’t have a single point of interaction until the comparison of points at the end, we actually got into trouble.
Had we made an Automa for that game it would just be a complex way to generate a semi-random number of victory points for the player to compete against. So, we had to take another route.
Real-time games are also challenging if you’re trying to create a bot-based solo mode because by a solo modes very nature the only one present to run the bot is the human player and they have their hands full playing their own game while clock is ticking.
We had to overcome that for Pendulum from Stonemaier Games, which contains real time segments. We handled this by having the two Automas that are in that game act as a singular player during the real-time segments. A single that just blocks action spaces using extremely simple rules that occasionally require a couple of seconds from the player, while in the non-real time part they’re treated as separate entities that score points and engage with the game.
You mentioned using two bots; generalising that, what are your views on automas being deployed beyond solo games and multiple in use simultaneously?
I’ve already mentioned Pendulum above and I’ll add Euphoria where we made a system that added two Automas instead of one because that game is generally regarded as working better at 3+ players than at 2. For that reason, we also made it so that the Automa could be included in a game with 2 humans to increase the player count to 4.
In Euphoria it worked because the Automa there is pretty smooth, so you can handle two easily.
Pendulum had 1.5 Automas so to speak because there were two but they merged into 1 during the real-time segments. In Tapestry we made 1.5 Automas in another way. A full one and a heavily slimmed down one that followed a subset of the full Automa rules to increase the tension and strategy in “tech track” part of the game.
That slimmed down extra bot can also be used in a game with 2 human players to make it feel more like 3-players.
Typically, though, we focus on just the solo experience because having to make an Automa that works at multiple player counts can lead to trade-offs that make the solo mode less than it could be. Also, developing for multiple player counts dilutes playtesting resources and time we’re able to spend on making solo the best it can be, which is what our mission is.
Another way I get my solo game fix is playing a phone or computer port of a game; these typically offer ‘AI’ opponents—how do you think they compare to automas? (I think it’s interesting that the digital Wingspan includes both your automa and AI, each with different strengths and weaknesses.)
They’re often rather different experiences because such AIs typically play by the same rules as the human player, while an Automa typically plays a different and simplified game that intersects with the game played by the human player at the points of interaction. Making such an AI play at a level that challenges a human player is very hard, though, beyond the resources of many game developers, so implementing an Automa can make sure that the game can present a challenge for even an expert player.
Can you say anything about what 2026 is going to be like for you and the Automa Factory?
Stonemaier will be releasing a couple games that feature our solo modes. The rerelease of Namiji is unfortunately the only one I’m allowed to name. Multiple expansions will also be released: Wingspan, Expeditions, and Finspan.
What’s next for you personally—are there more game or automa ideas waiting to burst out of your head? Can you give us any hints?
After Skoventyr was finished, I promised myself not to pitch a new game to a publisher for some years because I wanted to reduce my workload. I still wanted to dabble in game design, though, because it’s fun and to hone my skillset.
So, I designed a game just for myself with no intention of doing anything with it beyond the first one or two prototype iterations. It turned out that it kinda just worked from the beginning and I had so much fun with it that I couldn’t help myself but show it to some friends. They also liked it and so I toyed around with it some more.
Then I asked a publisher that seemed like a good fit for the game whether I could pitch it to them at Essen ’24. They told me to just send the files and so I did. Fairly soon thereafter I got a mail from them saying they wanted to publish it.
So, now it’s in small scale external playtesting and we’re trying to get ready to go into full-scale testing soon.
Unfortunately, I’m not allowed to say anything about the game itself and it’ll be a while yet before it’s published. Even more unfortunately, this story shows that I’m really bad at keeping promises to myself.
Additionally, there are always more Automas on the way but none I’m allowed to talk about.
What else can you tell us about Morten Monrad Pedersen, the person beyond the game industry? What do you do in your spare time, if such a thing exists?
I sporadically write a blog whose name indicates how much spare time I have. It’s called Thematic Solitaires for the Spare Time Challenged. In the past year or so, I’ve had more spare time, though, as I’ve gotten my workload somewhat under control and my son is now old enough that parenting takes less time and I even have a gaming partner in him.
The spare time that’s not spent with my wife and son is often spent on solo board games or going for walks. Going out to dinner with friends or to the cinema to see the latest sci-fi movies is also something I like to spend my time on.
While I have zero musical talent, music is very important to me and I listen to music on a daily basis. A lot of my design work is done with music in my headphones.
Finally, I’m currently getting into the habit of meditating and practicing mindfulness every day. So far it has been a great help to me on a personal level.
Where can people find out more about you and what’s on your mind?
The two best ways to do that are to subscribe to my blog, Thematic Solitaires for the Spare Time Challenged and to the Automa Factory newsletter.
Thank you for your time, Morten.
Thank you for being kind enough to provide me with a podium from which to yammer.
