In part 1 I introduced how I write a design brief for solo modes and provided a couple of quick examples of how that brief shaped the process. I also laid out the start of thinking about 3 types of solo mode. Remember: These are not distinct brackets but rough types that can easily overlap with one another. In this part we’ll look into those types in more detail, but first…
Co-op Games
I’m not keen on co-op games in general. I’m even less keen on them as a solo experience. This is my personal preference. I don’t like getting picked on by games themselves and I don’t like having to run multiple characters with a shared goal. Co-op games are already designed with a lot of solo principles in place. The player/s must beat something to win. They can also lose in certain situations. There are restrictions in place that make it harder for the player to win. Co-op games lend themselves to being played solo simply by 1 player taking on the game with 1 or more characters that would be used by separate players in a multiplayer game. Very little has to be done to make these work and if you like the challenge of co-op games then if its a good co-op game it’ll probably be a good solo game. These fall firmly into the challenge game type and provide us lots of useful information and tips to apply to competitive solo games.
Competitive games
These are the solo games I enjoy the most. That does not mean they need to have an artificial opponent to beat. A Feast for Odin would be considered by many to be a beat your score solo game. It is. However the simple and ingenious twist of using 2 sets of workers on alternating rounds makes it a challenge, as we have defined it. A Feast for Odin is an excellent solo game, Caverna on the other hand is only ok, the difference is the challenge of that alternating worker system, it takes a small element of an AI by having certain spaces blocked each round but it was you that went and blocked them! A lot of competitive games can make great solo games but some don’t. If they are poor 2 player games that just don’t work for some reason or are very spatial then an AI style solo mode will be very hard work, not just for the design team but also for the player. The movement rules in my Concordia solo mode are the biggest barrier and the hardest thing to work out for the player. Spatial awareness requires intelligence and assessing a particular situation. Without an enormous flow chart system for decision making this is hard to replicate and only so many solo gamers are prepared to invest in that effort to run an AI. Don’t write off solo modes because they don’t have an AI system, there are some really smart designs that use puzzles and challenges to great effect. By writing a brief for how a competitive game can be experienced by 1 player we can find new ways to make a solo mode.
Puzzles : A solo gameplay experience that has a single solution, you either win or lose
A lot of story based games are predominantly puzzles, Gloomhaven for example. I only play the solo scenarios solo, as explained above, and there aren’t many different ways to win those, they are very much puzzles to solve and once they are solved you move on. So to make a puzzle based solo mode you need one of 2 things; a lot of different puzzles or a modular puzzle with variable elements. Gloomhaven is the former, Mage Knight is the latter. Both have their own strengths and merits. A variable puzzle framework often needs an element of randomness, but that randomness will need to be very carefully tested. Chocolate Factory has this with the combinations of weekly targets and daily demands. Every iteration of that design I tested 12 times, twice with each weekly target, using all of the daily demands with each target. That’s not every combination but it was enough to make sure nothing was too hard or too easy (mainly too hard, it’s a tough game anyways!). Puzzles are win/loss, you solve it or you don’t. A sense of threat about losing is important, time is a great element to add here as a restriction, you can also use an outside force, which has elements of an opponent or twists and turns during the game taking things from challenges. A puzzle type game gives a solo player a great experience and is perfect for story telling over a period of time with either ever changing puzzles or a huge number of fixed puzzles joined together.
Challenge : A set of restrictions or a framework that sets a target for the player. These have a sliding scale of success
The infamous beat your score systems fall within this category but there’s so much more you can do here. By taking elements from the other types, a challenge solo mode can be as rewarding as any other. There’s a lot here that can also be included in the other two types. Ultimately giving the player something to beat is a key linchpin of solo design. Story elements can again be used to give the challenges progression and randomness can be used to make them variable. The key for me here is only giving the players a framework to play within and giving them a much bigger scope for how to do it than if it were a puzzle. Mage Knight does this through the character progression and deck building. A Feast for Odin through the worker placement system. Gaia Project through all the different races. A challenge framework should be thematic. Challenges are not win/loss like puzzles. There might be scores involved, with ranks applied to certain scores or perhaps there’s consequences for how well you did. Find ways to inject the threat of loss by controlling outside elements that affect the player during the game. Obviously anything passing from game to game needs tying into legacy or story telling within a game but there is so much design space to explore here, it is very much more than beat your score.
Opponents : An artificial opponent that replicated a human player, or an outside force that the player competes against
This is perhaps the best known type of solo mode and it has exploded in recent years. David Turzci and Morten Monrad Pedersen are the most known in this category but it is no coincidence that Nick Shaw works alongside both and is perhaps the hidden wizard behind the curtain in many of their designs. I’ve been very lucky to start working with David and Nick recently on a few projects and I’m continuing to learn a lot from them. I am sure David has a number of blogs you can read on the subject. Much of this section I have learnt from the aforementioned gentlemen.
An opponent type solo game sets out to do something very simple in concept; give you an artificial opponent to play against. This should be like playing against another person. Designing that opponent is a lot harder than you might think. It needs to be competitive and easy to run. These two aims will often oppose one another. Go back to your design brief and your theme, remember these are the biggest things you have to help you make decisions. Here are a few keys points I use:
Interaction – Where do players interact and how can this be replicated? This depends very much on the game itself. Play it 2 players and write good notes of the interaction points. Getting these right is the key to success.
What can be ignored – To cut down on admin the opponent should ignore anything fiddly, make it as simple as possible. In Eternal Palace for example the AI never gets or spends any resources, this cuts down loads of moving components and makes the decision tree much easier.
Randomness vs cheating – This is a key balance. To replicate another player the AI needs to do stuff you don’t expect. This requires randomness. You could just roll a dice. I’ve already mentioned I don’t like dice driven AIs. If it fits within the game and can be used really smartly, I’m all for this (David and Nick do this particularly well). I prefer a card draw based AI as it enables a balance to the randomness. But if a player just acted randomly all the time they’d lose, as we are dealing with games that require some strategy here. So the bot needs to cheat to make up for acting randomly. This can be seen in my Concordia variant where Automus earns maximum money, regardless of when he draws his Senator card.
A player shouldn’t make decisions for the AI but can influence them – Golden rule that works in the majority of cases. This ties into the interaction. If a player has too much control it’s akin to playing the game yourself which is not what we are after. A push/pull here can be a really interesting thing to explore, if I do that then the bot gets that, which is more important to me? Villagers has you drafting cards for the Countess but there’s enough other stuff going on to make that work. This goes hand in hand with making it simple. We will often create a decision tree, this can be complex like trading in Concordia or simple like the 3 tier cards you’ll see in Scrumpy. Basically it boils down to; can it do this? Yes – do that, no – do something else. Balancing and polishing these systems can take some time but it will be the core of any system.
Do players get better? If so, how does the opponent get better? This is my favourite thing to explore. How does a bot improve as the game goes on? I used Concordia as inspiration for Scrumpy as each time the AI deck cycles a new advanced card is added, over the course of the game the combination of these advanced cards sees the bot player evolve to have a particular strategy.
Can the opponent have a strategy – I’ve just spoken about strategy that evolves during a game but this is an element of artificial strategy and is something I handled differently in Reavers of Midgard where it is defined from the very beginning of the game, by going against one of three distinct bot players. I’d love to build an AI system for Lords of Waterdeep one day where each lord plays differently. Not got round to that yet. You’ll need a good grasp of the game to work this one out. My playtesting methods, from another blog, may well help.
More than one AI – Some games don’t play well at 2 players so you might already need some sort of opponent in a 2-player game, that means for solo you’ll need to add a second AI as well! This is a toughie as many players will resist running 2 AIs or using dummy players in a 2 player game. It may be that the game is just not suited to low player counts. It can be done however, in a recent development project I had a game with a dummy player mechanism at 2 players and a full AI. I rebalanced the game so that it appeared these two forces were working together and streamlined some of the admin. Simple changes, but with a good thematic tie in it reduces the barrier for players. Concordia Venus was a different beast as I had to adapt the Automus AI to act as a second AI as part of a team. Even David Turczi reckons a bot as a teammate to a player won’t work but building a subtly different version of Automus to act as his teammate was a good challenge.