This ancient game, which was established in 2002 and released for free in 2006, has directly influenced a series of masterpieces including "Minecraft", "Ring World" and "The Great Immortal Cultivation Simulator" with its advanced concepts that transcend the times, and indirectly spawned a market plate of tens of billions of dollars.
Tarn Adams is one of the two co-creators of Dwarf Fortress and has been the game's sole programmer until now. At the end of '22, he and his brother released a paid version of Dwarven Fortress on Steam (alongside the free version) in order to raise money for treatment, adding basic graphics and UI to the text-only original.
The game was a huge success upon its release. To date, VGI estimates that it has sold more than 700,000 copies and generated more than $17 million in revenue. The most terrifying thing is that such an extremely rough art game has an average player online time of 82.2 hours, and the median online time is 41.8 hours.
GameDeveloper interviewed Adams earlier this year to discuss topics such as balance, product ideas, and how to stay interested in the long term.
For all solo teams who are interested in deep cultivation in simulation management, strategy, survival, procedural generation and other tracks, Adams, as a founding figure in this field, has great value and is worth learning from Xi.
The following is a translation, some of which are abridged:
Dwarven Fortress has a very long history. It's been over 16 years since the game was first released to the public on August 8, 2006. To be able to continue to improve after such a long time is an accomplishment in itself. How did you maintain your interest in the project after all this time?
It's easy to stay interested because there's so much to do. We're just trying to be as comfortable as we come along, to face what happened naturally, and to try to get where we want to go. Basically, it's the goal of the mythical and magical version (laughs).
• What was the experience of working with publishers and publishing through Steam?Did you like the new system?Does it add to the workload?
Everything is ok and Kitfox (the publisher) is fantastic. Now I have Discord to manage and version control, so the process does get a little complicated. And we haven't tried to do a real feature curve like we used to, because the Steam version is still quite new.
I remember that Dwarven Fortress was originally a 2D game, with only one "Z layer", and digging mountains meant digging tunnels to the right side of the map. That's when Dwarf Fortress was first popular with players. How did you transition from that playstyle to your current 3D environment?
It was probably in 2008-2009 that I remember that the core problem was liquids. When the player builds bridges over lava, water and connects them with levers, none of this can be rationalized without going through some very complex logic and data structures. So we completely switched to 3D slicing.
This poses a lot of problems, such as the tightness of the game flow. We wanted to be able to tie all of our procedural content together with no dead zones, so that people could flow from one story to another as they explored.
But it's harder because basically you can't find a point or a line in three-dimensional space, everything is a slice or a chunk. You need to have enough variety to mix up a lot of little things. Doing so would require a lot of rewriting the map code.
This is the next big task, and we're looking at restoring or iterating on our original 3D content so that it really works, and can be properly expanded and modularized, unlike the current cave systems and underground sites and features.
This is also an obstacle to our vision of mythological content, as the locations involved are not supported in the current system. We are still thinking of ways to get around this obstacle. If versioning works, we can also create a branch while we continue to patch and add features, but merge conflicts can be quite serious and must be handled with care.
Many popular games are designed to be more geared towards hardcore players. This makes it more exciting for experienced players, but it also hinders new players from entering the game. What do you think?
We try not to do that, which is already a very deep philosophy in game design. "Don't design games for experienced players!".
Some SLGs, fighting games, sports games, RPGs often fall into this situation and can barely attract new players after being online for a while. If you don't want to attract too many new players and prefer to be like a club, then this probably doesn't matter. But we've always wanted more people to tell stories with our games.
Now you are maintaining two codebases, one is the paid version of Steam and the other is the older version. Who is responsible for the design of the new UI?Are there any challenges in separating the code?Are you worried about others working on your previous code?
The new UI was mostly done by Zach and me, but there were also other people involved, as menus were an ongoing topic of discussion in our early days when we were working on art design.
As the work progressed, others also contributed. Later we also hired a UX expert to help us out, which is why we have a beautiful building menu and main interface.
There really aren't two completely different codebases. Switching between the classic version and the graphical version only requires exchanging a few characters. This comes with some limitations, but it makes it very easy to maintain the free version.
This year we introduced a new programmer, Putnam, and so far there have been no problems.
I've always had a question about Dwarf Fortress. The game spends some time generating the prehistory of the world before it starts, which serves as the basis for the player's own dwarven worldview. What I want to know is, once the spawn is over and the player starts exploring and building a settlement, is the spawn process over, and if not, can the player influence the rest of the process?
Around 2014, we released a version called "World Activation".
This is the first time we've put world-generated stuff into gameplay. However, there are a lot of incompleteness. For example, there will be sports matches in world generation, but you won't see them in the game. In addition, there will be more details after the world generation is complete, such as armies and invasions, so it will take a lot of work to coordinate these things.
Players can basically influence everything in it. The game keeps a record of what they do in Legends mode, just like they do in world generation. The data structure of all historical events is the same across the various patterns, and some patterns may have more detail than others.
World generation is based on a seed, you can generate a world for 110 years, generate the same world for 100 years, and then play on your own for the next 10 years.
Legendary mode lets you compare what's going on, including random people or locations, and more. In a 110-year world without player influence, there may be someone who has lived their life peacefully, but in a world under player influence, that person may die in 105 years in a saw trap you placed.
· The influence of DND (Dungeons & Dragons) on RPGs is deep-rooted, such as the use of HP as a universal measure of biological health. I remember you saying that you don't like the concept of HP. Dwarven Fortress ditches HP in favor of injuries to each body part of a unit. In battle, it is also possible for powerful monsters to be killed by a single blow to the head. How does this "simulated reality" mentality affect game development?
Yes. The core of this is that you have to follow this set of mechanics everywhere in the game, or you will quickly get story conflicts, such as getting two skulls for a creature or something. In addition, we are able to use the body parts system in many places. Slaughter, toxins, syndromes and the like have become more interesting thanks to this system.
At the same time, the hybrid system is interesting - even a simple "HP+ wound system" can cover things like the "Lucky Hit" you just mentioned, without involving other complicated mechanical changes. There's a lot more to do in terms of the damage system (improving MP with the spell system, etc.).
Recently, there has been a lot of buzz on the Internet about new advances in AI content generation, which allow people to feed a large number of works into machine Xi algorithms, and then generate new works based on those works. Dwarven Fortress uses a traditional procedural approach to content generation, do you think this new system has a future in similar games?
I have reservations about that. I feel like I always need to expand on what I'm working on, and moving to this AI black box makes things more difficult, so I don't use AI. Although you can think of our world generation as a kind of AI in some way.
Everything depends on how you want to use and understand what you're making and what you want to do with it later.
Dwarven Fortress has been balancing gameplay with simulated reality. Which side do you think is more important? Are you worried about the development getting out of control due to the complexity of the system?
I don't feel like I have to make a choice. An innovative mechanism needs to do both. Players need to be engaged, interested, and exploratory, and the simulation needs to give back in a coherent and unexpected way.
Players should have a reasonable expectation of the impact of their actions. For example, in terms of myth release, if we were going to have the world destroyed by a meteorite at 250 years, we needed to make the process interesting, not just the sudden destruction of the whole world.
Similarly, if we have a simulator that generates random earthquakes and tsunamis, it may seem a bit excessive to destroy a coastal fort completely randomly. But at the same time, it is up to the player to decide - the fun of "disaster" depends entirely on frequency and means of prevention.
I don't think there's a set answer to that.
Dwarven Fortress is known for its fun world-generation. How do you translate the abstract rules of world generation into game events that players can interact with and even influence?
This process takes place on several different levels.
At the most basic level, world-generation leads to many situations, such as the proliferation of locations, wars, and diplomacy. The player is initially faced with the original outcome of these situations, and as the game continues, sometimes the mechanics are still working, sometimes not, depending on the situation.
This is followed by all the historical figures, including dwarves, goblins, or monsters who arrived as immigrants, as well as a large group of tourists and others. Worldgen builds on these characters in an ongoing way, as well as the relationships between them - at any point in time when the player might stop worldbuilding, existing historical figures are fulfilling their roles, and there is a realistic historical record.
So, starting with civilizations and places, to historical figures, there are many things that can be done. Specific events led to specific diplomatic relations. Roughly speaking, you can think of it as a zero-player SLG with no player intervention and only AI opponents to operate, and then cut in and let the player control one side in the middle of the game.
How do you balance the relationship between the different civilizations in the game world? Do you try to deal with civilizations that get stronger over time and eventually take over the entire world, or how do you keep the game full of interesting conflicts?
We try not to mess with it, because when the mechanism changes, those interventions will decay or overdose.
The best-case scenario: things happen because of the underlying mechanism. In places where the mechanics are incomplete, like Stronghold Siege, we'll still be reinforcing some things in a "gamified" way, but we're trying to minimize that behavior.
For a civilization that is too strong, we have added mechanisms such as alliances to slow down its development. We plan to add more mechanics, such as task forces to hunt them down even in peacetime, to better control them.
If their power affects the balance too much and must be weakened, that's fine. But we prefer to keep one side strong, just so that other civilizations can react more strongly to enliven the whole game. This is a dynamic hedging process.
- In the game, dwarves use experience and equipment to improve their abilities. Experience is like a reflection of a character's personal history, and equipment is the strength of the fortress itself. Do you think this distinction is useful, and which is more important to the player?
Both are important.
We want players to be both good at using their resources and treating their soldiers well, giving them time to train, and keeping them happy and healthy. We wanted the extremes to look extreme - the game's Adamantite weapons and legendary skills stand out.
In general, when it comes to story and world-building, we want people to be heroes, which means that skills can overcome some of the equipment and quantity deficiencies, although it may not be as exaggerated as Dynasty Warriors.
But once we have more options and more flexibility in terms of mythology/magic, I think balance issues might arise as well. At that point, we should direct the player to a more "realistic" world, and the importance of individual skills will decrease.
Rectangular grids are an old but reliable way to build game worlds. It's simple, efficient, and easy to map each tile to a character in a text display. Are there any shortcomings in your experience?
I don't think it's convenient to move diagonally. Hexagons are undoubtedly superior in this regard, but of course, hexagons are inferior to rectangles in other respects. Any mesh has a lot of drawbacks over representations with higher resolutions, such as pathfinding, how many creatures can surround another one, visual representations, the underlying complexity of the map, and so on.
But grids are easy for computers to handle, which gives me a lot of flexibility in content creation.
Can you talk about future development plans?
We will continue to develop. It's important to keep the game playable, and that's obviously something we've been focusing on for the past few years.
We've just rejoined the arena, so the next step is to make some optimizations and then update Adventure Mode to the next version while completing the Linux/Mac port.
In addition, we will continue to restore classic features such as keyboard support. Then, we'll set to work on refining the villain system, improving armies/sieges, and continuing the work of mythology/magic.