AI War:Design Complexity

Why The Complex Interrelation Of Ship Damage Mechanisms?

Q: The game has an enormous number of relationships between all of the various ships, which is impossible for anyone to remember off the top of their head. Class X is stronger than Class Y, fine -- but with dozens of ship classes, this makes for tens of thousands of combinations. Wouldn't it make more sense to have overall "armor types" and "shot types" that then people could actually remember?

Response from Chris Park, AI War's Lead Designer:

Armor Types and Shot Types

First of all, the game does model both armor and shot types, after a fashion. However, they are 1:1 correlated with the actual ships in the game, rather than being an overall armor class that is used by many ships.

Take the following real-world example: a Chevy Malibu and a Porshe 911 and a Hummer If you were to fire bullets of varying sorts (differing calibers, even amongst "shell" type ammo) into those various vehicles, you would get different damage characteristics to all of them. For example, placement of engines, size of passenger compartments, material and amount and position of armor, plus other internal safety features, all make for an extremely complex real-world situation. And, for that matter, as changes are made to each line of car with each calendar year, you'd see different performance characteristics per car per year in order to make all of those realistically modeled.

So, what AI War is doing for the sake of sanity and CPU efficiency, is generally modeling the interactions of all of the various shot types from various craft against the various hull and construction characteristics of other craft. It can be assumed that the caliber and other factors is a bit different even between "shell" or "laser" shot types that are otherwise the same. So that's what is being modeled -- it's very different from a "job class" for human avatars in a game like TF2, because those refer to skills and there is very little difference in shooting a big skillful guy or a small medic guy with a sniper rifle -- they both die pretty much the same. Giant spacecraft and a variety of weaponry are a whole different animal.

But Why So Much Complexity?

Okay, so that's what is being simulated -- but why? To a mathematician, the design of the mix of ships in AI War is not mathematically elegant, and so a few particularly math-oriented players have tried to get me to make it more simple and mathematically expressible in order to make it more pleasing to those sensibilities. I don't consider myself a math person, but I can sympathize with that viewpoint. I have very good reasons for doing otherwise, though:

1. I keep the game balance in a semblance of disorder and a least slight-imbalance on purpose, and always have. If there are major exploits, of course I fix those, but those are comparably rare. But when I introduce a new ship, I intentionally seed it with values that will put it just a bit off kilter with other similar ships, and against the expected foes. Usually this is via special abilities, but it's also in terms of costs, attack powers and attack bonuses, shielding, and health. I like for all of the ships, even amongst fighter-type ships for example, to be slightly unique in subtle ways in addition to the major ways. In a second, I'll get to why.

2. In some ways, I guess I am a math-oriented person, despite my claims otherwise. It comes from the upbringing. When I play boardgames such as Descent: Journeys in the Dark, I'm always the rulekeeper/game master, and I'm the one that keeps up with everything and even helps my "foes" reason through all their possible attacks and scenarios, and I'm extremely, extremely quick at coming up with the optimal solution. This is also important. I'm also very, very good at doing this against AIs in RTS games, and with the mechanics in general with RTS games. When I start playing a new strategy game, like many people who enjoy AI War I'm sure, I'm learning as much as I can about it and crunching the numbers mentally and collating and comparing in-game data against my hypotheses in order to find my personal optimal strategy. And when I eventually do? I've lost all interest in the game.

3. What most strategy games have in common (the good ones) is that they are designed by math-oriented folks. Those designers use all sorts of techniques to make sure that the balance is as pristine as possible, they try to keep things simple while also being complex, and generally they are huge proponents doing automated/adaptive unit balancing, etc. It's a good approach, generally speaking. Certainly a must for a competitive pvp game. But all those games bore me to tears after 6-12 months of biweekly play. I'm tired of being bored, and I certainly don't want to be bored by my own game, so this goes back to why I designed AI War the way that I did, in a lot of senses. Escaping this eventual fate with AI War is my core motivation, and why the idea of making things simpler or more orderly strikes me as hugely undesirable.

4. Therefore, I keep things muddy with the ships. There are a lot of them, they are complex, they have attributes that interact in varying funky ways, and in general there is too much for even me to remember despite the fact that I spend so much time with the game. I can't understand it all in one go. I haven't been able to find a best strategy in the 13 months I've been working with the game (as of the time of this post). That's victory for me! I have kept myself entertained for 13 months, and I suspect I will keep myself interested for several more years if not far longer. If I'd made this simpler, or more orderly, I'd already be done with it and bored.

AI_War:Fleet_Command

Why Does The Game Force Variety On The Player?

Q: In many aspects of the game, I am forced to try something experimental rather than being able to craft exactly the scenario I want. For one example, why can't I choose my starting home planet and my starting bonus ship type separately? It is hard to find a map with both exactly the starting unit and starting position that I want.

Response from Chris Park, AI War's Lead Designer:

You'll notice that everything about this game centers around variety:

• It forces you to choose either ships or starting positions you may not prefer in the lobby
• The AI types have some pretty powerful and game-altering abilities in many cases.
• No two games have exactly the same ship mix
• The expansion is geared around adding as much variety to an already-huge game as possible.
• The maps are randomly-generated only, and with the added styles more recently they have all been focused on creating new game modes, rather than just looking cosmetically different.
• And finally, wherever possible I go out of my way to make the complexity multiplicative, in terms of letting players have as many minor factions as they want in games, and more capturables, and so forth and so on.

There are many other possible examples, but those are perhaps the most central (going along with the explanation of the complex interrelation of ship damage mechanisms, above). All of those overlapping, crisscrossing, muddy systems create an environment that feels increasingly real to me from a strategic standpoint, and increasingly less meta-gamey. I can't remember all the attributes and stats and strengths and weaknesses of all the ships, and so every time I start a new campaign I am presented with practically a fresh game. Maybe autocannons will be great; maybe they will be not even worth building. It all depends on the relative ship mixes, and I'll have to decide in that game. Some ships might not have any use in one game, or might be marginalized, and in others they might hold one key role or a different key role. If you always play your favorite ships, you start falling into a rut, which is why the game so ardently encourages experimentation at every turn.

Overall, I'm trying to emphasize exploration and The Unknown in the galaxies, and to a certain amount that means letting the procedural generator do its thing, rather than letting the players script out every last aspect at the start. If players script out everything to just the perfect way that they want, then that's the only way they'll play and they'll get bored of it quickly -- same as finding optimal build paths in other RTS games. In Chess, the only thing keeping every game from being the same as the last is your opponent: your opponent does something different, and that changes the strategic landscape, so you must do something different as well. In RTS games, which have a lot of design challenges not present in turn-based games, I posit that the scenario must vary. Even online pvp RTS games tend to get highly predictable compared to, for example, Chess. The only way I've been able to think of to combat that is by having a wider variety of potential game states and scenarios for players, so that they don't fall into ruts or learned patterns, but rather must instead evaluate every situation fresh.

To put it another way, it's to save the math-oriented people from themselves. If they are anything like me, they want very badly to understand everything around them, and in general they do a great job of it -- but especially when it comes to games, that is the game. They come, they experiment, they understand... they get bored and leave. AI War, for purely selfish reasons, is built around an entirely different premise that I can probably claim no other RTS has ever tried before: it's built to keep myself and people like me entertained for really, really long stretches of time.

Followup Regarding The Randomized Seeding Of Starting Ship Types, Specifically

Q: I miss this feature, we are wasting a lot of time clicking through maps in order to get the ships we want. I never play a casual game of AI War. This game is a real challenge. I am playing against AI level 7, and I win perhaps one out of every ten games. Given this sort of opposition I always have a plan, and this includes knowing in advance what kind of bonus ship I shall try out this time around. Your argument is that if I use the same ship type config every time, the game will be boring in the long run. This is only partially true. My opening moves will tend to be the same, but when I find an advanced space lab, randomness sets in, and in most games this means a change of plans. If the day arrives, where my opening config is so strong that I am sure to win the game no matter what other ships I acquire during the game, then your argument holds. But I am afraid that day never comes. And if it should come, I can always go to AI level 8. I respect that the final decision is yours. But you changed your mind on other topics, so maybe we can make you change your mind on this one, too?

Response from Chris Park, AI War's Lead Designer:

I've changed my mind on a variety of topics, for a variety of reasons. I try to be as open-minded as possible. But, this is not one that I'm currently feeling at all inclined to change on. From one angle, really, how much time are you wasting clicking through maps to find ones that you want? A minute or two per 8-12 hour game? That's not something I am really that worried about -- I think of it rather like re-rolling for your character at the start of a great many RPGs. Yes, you can keep going until you get almost exactly what you want. But there's a huge pressure to simply settle for something that is sort of what you want, and thus try something new.

In RPGs, that mechanic tends to be based around a pure gamble -- if you find something almost right, and then click re-roll again, then you've totally lost the one possible scenario and might well not find anything remotely as good for a long time. Of course, some RPGs compensate with a certain number of skill points that you can then allocate as you wish to buff your character further. With AI War, you can copy your favorite seed so far to the clipboard, and then go back to it if you are frustrated (or copy a new one to the clipboard when you find something better).

I am no stranger to re-rolling map seeds in order to find a "better" map than the current one. Sometimes I just don't like a layout, or it doesn't have any ships that I am that interested in playing at the moment. But what I end up settling on in the end is something a bit random, with positive and negative aspects, which I view as highly desirable. With multiplayer, this might be compounded with certain players really wanting a certain ship. If it's only two players, then you can probably find a map with a what both of them want fairly quickly (in my experience), although that might then create geographic constraints. Then the remaining players (I tend to play 4-player) are left to pick from amongst the remaining ships, which introduces yet more variety.

I firmly believe that this process leads to highly beneficial results, even if it does result in a minor bit of aggravation to some (or even a fair number of) players some of the time. This is something I've thought a lot about, and have had ample chance to witness in various forms in all my years of multiplayer gaming, and I feel it's important. If you look at a lot of modern FPS games (especially team-based ones), the same general theory is in play in a great number of them: weapon/vehicle drops are often random, and class slots are often random, so in a given game it might be rare for you to get exactly everything that you want, just the way you want. So instead you have to experiment, and learn to function as part of your team, rather than just always going with your ol' standby.

You can claim that this reduces the strategy of the game, but there again I disagree: real generals don't get to set every last parameter for the battles they will go into, and they don't even really get to choose all the force they will have (if you have no more tanks, you have no more tanks). So a big part of their strategy, at the start especially, is figuring out how the heck to deal with the hand they have been given. There are a huge, huge number of very useful ships in the game, such that there should be multiple top-tier ships in almost any given map seed. Thus I really don't feel that this affects the difficulty of the game overmuch, it simply makes players use more than a single overall strategy (a primary goal for the game's design). Plans never survive contact with the enemy, right? Consider the map generator your first contact with the enemy -- as would be the case for a general in any real conflict, whose first order of business is finding out what the general terrain looks like and thus where the enemy can be presumed to be, seeing the possible deployment points given that data, and seeing what sort of forces he/she has been given from higher command.

AI_War:Fleet_Command