Game Design Games

Playing Pieces, or Free Agents? A Question of Unit Agency in RTS games

in real time strategy games, units tend to be, well, pretty stupid. They stand around doing nothing, unless ordered by the player. They have no motivations, no drive – shoot, in many cases you’ll find yourself lucky if they shoot back when they’re attacked by the enemy. They act more like chessmen than proud space marines, Mongol warriors or vicious insectoid aliens. Heroes, be they supposedly suave secret agents, supernatural demigods, or hotheaded paladin, are all likewise bereft of actual character and are able to be left, forgotten, in seldom frequented areas of the game map to twiddle their thumbs while their armies do all the work.

Where are the heroes, the free agents? Where are the generals, the mavericks, the characters? Where is the feeling that units are, well, people with a will and mind of their own?

Let’s talk about that, shall we?

Dune-2-androidIn the Beginning…

In the beginning, there was Dune.

Now, of course, Dune 2 was not born in a vacuum – it was the result of lessons learned from other games both physical and digital, but it gave us perhaps the first truly modern RTS experience and so provides a convenient starting place, as well as an opportunity to make a Bible joke.

Early RTS games were necessarily constrained. They were trying to bring to players an immensely complex experience, even a connected set of complexly interconnected experiences, and processing power being what it was, developers were hard-pressed to execute on their visions in a meaningful way. From computer graphics to game intelligence to the number of units a game could support on the screen or within the context of a match, these games were shaped, forged like diamonds by the technological pressures of the day.

Necessarily, unit behavior was limited, and has improved over time, as processing power and AI design methods have improved to allow more autonomous behavior to be built into unit designs. No longer would tanks or infantry stand around and mindlessly allow enemy tanks or infantry to wail on them unopposed. Processing power and consumer demand met in ingrained unit behavior that improved the play experience in RTS for pretty much everyone.

And yet, now, we still see most RTS games limit the artificial responsiveness and ‘intelligence’ of their units. Units in most RTS games are treated like game pieces, not like, well, people. Why is that?


Virtual Board Game

It all comes down, of course, to the designer’s philosophy behind creating the game. Or the player’s expectations when playing the game. Or, ah, a conflation of these 2 admittedly general categories. To that end, I’d like to look into 2 of what I see as the most common types of strategy game and their respective approaches to unit design, specifically as regards unit behavior and autonomy.

First, and I’m sure there’s exists a better term for this, are what I think of as virtualized board games.

A fairly common assumption behind the roots of the strategy genre is that RTS games evolved from board games or, more specifically, tabletop war games. In these games, the player is necessarily a large portion of or entirely in control of the movements of their pieces. Rules such as terrain or morale could alter piece (unit) behavior, but ultimately players are in control of when, where and why their pieces act.

This has a number of implications. First and foremost, in a real-time strategy game it has become an expectation that a player be able to immediately determine the outcome of a given action. This is why, for instance, RTS games tend to seldom use damage ranges when creating unit damage tables. A unit moves so fast, does so much damage, and behaves in precisely such a fashion, based on circumstances. In a large part, units being ‘playing pieces’ is a consequence of players and game designers intent to provide complete knowledge of game context – clear rules of engagement and understanding of the outcomes of an engagement.

Such niceties as units representing actual, thinking, people or entities is antithetical to the design concepts of most competitive strategy games, just as variable unit damage output (free of other mitigating conditions). This is also a large part of the reason that StarCraft 2 and similar games don’t use unit skins – it’s incredibly important that a player be able to look at a situation and know exactly the capabilities of the units involved in that situation, and skins can throw this off.

Likewise, seeing a unit and not knowing precisely how it will react can throw off player expectations, to say nothing of units potentially putting themselves at risk or being out of position.


Sim City

But this philosophy is not the only one in games design, not by a long shot. Core strategy titles like Relic’s Company of Heroes and Dawn of War titles edge into seldom-trodden RTS territory, using limited but noticeable unit behavior profiles such as infantry ducking and covering for a brief time when exposed to artillery fire, being ‘smart’ enough to hide behind obstructions to protect themselves, and units having variable damage output.

Not to get too far afield on the subject, but I think it bears noting: while StarCraft is a game of pure, brutal efficiency and economy, Company of Heroes is a game of managing risk, of firing arcs and overlapping fields of fire. It is a game of precise army composition (in, I think, a way that StarCraft is not, by design) and obsessively covering your bases (again, in a way that I think StarCraft is not). While I think StarCraft has a much higher skill ceiling for competitive play, and is a more challenging game to play than Company of Heroes, I think that alternate design models that force you to think in different ways about how you handle your units are often overlooked in a genre that has become obsessed with chasing the rather nebulous ‘e-sports’ cachet seen in StarCraft.

untitledAlso worthy of note is that many more casual players of real-time strategy games get downright pissed that their units are dumb. Players of shooters, RPGs etc have come to expect an immersive experience that I think most, if not all, RTS games are not designed to provide in any way. Most RTS eschew physics, many eschew collision, and even line of sight or terrain in subservience to the all-important rules interactions that define these titles. To the designer and the competitive player, it almost doesn’t matter to some extent what the game is portraying – take StarCraft’s rules and put ancient Egyptian character models on the Protoss and you still have the same game.

But some players see a Space Marine and ask why they act more like a Roomba than a soldier, or why their Mark IV Space Chariot in space RTS X doesn’t engage that Flying Space Mutant Frigate that’s 1 px outside of some magical attack radius.

Where most RTS games design is handled pretty much the same way as Chess or any other board game might – that ‘units’ are no more than game pieces designed to interact according to predetermined, abstracted rule sets, that’s not all that we see. Few games take alternative approaches to unit design in the RTS genre, but there are games like Majesty in which the player does not control their units, who act, in that particular game at least, more as independent agents the player can coerce into actions by appealing to the motivations of these virtual entities. Now, in Majesty, the simulation here might not be particularly complex, but the dynamic between the player and the heroes they require to defend their fragile kingdom does represent the notion of the unit as a free agent with its own will.

Typically, strategy games in which the characters are represented as individuals tend to be, or feel like, simulations. Instead of pieces on a gameboard, whose rules could just as easily be a complete abstraction from reality as intended to represent some aspect of warfare, simulation rule sets tend to explore complex relationships. Now, this doesn’t necessarily have to be personal relationships – The Sims series, for instance, attempts to model the dynamics of human happiness – money, things intrapersonal relationships, almost a virtual representation of Maslow’s Hierarchy, come to think of it.

Taking a look at the Tropico series, we see economic and political relationships at play, as the player is asked to manage citizen happiness (to a point), and production efficiency of various cash crops, to a point. And maybe that’s part of the difference: competitive strategy games are all about pure, brutal efficiency while simulation type games are more about balancing disparate factors that may not have simple economics at their foundation.

Certainly, any simulation of human happiness would have inherent within it factors other than domination by force or min/maxing a resource system (not to say these simulations couldn’t be min-maxed: I’m sure in many cases the models that the Sims etc use are able to be gamed like any other mechanic).



The current landscape of RTS design is firmly rooted in providing the player as much agency as possible: full knowledge, should they have the presence of mind to see and process it, of the capabilities and behavior of their playing pieces and full and accurate representation of the outcomes of any action. From unit damage, to behavior to design, most of the current generation of RTS are concerned with boardgame-like interactions between playing pieces and player.

In my humble opinion, as with many other fundamental RTS design elements that tend to go unconsidered, I feel that there is room to play with varying levels of unit agency. Not just for the fun of it or for experimentation’s sake, but in the pursuit of unique and interesting competitive experiences. Or, I suppose, in the case of single-player RTS games, in the pursuit of more in-depth interactions and cohesion between story elements and what’s presented in the gameplay.

While I am unsure that additional unit autonomy would be an interesting mechanic in RTS games, I do feel that some additional behavior modeling and AI systems (see Company of Heroes series here as a for-instance) could produce some incredibly interesting in-match interactions in competitive games. Applications range from ‘convincing’ preexisting populations on the game map to side with your faction to support your efforts in the match, to the challenges of passing orders through a virtual chain of command, to game mechanics which model armed forces under your control gaining and losing confidence in the player’s ability to command (defection etc)

Of course, these sorts of things are easy to say, while  the full consequences of implementing them into a game or building a game around these interactions are more nebulous indeed.

Finishing Thoughts

Unit agency is a spectrum, and I think there is a huge untapped potential for RTS games to experiment with giving various degrees of autonomy to units. The very struggle to ‘convince’ units to do the player’s bidding, or convince neutral AI on the map that your side is the ‘right’ one to support, could potentially be very interesting mechanics that could make an appearance in competitive games.

eSports’ current obsession with removing random elements and giving the player unfettered control of the workings of their play pieces is understandable but I do not think it’s necessarily the be-all, end all of strategy game design. Moreover, increasingly it is clear that mechanics which make interesting, balanced, and engaging multiplayer experiences do not necessarily make for interesting and engaging single player experiences. Single-player RTS have a lot more room to break the mold than do multiplayer RTS games.

In conclusion, I think it’s fair to say a couple of things about the entire concept of units simulating intelligence and decision making in RTS games:

  • by and large, competitive players do not want it
  • the very nature of highly competitive RTS games precludes it, though to some degree counter-examples exist
  • it is interesting to think about in the context of competitive games, and would be interesting to see competitive games experiment with it more
  • there is ample room for unit autonomy and agency in more single-player focused games
  • the implications of not being restricted to think of a single player game the same way as a multiplayer game have not been adequately explored, and unit autonomy is one of the largest sources of differentiation available to designers of single player RTS experiences.



  1. Thanks for republishing this article, I think it’s a very good discussion to have and I hope that I can recreate my thoughts on the topic. I broke down my discussion of this into groups. Let’s start with how improving unit freedom impacts the accessibility of the RTS genre on new players.

    Accessibility: For those entering the RTS genre one of the first thing they notice is the impact micro has on the game. While learning macro/econ management is for the most part more important for new players, the difference in unit control and micro is something players notice right away both from their own game experience and if they watch the game through streaming. By adding more unit freedom it eases up some of the pressure on new players. For example in the Company of Heroes series units tend to have more freedom and that freedom tend to help new players. When they come under fire they will dive for cover, even if the player doesn’t notice right away. They will adjust, give chase if needed or retreat if overwhelmed, they provide an amount of leeway to new players. Now they aren’t removing the skill ceiling, a player actively controlling their units still will have the advantage but the gap is smaller than in other RTS games. This raising of the skill floor is a nice benefit for those unfamiliar with the RTS genre, and you can tell that CoH was built around this unit AI and freedom, it wouldn’t work with traditional unit control. But as any CoH player will tell you, this has a downside as well.

    “Obey me minion!”: Something veteran players who spend time with CoH will learn very quickly is that the freedom the units gain is the lack of immediate control the user has on his units. In a traditional RTS while a unit will follow your orders you can change their orders regardless of the circumstance (under fire, retreating, standing around) and their reaction is as fast as the player can issue new orders. This is the foundation for most micro in the standard RTS. In the CoH series, and likely any RTS with greater unit freedom, unit control isn’t precise, the units don’t respond or obey immediately, or at all sometimes, and this can lead to great frustration. Now I understand why the developers did this, CoH is a much more realistic RTS, units under fire won’t advance or hold their ground in an open field, if pinned they won’t retreat or relocated. The only time in-game the unit will respond immediately is the forced retreat option and the unit will return to your HQ, but you can’t control that unit while it does that. The greater freedom we apply to a unit the harder it will be to control, if they applied an instant response command to CoH the game would lose it’s uniqueness and credibility when it comes to combat and cover design. But let’s take this to the far side of the spectrum and see where that leads us.

    The realistic extreme to this idea: CoH series goal is to try and create a more realistic military experience and both the benefits and frustrations that come with that are realizations of this goal. But to take this to the border of this idea, we approach reality and the pipe dream, that’s been attempted but never really succeed at, of game developers which combines an RTS/FPS where the units the commander (the RTS player) are players as the shooters (the FPS users). This idea is enticing in theory for a number of reasons, for the commander you will have units with intuition, as watching a huge map can be overwhelming, having your units watch out for themselves alleviates some of that pressure. If a shooter sees something that the commander misses the unit can investigate themselves. For the shooter they have the benefit of knowing someone is watching over them, using information from other units to provide an overwatch. The problems arise in this idea quickly though when you realize this only works in the perfect situation, when a team knows each other, has experience with each other and has each others best interests at heart. In reality this would likely be a very small minority of players. For the average player they would suffer a negative experience. An incompetent or trolling RTS player could send their FPS players into bad situations, realistic sure, but not fun. A quality RTS player who is saddled with even a few poor quality or negative FPS players could cause the game to quickly spiral out of control, realistic to be sure, but not fun. I’ve said that twice because that’s the ultimate question, is it fun?

    But is it fun?: This is the question every game developer must ask themselves, and the more I think on the idea, increasing unit freedom seems to decrease the fun factor in the genre. CoH found a nice spot to be, the intelligence they provide their units is both unique and matches the tone it’s setting. This is one aspect where more freedom would be allowed, if it matches the tone of it’s setting or goal. I’ll discuss this in my conclusion but for now I want to get back to the “is it fun” question. With the modern staples of the RTS genre, the greater freedom would only increase frustration, which is never a goal for a game developer.

    Wrapping up: This is running long so let’s conclude. I am all for creative ideas in the RTS genre which tends to be a tad stagnant. CoH is the largest change the genre has seen in a long time and thinking about how it came to it’s design does tell me moving down the scale of unit freedom isn’t all bad, but it has to have a PURPOSE. The game’s setting is the only way it would work. CoH gets away with units disobeying or ignore orders because it gives the sense of war, when a unit is scared it takes cover, it will ignore a commanding officers orders if they feel it will get them killed. If a game developer can create a work or situation that creates a logical and intuitive reason for WHY units have freedom and may not always obey their commander, it creates a sense of belief for the player. I’m not sure how to do it, but hopefully someone more creative than me has some ideas.

    As you said, this is a spectrum and even in perfect unit obedience games like modern RTS games such as Starcraft 2, the units still possess some freedom. A few examples of this: If a unit is attacked by a unit with a range greater than it’s own it will gave chase unless it’s on a hold command. When you order your army to attack move and they engage an enemy force, they won’t merely attack the closest unit once the battle starts, they have targeting priorities, going after the most dangerous units first. This level of intellences is both a design choice but also exist because of improvements in technology and AI creation, it didn’t exist in early generation RTS games like Age of Empire, the early Warcraft games and the Command and Conquer series. Improvement in computer programming has made this balance work.

    I am a fan of the current RTS paradigm on unit freedom, units that aren’t dumb as rocks but still will follow your every command as fast as you can issue them. Having said that CoH changed my view a bit, not every RTS can pull of that style of game play, it works because of the amazing work by Relic. If a game developer can create a world and a reason to change the way we view unit freedom, I’d gladly try it, but I’m wary of giving units freedom purely for the sake of it, it doesn’t seem fun, but I have hope.

    This is rambling and very long (hey I kept it under 1,500 words at least!), but it’s a topic that I feel would be great to have a round table about, a lot of things to discuss and theorize about.

    Liked by 1 person

    1. A lot of this was informed by talking with progamers and developers. I’d like to see more independent unit AI but I think that’d be better served showing up first in RTS that target the single player experience. Which I’m in favor of, actually 🙂


  2. Hey, just stumbled across this. Good read.

    Personally, I’m all for more intelligent units in an RTS. I hate feeling like my units won’t breath unless I order them to. I’d prefer units that acted with a bit more autonomy, reacted just a bit more intelligently. Myself, I find it a mistake to think that such unit behavior isn’t as predictable, since most of it would just come down to units being smarter and carrying out a task based on their available surroundings. As long as a player is aware of what the unit’s most likely course of action is during a scenario, they have the freedom to selectively ignore or interfere with that unit based on their knowledge.

    Anyway, three stories of unit AI, good and bad, that I’ve seen.

    SCII: Here we have a bad example. Playing SP (where the problem still persists), queued a series of attack orders on the enemy base, moved away to manage some other things on the map … after a moment noticed that not only was the base still standing, but I was losing units. As it turned out, my units were being really stupid. The first marine to get within range of firing had done so, stopped, and started shooting.

    The units behind him had gotten confused, and couldn’t figure out how to move around him. This lone marine was standing there shooting, in a wide-open space, and behind him was a giant triangle of units bumping into each other, all incapable of spreading out and getting around the lone marine, who was still shooting.

    Worse, the AI had retaliated with several units, none of whom should have been any harm, but who were cheerfully murdering my army because even though they were under fire, they weren’t intelligent enough to react to that, and were just milling around dying. That was kind of a breaking moment for me with SCII, because any suspension of disbelief I had that the game was even supposed to be a representation of a war was gone.

    C&C3: I’ve got two good ones here. The first was an infantry squad I left standing in the middle of a bridge on guard mode. The moment a tank started coming for them, they ran off of the end of the bridge and into cover behind the bridge rails, where it would be very hard for the tank to run them over. By the time I hopped over to look, the tank had been wiped out by my other defenders, and then the infantry moved back to where I had told them to. They didn’t get run over because they moved.

    The second one was using a Titan, a walking mech that can crush other lighter vehicles beneath its feet. I had it taking a resource area, and the enemy sent a tank. No problem, Titan versus light tank is a predetermined battle. Then, the tank moved a little closer, closing the distance, and without any action on my part, the Titan stomps forward and smashes it, then continues on to where I wanted it to go. Saved a good 1/4 of its life by choosing to crush the tank rather than shoot it. No input from me.

    Stuff like that makes an RTS. Those units actually feel like soldiers, even when they do dumb stuff.

    I’ll take a much more intelligent unit base any day. You’re 100% right that there’s a ton of potential there, and I keep hoping that someone will take it. I think once they do, that’s when we’ll see a modern RTS that truly starts hitting the “competitive” sphere in real fashion.

    On a side note, my buddy has made the observation that among current “e-sport” RTS games like SCII, MOBAs are eating them alive for playerbase.

    Liked by 1 person

Comments are closed.

%d bloggers like this: