Cloaking is awesome. At least, some types of cloaking. Depending on who you ask. Let me start again.

Cloaking is polarising. Some love it, some hate it, and being on the receiving end of an invisible strike can feel pretty terrible. To complicate things further, the specific mechanics of cloaking vary wildly between games. Sometimes cloaked units are near-immune to damage and can fight with impunity, while cloaking in another game might be dispelled by the slightest touch. But where it gets really interesting is how a single set of cloaking mechanics can be used in different ways, within the same game.

Zero-K has a lot of cloaking, which equates to a lot of ways for it to go wrong. In addition to personal cloak, cloaking fields were added early in Complete Annihilation as a way to differentiate Arm from Core, and we have been nerfing them ever since. Over time, we started thinking about cloaking in terms of two distinct use cases.

• Tactical Cloak: The enemy knows where your units are, but has trouble engaging them.
  
• Strategic Cloak: The sneaky sort of cloak, where the enemy is unaware that you have units in the area.

Dark Templars from Starcraft are quintessential tactical cloakers. It is abundantly clear when a base is full of Dark Templars, from all the dead workers and sci-fi noises, but they cannot be fought without being revealed by a special detector-type unit. This example relies on Starcraft having very few ways to damage cloaked units (they even remain cloaked while fighting), but the concept of tactical cloaking applies across all games. Tactical cloaking in Zero-K, and many other games, depends on uncertainty in the precise location of a cloaked unit. Sure, units should shoot at "empty space" if they know a cloaked unit is there, but it does not take much uncertainty to make such a shot highly questionable.

It could even be argued that Starcraft lacks "true cloak" (if not for the fact that arguments about the "true" anything tend to be pointless), since units reveal themselves to the enemy with a faint shimmer (except for burrowed units), which probably helps cloak feel fair. Mechanics like these are unavailable to Zero-K though, since such a disconnect between player- and unit-knowledge is against the rest of the design.

Fighting the UI thrives on situations where the player knows something, but is unable to convey this knowledge to their units. Cloak is relatively easy to dispel in Zero-K for this very reason, with the goal being that a unit should only stay cloaked when there is real uncertainty about its position. To start with, structures cannot cloak, since nobody wants to walk their units through how to shoot at a cloaked turret. Units also decloak when they fire, since the sudden appearance of a projectile is a dead giveaway, and when they take damage, since a projectile disappearing mid-air is also quite suspicious. There are non-suspicious ways to take damage, such as walking through fire, but they also decloak for consistency and to provide cloak with more counters.

Cloak is also pierced by straying too close to the enemy, with "close" being defined by the decloak range of the cloaked unit. This range is an important balance factor for personal cloak, with some personal decloak ranges being quite large. Other units have a consistent size-based decloak range, for when they enter a cloaking field, and cloak fields also apply this range to personal cloakers, if it is an improvement. Cloak field projectors cannot have their decloak range reduced, and mobile projectors have large decloak ranges to keep tactical cloaking in check.

Proximity-based decloaking allows Zero-K to completely avoid passive long-ranged detection abilities. Active detection abilities exist, such as napalm missiles and Sparrow's reveal-on-death ability, but cloaked units are usually countered by a combination of base design and patrols of light units. We like the nuance involved in this system, it has more play in it than having units that passively negate all cloaking in an area.

A lack of widespread detectors allows strategic cloaking to be more viable late into the game. This is surely the dream of everyone who hears about cloaking, they want to be James Bond; sneaking into the enemy base to blow it up. This is great, but Zero-K also wants to be a fun multiplayer game, and having your base blow up out of the blue is not great. So strategic cloaking is tuned mostly for hiding army movements within your base, and in no-mans-land. Still, infiltration is the dream, so it is always on the table. Our goal is just to make letting a cloaked army into your base feel like an avoidable mistake, and besides, Antinukes still provide sneaky infiltrators a way to blow up an entire base.

As with most definitions, there are edge case that fuzz the distinction between tactical and strategic cloaking. A cloaked army might start out with strategic cloak, turn tactical when they strike a target, then shift slowly back to strategic after the survivors disengage. Particularly popular uses of strategic cloaking can even take tactical aspects when become standard enough to anticipate. DotA's Smoke of Deceit is a great example of this, since it has a massive decloak radius, so is clearly meant as a strategic cloaking tool, yet skilled opponents will anticipate smokes and respond when they feel they are a strong possibility.

Edge cases aside, what is the point of these categories? The point is that player feedback tends to revolve around adverse events, and events involving cloaking tend to cleanly fall into one of the two categories. Sorting feedback is great, but where categorisation really pay off is in how it informs the sort of balance change that might be required. A problem with strategic cloak suggests nerfs to speed or detection range, as with Conjurer after it gained a cloaking field. Problems with tactical cloak suggest nerfs to recloak time and the area that a cloaking field can cover.

As alluded to earlier, cloaking fields have been overwhelmingly nerfed since their debut. In part this will be down to them being figured out, but there was probably also an initial bias towards making them powerful. Units used to decloak for much less time when they took damage, and cloak fields have been steadily getting smaller. Structures used to be cloakable until we realised that cloak is boring or frustrating on most structures. Cloaking your Antinukes and Singularity Reactors is obviously good, but removes the whole point of these structures being juicy weak points.

Tactical cloaking with cloaking fields has caused more problems than all other forms of cloaking. Personal cloaking is fine, since any issues with it are tied to individual units, and balancing units is relatively straightforward. Even the Phantom's ability to fire without decloaking has been balanced. Cloaking fields are a different beast though, since they give cloaking to everything, and our principles prevent us from carving out arbitrary exceptions. Skirmishers and artillery have been an issue, since cloaking them can make it too difficult to retaliate. To solve this we nerfed the recloak time for non-personal cloak to six seconds. This still lets armies make a cloaked surprise attack, but leaves the units decloaked as they fight.

Bombs are the most exciting units to put under a cloaking field, and the trickiest to balance. We have tended to nerf the field projector itself, decreasing field radius and movement speed, while increasing decloak range. The aim is not to remove cloaked bombs, since that would remove the joy of figuring out a cool combo, then realising that it works. Rather, the goal is to give the defender something to do. This is part of making cloak feel fair; the interaction between attacker and defender should not end as soon as the cloaked units are revealed. In the case of bombs, finding and destroying the cloaking field is a consolation prize. This is part of what makes Skuttle so hard to balance compared to Widow: they work alone.

In addition to structures being uncloakable, cloaking is disabled in a few more cases. One is the fact that shield-projecting units cannot cloak. This is a unit intelligence thing; if you can see the shield, then you should be able to pinpoint the unit that is projecting it. Technically the shield could cloak too, but then projectiles would collide mid-air, or shields would have to pop into existence when the unit decloaked, which would look messy. Another exception is cloaking underwater; fully submerged units cannot cloak. This is mainly due to a lack of good ways to detect underwater units by conventional means, and because underwater units already require sonar to detect.

The relationship between cloaking and sonar raises a good point; is cloaking just an extension of line of sight? Vision, radar, cloaking and sonar could be unified under the banner of How Not To Be Seen. In fact, most of what I have said about tactical and strategic cloaking could apply to line of sight and radar. There are times when a lack of vision makes it hard to fight, and other times when a lack of vision hides the fact that there is anything to fight. So there are parallels, but also many differences, the exploration of which will have to wait for a future article.

it is finally time to talk about weapons, starting with how they are fired. The weapon design space is huge, thanks to projectile physics, and we put it to good use. But such freedom is a double-edged sword, since much of the full design space contains frustrating or incomprehensible mechanics. There are interesting games in this space, but Zero-K is first and foremost a strategy game, which means it needs to be stable and predictable enough to reward planning. So, to combat the full freedom of physics, we installed a few guard rails around weapon design; a loose set of rules that, if followed, limit how badly a weapon can break the game.

There are rules for many aspects of weapon design, but those around aiming and firing are some of the most important. The rules consider the following three aspects of our design goals.

• Intelligence - Aiming and firing are just abilities, and units should be good at using their abilities.
  
• Predictability - Players should be able to learn how units aim and what causes them to fire.
  
• Verisimilitude - Weapons should feel like they can be used in ways that make sense within the observable fiction of the game world, rather than just being a bundle of arbitrary mechanics.

As you might have guessed, projectile physics causes most of the conflicts between these goals. This specific set of problems is mostly limited to descendants of Total Annihilation, since projectiles have a minor role in other types of games. The standard RTS projectile is a animation and a short delay between firing and dealing damage, since they are guaranteed to hit their target, and only their target (unless something unusual happens, such as death or teleportation). This limits just how stupidly units might misuse a weapon, and it allows weapon fire to be quite predictable.

Total Annihilation probably used projectile physics to enhance verisimilitude, but I would disagree that this means other games inherently lack it. Every RTS simplifies many aspects of reality, and simplified projectiles are quite popular. How closely a simulation hews to reality is only loosely related to the experience of a reasonably consistent and believable game world. There are, of course, many players that seek out strict realism, but Zero-K does not simulate projectiles to be realistic. We use physics to make intricate, yet fundamentally simple, mechanics that reward intuition and creativity. But I digress, physics generally seems to cause problems for the goals above, rather than solve them.

Physical projectiles are hard to use. You shoot at the enemy, but then the enemy moves and the projectile hits empty ground. So you shoot where the enemy is going to be, but they dodge the other way, or you miss due to innate weapon spread. Units do a decent job at leading their shots, but many still miss. This makes deciding when to shoot of utmost importance, but this decision has to be quick, which risks making it unpredictable.

To make matters worse, the sense of verisimilitude subscribed to by Zero-K says that, ideally, weapons only provide two in-world abilities:

• Change where the weapon is aiming.
  
• Fire the weapon.

In other words, we think of units as having the same basic abilities as a player in a first person shooter. In practice the implementation does not go that far, but we do at least let units blindly fire at the ground, since a cloaked bomb might be there. But even this limited ability means that, in theory, players could manually control every shot, which raises the spectre of fighting the UI. So units have to be smart enough for players to not need to resort to manual control.

We could write some whizz-bang first person shooter-level AI, but then battles would amount to little more than watching units careen around the screen. This brings us back to predictability, so perhaps intelligence and predictability are fundamentally opposed. Worry not, the deadlock can be broken by imposing limitations upon weapons that are designed to make them simpler, so that units can be reasonably smart without breaking the "intelligence budget" of predictability. These limitations tend to be quite gamey, beyond the realm of pure physics, which causes trade-offs with verisimilitude.

Consider the ubiquitous range ring and its implications for homing missiles. A Fencer can only fire missiles at enemies within a circle around itself, but an enemy that dances just out of this circle will still be hit, which is a bit weird. What happens if we let just the goals argue it out?

Verisimilitude: That's ridiculous! The missile can demonstrably hit units beyond its arbitrary "range". Fencer should be firing in more situations, at everything that it can physically hit.Intelligence: Hmm... sounds exploitable, we don't want to fire at things barely within physical range. Borderline targets could (and should) just step back and let the missile run out of fuel before impact.Predictability: We could just only let missiles fire at structures, not units. Those are sure to stay still.Verisimilitude: That's super arbitrary!Intelligence: How about we use the target's max speed, accounting for slow and EMP, hill climbing, facing, angular velocity....Performance: uhh...Predictability: And how is anyone meant to know which combination of factors will or will not result in a Fencer firing a missile?

At this point we might placate verisimilitude with some mumbling about "missile lock range", but background lore is no match for what players can observe ingame. Besides, we have many ballistic weapons that could also shoot a bit beyond their range ring, and this cannot be explained by any sort of "missile lock". So the the fact is that verisimilitude lost this battle, and the best we can do is make these situations as simple and consistent as possible.

The verisimilitude of Zero-K demands that units can do everything that they look like they are capable of, predictability demands whatever is simplest to communicate to players, and intelligence only cares that the resulting ability is hard to screw up. Designing a weapon with these constraints in mind looks a bit like this:

• Imagine the ideal form of the weapon, packed full of verisimilitude. A Glaive holds a gun in its hands. It can actuate its arms in any direction and pull the trigger at any time. It is a real robot.
  
• Next consider the proper use of the weapon. Should Glaives spend their time shooting wildly into the sky? Probably not, unless an enemy is flying overhead.
  
• Then impose as many behavioural or mechanical limitations as you like, provided that you only cull the stupid or pointless ways that the ideal weapon might have been used. These should help the unit avoid being stupid. Shooting into the sky is almost always pointless, so Glaives only do it when there is an enemy flying overhead.
  
• Finally, add a small set of standard limitations, such as range rings. Consistency is key here, since the goal is predictability, and arbitrary limitations hurt verisimilitude the most.

In practise, the process is not as simple as working through four steps. Most limitations affects all three aspects simultaneously, and sometimes you need to revise the ideal version of the weapon to make the downstream implementation workable. It is also possible to start at the implementation, then work backwards to figure out what sort of ideal weapon would result in it.

One of the most important corollaries of the process is that weapon behaviour cannot depend on the existence of a target. This is known as the No Void Ray rule, as it was developed back when Starcraft 2 had a unit that gained damage the longer it fired. The unit was redesigned in the first expansion, probably because shooting your own stuff is allowed in Starcraft, but incentivising players to do it looked a bit silly. In any case, Void Ray would not even make it into Zero-K, since there would be nothing stopping it shooting into the sky when idle. Maintaining its damage bonus is smart, so we could not in good conscience prevent it from doing so. This would make the mechanic pointless, so we would avoid adding it in the first place.

The same rule applies to lock-on mechanics, or mechanics that penalise a unit for switching targets. A weapon shooting down a line of enemies should not care which unit is "aiming at", since it all involves aiming in the same direction. Any mechanical distinction would drag the interface abstraction of "targeting" into the realm of the game world. There should be no "best command" among those that cause the unit to aim and fire in the same direction. Commands are meant to help the player manipulate the game world, not be themselves be something that require optimisation.

There are ways around the No Void Ray rule. For example, the weapon could apply an escalating status effect on the target that causes the target to take more damage from that type of weapon. This would look a bit like a Void Ray, and even solves the problem of charging by shooting at your own stuff. However, this particular mechanic would be discouraged by Physics vs. Formulas, but in general there are often ways to make a design work within the weapon design guidelines.

Perhaps it is best to think of the weapon design rules as a set of guarantees, rather than a set of restrictions. Working within the rules keeps Zero-K consistent and delineates a relatively safe part of design space. These restrictions can aid creativity, and not only via the usual refrain, but because the safety guarantees free up brainpower that would otherwise be spent making sure that an idea does not completely break the game. For example, the rules inherently disallow designs that incentivises players to shoot at their own base. So the rules can be bent, but more care is required, because the rest of the game is built on their assumptions.

Anti-air weapons are the most blatant violators of idea that weapons can be fired in any direction. Being so blatant is great actually, since a large and well-defined exception feels much less arbitrary. Every weapon falls into one of the following categories.

• Weapons that can only shoot at aicraft. Units with these weapons have "Anti-Air" in their two-word role description.
  
• All other weapons. These weapons shoot at aircraft if they have a decent chance of hitting. Melee weapons can and do hit planes, when the wielder is thrown high enough.

We could have avoided creating the special Anti-Air category. One approach would be to make aircraft significantly slower or flimsier, to allow their best counters to be weak enough to hit ground as well. Unfortunately this would make air units more like ground units, or just weaker overall, violating Quant's Rule. We also considered giving anti-air weapons upwards-facing firing arcs, to allow them to shoot into the sky without hitting the ground. This has poor implications for predictability, as an inverted cone would give anti-air variable range depending on terrain and cruise altitude, as well as make for great artillery against cliffs. An inverted cone also just kick the can down the road, as it raises the question of why this amazing weapon cannot be built at an angle.

None of the alternatives to Anti-Air were as good as defining a simple category that gives air the freedom to be powerful and different. That said, most games with air have Anti-Air, so perhaps the more interesting part of Zero-K is the lack of strict Anti-Ground. This choice is a type of design minimalism, we started with projectile physics and then added as few mechanics on top as seemed necessary. We could have widespread Anti-Ground, either as a targeting category, or just by making planes fly so fast and high that nothing but Anti-Air could hit them, but the game works fine as-is.

Truth be told, ground units shooting at planes is a feature, not a bug. The goal is not realism, the goal is the feeling of a fairly consistent world, and we like the extra interactions between different parts of the game. In a sense we picked our flavour of verisimilitude because we like where we end up when taken to its logical conclusion. But now it feels like this article is about to fold back on on itself. Just remember, for future articles or discussion, that Zero-K units are generally allowed to aim their weapons and fire at whatever they want, provided they agree to avoid using this freedom to do anything too pointless or stupid.