Cold Takes usually cover a concept used in the design of Zero-K, but this one is about a term invented by players to describe an issue with the game. This issue involved the "retreat range bonus", but the term was initially perplexing because it describes something that is both emergent and ubiquitous. Nobody could tell me how other games dealt with it, and it seemed very hard to remove without breaking some of the basic rules of Zero-K.

The retreat range bonus is the apparent range gained by units as they retreat from a battle. Consider two identical units called Chaser and Runner, locked in a battle where Chaser follows Runner while Runner tries to retreat. The retreat range bonus is the observation that, at a certain distance, Runner will be able to fire, while Chaser cannot. In other words, it appears as if Runner gained bonus range on Chaser merely by retreating. Taken at face value, this looks like a weird arbitrary bonus, and the type of thing Zero-K would certainly not want.

The paradox of the retreat range bonus is that it was not added; it is just the inevitable result of the common type of game physics. If a game physically simulates its projectiles, then it probably has some form of retreat range bonus. Many, possibly all, first-person shooters have it, and there might even be a term for it in that space. Games like Supreme Commander and Planetary Annihilation should have it too, but to this day I have not found any discussion of retreat range in those games. I am often met with confusion when I ask people about it, even after explaining the generality of the mechanism.

The retreat range bonus is caused by units leading their shots. In other words, to hit a moving target, you need to shoot at where it is going to be, rather than where it is. In particular, for a unit to hit a fleeing target near the edge of its range, it would have to fire at a point beyond its range. Doing so would violate the No Void Ray rule, i.e. that the capabilities of weapons are independent of their target, so the unit cannot shoot at such a target. In the case of Chaser and Runner, both units are just within each other's range, but Chaser is moving towards Runner while Runner is moving away, so only Runner gets to shoot. The effect is compounded by the fact that units predict when an enemy will enter range, and fire towards the edge of their range in anticipation.

The retreat range bonus is fundamentally caused by a lack of Galilean relativity. Galilean relativity only really shows up in actual sports, and is responsible for mechanics such as taking a run-up to throw a ball faster. Simulated games tend to avoid it, and are right to do so. Humans are pretty good at accounting for relativity when doing human things, but not when shooting pretend rocket launchers. Imagine firing while strafing, only for the shot to go wide due to your own velocity. I would definitely play a Zero-K mod with Galilean relativity, just to experience the nonsense, but I expect it would make units quite stupid, as many of them would want to suddenly lurch towards their target for extra range as they fire.

So far so good. The retreat range bonus is baked into the game, and removing it would make less sense than leaving it in. The problems arise when we consider the tactical implications. The Chaser-Runner battle can be in one of three states, depending on the distance between the participants:

A) Neither unit can shoot at the other. B) Only Runner can shoot. C) Both units can shoot.

Runner would really like to deal damage for free by keeping the battle in the sweet spot at B. Chaser absolutely does not want to take damage for free, and would either like to kill Runner or chase it away. Both sides can control the battle by slowing down, for Chaser, this widens the gap from A to B to C, until Runner escapes completely, while for Runner slowing down closes the gap. This shared control of the situation creates a tight feedback loop of manoeuvring and fakeouts, with the potential for Runner to push its luck for a shot at killing Chaser at no cost. This results in intense bouts of clicking, especially in the hands of hardcore players with low latency.

The retreat range bonus came up as part of a broader set of feedback about the swinginess of the early game. In short, some 1v1 players felt like too much hinged on the first few raider engagements, and that too few games made it out of this stage. This was by no means a universal take, but it was prevalent enough to be worth some experimentation, as we could end up somewhere that most people would prefer. Chasing feedback from players that are particularly sensitive to an issue can be useful, as this can reveal things that generally improve the game, yet were below the complaint threshold for most people. The feedback and testing culminated in the Superfluid Update.

The Superfluid Update removed the retreat range bonus for raiders. These units have low range and fast projectiles, which made for small sweet spots, but it was large enough to dominate the opening. Some players were against removing the retreat range bonus as they liked the tactical options and its potential for large swings. Others just liked being able to furiously micromanage a raider to kill many times its cost, but this was offset by alleviating the frustration of players that were on the recieving end.

The deciding factor in the removal of the retreat range bonus was the fight with the UI inherent in having units die by being baited into the sweet spot. The other option was to improve our unit AI to account for the sweet spot, but I felt that such an AI would be too unreliable and intrusive. These fights were not just about micromanagement; there were decisions to be made. For example, a fight between units of different speeds might have one try to push through the sweet spot, at the cost of taking some damage, and whether to do so depends on your strategy and the whole state of the game. The player needs to be able to implement these choices, but the remaining options are state toggles and special commands, both of which are much worse than letting players rely on simple move commands. Essentially, removing the retreat range bonus was a case of tweaking mechanics to preserve the power of the smoothest and most fun part of the UI. Feeling good to control is much more important than any specific interaction.

On the technical end we removed the retreat range bonus by forcing units to not lead their shots, then hacking physics to cause their shots to lead themselves. Every time a Bandit shoots, its projectile is reoriented slightly to point towards the predicted location of its target. This lets raiders shoot very slightly beyond their range, breaking the No Void Ray rule, but their high projectile speed makes the effect barely noticeable. Unless you pause the game, zoom in, take a screenshot, then draw some arrows. If you cannot unsee the small deviation ingame, then I am sorry, but that is the risk we take when pulling back the curtain.

As for the Superfluid Update itself, the rest of it involved making raiders and constructors take a bit longer to kill, and generally making units feel more responsive. These changes were walked back over time, to an average of the old and new states, but this was by design. I prefer to overshoot reworks to probe the outer edges of the reasonable area in the design space, to make a better map with which to navigate to the final state. The raider retreat range bonus has not returned though, and predictions of the raider game becoming trivial or pointless did not eventuate. Raider battles still involve a lot of position adjustment as players try to keep all their raiders firing on a target, while trying to block enemy raiders behind each other. Still, it would be good to hear some of the feedback from players who were there six years ago, to see how the changes have aged.

Index of Cold Takes