Cold Takes/15 - Experiences With Veterancy

Cold Take #15 - Experiences With Veterancy

♥

This week we're back to Sprung's series on unit improvement mechanics missing from Zero-K. After Mighty Morphing, it's time to tackle veterancy. This one still exists, well, sort of.

Already in Zero-K's oldest ancestor, Total Annihilation, units could gain veterancy via kills. The TA interface was fairly cryptic as to what exactly this meant: it provided no further information beyond adding the single word "Veteran" when mousing over units with at least 5 kills. Those of us who spent days shelling enemy bases with Big Berthas often noticed the improved accuracy. With a lot of effort it was possible to discover a minor increase in health as well. Frontline combat units were out of luck though, with veterancy being quite hard to obtain and having barely noticeable effects.

Zero-K is not constrained by the TA engine though, it runs on Spring (now Recoil), which expanded upon veterancy from early in its development. So by the time Complete Annihilation came around, veterancy did three things:

• Increase the health of the unit.
• Reduce the reload time of the unit's weapons.
• Reduce the inaccuracy of the unit's weapons.

Spring also moved to a fairer system of doling out experience: damage dealt is counted rather than kills, and the cost of the attacker and defender are taken into account. This means a cheap glass cannon like Glaive can rack up a lot of experience by shooting at a commander, while a Detriment can destroy whole armies without learning much. In a sense, veterancy depends on how well a unit is expected to do, given the resources invested into it. Higher veterancy units also gave out more veterancy when damaged, and the bonuses were applied smoothly with formulas rather than at set thresholds. Veterancy was more attainable and impactful, but this revealed problems.

Consider increased health, at certain health thresholds a unit starts taking extra shots to kill. On its own this is fine: C&C games, where veterancy grants both health and damage, tend to make anti-infantry units such as Tanya deal either 90% or 110% of basic infantry's health in damage per shot, specifically so that it flips between taking 1 or 2 shots depending on the shooter's and/or the target's veterancy. One of the things that prevents such elegance in ZK is that the rock-paper-scissor relationships between units are not as sharply delimited or as closely followed. You can't have simple rules like "basic infantry with 2 chevrons can survive a commando hit" if there is neither basic infantry nor commandos, nor discrete chevrons for that matter. Role counter relationships such as raider (roughly corresponding to basic infantry) and riot (Tanya) exist, but there is massive variety within them. Players would have to learn and memorise a huge relational spreadsheet of how each of Zero-K's dozen raiders fares in a fight against any other raider at any given pair of ranks to stay competitive, not to mention all the other unit types.

Reload time caused even more trouble, since time is actually discrete. In layman's terms, this means that reload times have to be in steps of 0.033 seconds. Again, raiders tend to be affected by this the most, as their fast firing weapons don't have much wiggle room. Glaive has a reload time of 0.1s, which is three steps of 0.033. From there, you can only reduce it to two steps, which represents +50% more fire rate. This is too steep! Another effect is that some units with slow projectiles were designed to land their projectile just before their reload finishes. This is a crude form of overkill protection, since if the target is dead you don't need to shoot it twice. A slight improvement to reload meant the unit would waste the follow-up shot on a unit that was just about to die.

Inaccuracy was usually the least problematic of the three, since most units tend to be perfectly accurate. But some are inaccurate on purpose, and a mandatory increase in accuracy would drastically redesign the unit on the fly, leaving you with something potentially useless for its original role. For example, an accurate Tremor fails to damage most of its usual area of effect, making it useless against cloak or wide terrain deformation. (As an aside, the original CA Tremor was just a tiny Disco Rave Party on wheels.)

Specific issues aside, all veterancy bonuses further reinforce snowballing in an already snowbally genre. Losing units makes your army weaker, but with veterancy it also makes the enemy army stronger. This necessitates some way to avoid even battles snowballing into victory after a few early pickoffs, either by diminishing the advantage or facilitating comebacks. C&C seems to achieve it largely by having units counter each other very hard, such that even elite units cannot hope to stand against their counter; there is also a cap on veterancy level so a unit cannot grow into infinity. Warcraft and DotA heroes give more experience and take longer to respawn when killed at a higher level, and each level is progressively harder to attain. Of course we could come up with a decent system, but is the extra complexity worth it? Besides, ZK already gives you a little extra for killing enemy units; but where other games do that via veterancy, ZK gives you wreckage to reclaim. This may not feel as cool as getting elite units, but it is such a deep and elegant mechanic we are quite happy with it.

Another reason to avoid veterancy bonuses is that simple stat bonuses affect formulas rather than physics. C&C again has a decent approach there: veteran units that reach the highest "elite" rank sometimes get an improvement in behaviour in addition to raw numerical stats. For example, heavy tanks get splash damage, or RA3 infantry gets immunity to dog stun. But at that point you become something more, a different unit in some ways, which should be reflected in its visuals. Rank icons, the way they are used in every other RTS, are way too salient for mere stats, and too un-salient for behaviour differences. The main hurdle here is that ZK artwork is essentially a loose bunch of assets collected from a wide gamut of volunteers over years, and there is simply no manpower for introducing such extras. Historically we actually had some models with a handful of progressively stronger looking variants, but that was just a drop in the ocean with barely any consistency. This is also one of the big reasons why we don't have research - a stronger unit needs to appear stronger visually as well, whether via research, veterancy, or any other mechanic. What you see is what you get.

Soon enough, stat gains from veterancy were removed, and technical capabilities were fairly low at the time so it wasn't really possible to implement things like weapon type changes. But veterancy is cool, we must have it! As hinted in the article on morphing, the solution was to let units with enough veterancy morph into whatever was a bit stronger and looked vaguely appropriate, and then sometimes not even that. Flea morphed into Glaive, sine they both lived in the "cloaky" factory at the time, i.e. Arm T1 KBot. Glaive morphed into Reaver, then Knight. Knight morphed into Crab, since Knight lived in the "spider" factory, Arm T2 KBot. This chain ended at Detriment. Many units could upgrade this way via veterancy, and many upgrade paths stayed well after factions were merged and factories shuffled.

People loved this! The fantasy of a Flea evolving into Detriment seemed really cool. Unfortunately, it remained a fantasy. Sure, people could do it in games vs AI, and there is one known case in PvP where a top 10 player did it as a show-off when toying with a first-timer, but chain-morphing never really happened in competitive games. There were just a handful of viable morphs that people aimed for. But in addition to being slapped onto effectively random units without any prior design, there were all the problems of morphing, and veterancy made it extra bad because it was so reactive. When your Knights get countered by skirmishers, Crab is the perfect answer - but you cannot plan for your Knights to get enough veterancy. You can merely hope one of them randomly happens to survive in combat long enough, and great power swings should not come from such unreliable sources.

Ultimately, these veterancy-based morph options were removed as well. As a fun fact, this revealed a minor factory design problem, which led to the creation of Redback. Spider did not have the "DPS riot" role filled by any spider at the time, but Flea was so cheap that you could reliably get one to high enough rank to morph straight to Reaver by shooting anything for about two seconds. The hole in the lineup became evident after losing access to Reaver, so Redback was quickly thrown together to fill it.

Modern ZK units still gain veterancy, represented by rank icons above them. Mechanically, it does absolutely nothing (beyond inspire modders). But getting icons makes players intuitively feel good and rewarded, so we keep it. Often though, players start to suspect that there isn't actually any benefit to veterancy, and they ask questions. Should I aim to gain veterancy? Should I protect my veterans? To which we say: yes! The effect of getting ranks is that your enemy no longer has the units you had to kill to get those ranks, and this is already something that directly helps your goals. Veterancy is also a nice way to explicitly gauge how well a unit is doing, though there is some bias against categories of units like artillery and superweapons that often shoot into the fog of war, since in order to prevent leaking information, only visible damage is counted.

The final clue to why we avoid rewarding veterancy can be seen by looking at high skill WC3 or DotA gameplay. And after all, what is levelling heroes if not an extreme case of veterancy? In both these games, you often want to kill your own units just to prevent the enemy from getting experience. In DotA the metagame is optimised so that each hero on the team gets the "correct" amount of experience (and gold, which comes from killing enemies as well so can also be considered a soft form of veterancy), with supports sometimes forgoing it in favour of cores.

In ZK we anticipate problems and immediately bite whatever bullets may appear in the future. The powerful UI would allow you to squeeze quite a lot out of veterancy optimisations such as selective self-destruction to prevent enemy veterancy, and holding fire so that units with worse scaling don't "steal last hits" from the better ones. High skill ceiling? Definitely. Mostly obsoleted by automation? Maybe. Do we want to find out, and possibly lead players into the weeds of optimising this kind of engagement minutiae? Rather not.