I've heard the term "perfect imbalance" thrown around a lot, which seems to imply that it is okay for different game options to be strictly superior to others as long as the 'better' options require more player skill to use.
What strategies does one employ, generally, to achieve this? And am I even understanding the principle correctly?
Perfect imbalance relies on players shifting their strategies (collectively forming the meta of the game) in order to neutralize strategic advantages others might use against them.
That means we have two main ingredients:
Intransitive mechanics are named in reference to the Transitive Property in math, which says that if we know:
A > BB > Cthen for a transitive relation we can safely conclude that A > C.
For a transitive rule and a finite set of options, there's always one that's not less than any other. If we think of our relation as being one of gameplay advantage, that means one option isn't strictly weak against any other option, though it might be equally advantageous as several others.
The risk here is that it's fragile - if we don't get that equality of advantage exact, then we end up with one option that's strictly better than everything else, and ones that are strictly worse. Nobody wants to play a worse option, so our game can become one-note as everyone piles up on the dominant options.
Intransitive or non-transitive rules break this. The most familiar example is probably Rock Paper Scissors, where:
Rock ---defeats--> ScissorsScissors ---defeats--> PaperPaper ---defeats--> Rock (the opposite of what a transitive rule would say, given the two rules above)Now there's no way to put these into a strict order of better & worse - there's always a wrap-around point somewhere, some way to "beat the best." This makes balancing these types of mechanics much easier than in the transitive case, because we're not relying on exact parity of numeric relationships to keep two options balanced - instead it's baked into the cyclic structure of advantages.
In games you'll often see this expressed as counter relationships. A sniper character is weak against a fast/evasive character who can close the distance and take them on at close range, but that character will often be lightly armored and weak against a heavily-armored bruiser archetype, who might in turn be weak to being sniped from afar...
If a group of players is dominating by using one of these strategies, say the sniper, then other players can notice this and counter by deploying more fast/evasive characters, dethroning the snipers or forcing them to switch strategies. This keeps the strategic landscape dynamic, fighting the tendency to settle on one optimal strategy or tactic.
This works even when one of these characters/options is, by some definition, strictly more powerful. I recommend reading In Schreiber's Game Balance Concepts blog for the full details & math treatment on why. There he shows that even if we play a version of Rock Paper Scissors where rock wins count double (making it, at first glance, twice as advantageous to play), there's still a way for players to balance their strategic choices that keeps the game fair and interesting overall, without the game collapsing to always playing a particular option.
It doesn't always need to be exactly 3 options, or all in a single loop. Your move / unit / character / build / strategy relationships can be a spaghetti fractal of little loops inside bigger loops. As long as no one option (or group of options) is strictly dominated by all other options, nor strictly dominates all other options, then a sufficiently large & motivated player community will tend to seek out a ratio of choosing these strategies that overall roughly neutralizes individual advantage in the context of the current meta - something called a Mixed Strategy Nash Equilibrium if you want to get into the game theory of it. :)
Just beware that we can't rely wholly on this self-balancing feature of player communities to absolve us of all design responsibilities. Every strategic choice I need to make as a player just to counter the dominant strategies in the current meta is one less choice I have available to express my own play style and preferences - it's a choice that's effectively already made for me, constraining my play space. A little shifting of the strategic tides as the meta evolves can keep the game fresh over time, but when most of my choices are dedicated to fighting against the same lopsided design then I'm likely to be having less fun, and go look elsewhere for my challenge. So we still need careful balance, even when we allow it to be imperfect.
"Perfect Imbalance" means it is acceptable for a game to have one or two game playstyles which are objectively better than the alternatives, as long as there is a feasible counter for them.
Let's take the RTS game Age of Empires II, for example. (Or rather a simplified version of it. I omitted some details for clarity. AoE fanatics, please excuse my oversimplifications).
There are 3 essential combat unit archetypes: cavalry, infantry and archers. Of those three unit types, Infantry is the one with the worst power/cost ratio. Anyone building only infantry will lose against an opponent who builds only archers or only cavalry. So with just those 3 basic unit-types, there is no reason to ever go for infantry. That's a glaring imbalance.
But there is a catch: We haven't talked about the non-essential combat units yet. Both cavalry and archers have a dedicated counter-unit: Pikemen and Skirmishers respectively. A cavalry-army is easily defeated by a pikemen-army and an archer-army is easily defeated by a skirmisher-army. But outside of these two situations, both pikemen and skirmishers are pretty much useless. There is nothing they can do except countering the two otherwise superior strategies. So if the enemy tricks you into wasting resources on preparing to counter a strategy she isn't actually pursuing, you will be in a pretty precarious situation.
Infantry has no such direct counter-unit. So you can't really do much wrong by building infantry. So Infantry is suddenly a viable option again.
