Gamorium's take

Fighting-game execution and opponent prediction across an enormous, loosely balanced roster.
the cheat-test profile
Each axis is rated on its own - how far one cheat would trivialize the game here. They don't trade off; low on all four means the game is an experience, not a contest.
cheat: Aimbot - perfect inputs
Combo strings and kombat timing decide exchanges
cheat: Out-predict an adaptive opponent
Versus humans, baiting and punishing habits is the meta
cheat: See what's hidden / map the RNG
No hidden information once movesets are known
cheat: A perfect solver / coach
A 62-character roster of matchup knowledge to internalise
Each axis is named by the cheat that would break the game along it. Execution is the hands. Reads is out-thinking an adaptive opponent. Variance is hidden information & RNG. Optimization is the solvable systems layer. A game's region is the top one or two axes.
These four aren't one rating split four ways - they're independent 0-10 questions. A 9 on one axis takes nothing from the others, and a game low on all four isn't missing data: it's telling you there's no skill here to cheat.
The basis is how the majority actually plays this game - not a niche mode or a personal run. For a game with very different modes (a story campaign versus ranked multiplayer), that's the version most people mean.
Ranked by shared genre first, then by how closely the cheat-test profile matches - a cross-genre tag means a pick matched on skill profile alone.

Mortal Kombat vs. DC Universe
Competitive PvP · crossover fighting game

Dragon Ball Z: Budokai Tenkaichi 3
Competitive PvP · arena fighting game

Marvel vs. Capcom: Clash of Super Heroes
Competitive PvP · tag fighting game