Rudo (Brig:LoR) vs. Cloud of Darkness (FF3)
Rudo is *very close* to winning this, which is amazing for a PC. Immuning status is a good start, and having powerful counterattacks is handy when the main body starts out only able to attack (and thus walk into counters - I'd certainly force the AI to be "dumb" rather than pass their turn, and would be willing to take the 75% HP threshold guess a little lower in the DL). Probably open with Clean Military Rule'ing out the Left Tentacle and hoping to kill the main body with counterattacks + regular attacks before dying... just... even as tanky as Rudo is, idk if he can handle it. Turn 1: Blow up left tentacle, eat main body physical + right tentacle lightning (1800 FF3 damage), counterattack main body. Turn 2: Attack + Counterattack, 1800 more FF3 damage. Turn 3: One final attack, then surely die to damage (no matter how tanky Rudo is, there's a point where it gets silly). Kinda doubt 2 attacks + 2 counterattacks are enough to kill the main body (with the damage cut, that's like 1.9 PCHP or so (~2.1 PCHP pre-defense), and CoD is listed at around 2 PCHP). On the other hand, there's taking CoD at a higher level than 51 (I know "casual" guides suggest L60, and I did it around L54-55 myself), although then the argument to deny a Particle Beam on turn 2 gets tougher... then Rudo needs to survive 3800 FF3 damage (1.53 PCHP to the party average in the topic, maybe a little less at L55) with his 1.34 PCHP and 1.67 PDur, but a decent amount of that damage was magical (Particle Beam + Lightning). It's very close! There's also YOLO charge the main body stratz, although Protect is an issue then... I guess since the main body is locked into a counterattack'd physical on turn 1, that might work anyway, since Rudo can spend his big damage on turn 1 before Protect comes down. I'll vaguely give it to Rudo but could be convinced otherwise.
HEAVY
Edge (FF4) vs. Worker 8 (FFT)
Oh boy, interp split for if Image is "null next physical attack" or "perfect evasion against next physical attack." W8 loses if it's the first and wins if it's the second, I think.
EDIT: Eh, if worker 8's Work penetrates DEfense UP in-game, I guess I"ll tiebreak for it.
Ike (FEs) vs. Lyon (Suiko5)
Yeah no.
MIDDLE
Ashley Winchester (WA2) vs. Bartholomew Fatima (XG)
So the bad news for Bart is that Full Clip ignores evade (according to Pyro's stat topic). The bad news for Ashley is that Full Clip costs 75 FP. The very bad news is that attacking and missing *does not generate FP* according to Pyro's stat topic. So Bart goes to full evade, attacks Ashley once with the Cobra Cracka to apply Poison, waits for Ashley to hit 1 HP from poison damage (which doesn't generate FP), and then finishes, right? Any flaws with this plan?
Empoleon (PKMN) vs. Squall (FF8)
Looks like Hydro Pump -> Hydro Cannon should KO from the Gen 4 stat topic.
LIGHT
Thief (FF5) vs. Palmer (FF7)
Thief is bad, but... Palmer level bad? HP Plus materia was storebought in Cosmo Canyon, so you should have it by Rocket Town...
Paula (EB) vs. Geela (TriStrat)
I don't think this combo has votership, so going to boldly vote here despite my EB knowledge not being very deep. The good news for Paula: Geela can't exploit her awful durability. The bad news: below-average magic damage is not how you threaten an infinite healer with above-average MDef. I think Paula 3HKOs here if she spams PSI Freeze Omega, so she needs to chain 2 Solidifies in a row OR get 1 Solidify right next to a very rare double. That's a 1/16 chance, but Paula only has 9 Freeze Omegas, so... not likely enough IMO, unless you hugely respect mildly above average EB speed.
Natan (SH3) vs. Aika (SoA)
Aika in Light?! RIP Light. I don't see a Confuse blocker on the SH3 list, so free win for Aika...
Pikachu (PKMN) vs. Locke (FF6)
Gonna assume Locke still 2HKOs with a Wing Edge, so he can avoid the chance of a Static proc that way. 'Chu's damage is average in the Gen6 stat topic, so despite going first, FF6 characters have a pretty flat HP curve, so assuming Locke won't be 2HKO'd. Pikachu can still win with Paralyze hax but nah.