Research What would happen if every pokemon battled every other pokemon in a full round-robin? I wrote a program to find out.

[relevartemit]

I've always enjoyed looking at the stats and data in Pokemon, and a while back, I had the idea to design a way to determine who would be expected to win a matchup between any two pokemon. Based only on stats and typing, and using a few assumptions about movesets and battle mechanics, this is possible. Then it's just a matter of running this code for every matchup that currently exists, and we can use the total win rate to get a ranking of every single pokemon from best to worst... at least in this very simplified view. I even sorted them into tiers.

The link to the Google spreadsheet containing the results is here. You can sort the data or download a copy if you want to mess around with it. It shows every pokemon's base stats, typing, win rate, tier, best win, and worst loss. The rules for the simulation are in the document; basically, I ignore items and abilities and assume everyone is constantly using raw-damage moves of either of their two types, or a non-STAB attack for neutral damage if that is more effective. Obviously, this makes it totally useless for competitive play, but it's not meant for that. It's just a somewhat-sophisticated way to see how they all look on paper.

I don't know if anyone else will enjoy this, but I think it's cool to see some of the patterns that show up. It's mostly what you expect, but there are some pokemon that might be a little surprising. Hope you like it!

 

[R_N]

Rank 4: Regigigas is hilarious to me.

Obviously with the ability ignored, it would be a very strong pokemon, but to rank 4th amid all the other 670+ legends? Its really impressive and not something I'd have expected.

And the first "normal" pokemon on this is...rank 20 Garchomp. Yeah that tracks.

e: how on earth did Happiny lose to Cosmog oh my god

e2: COSMOEM BEAT HOW MANY POKEMON????

e3: Ah okay I didnt see you gave EVERY pokemon a set of attacking moves. Misinterpreted that ruling

 

[Hugin]

Genesect at 37 is wild to me. That feels about correct; depending on how you count, there's 30-40 mons in Ubers currently, so Genesect being pretty low on ubers viability means there's 30ish mons that are better than it, which is close enough to 37. But literally everything that makes Genesect good(ability, coverage, ESpeed, boosting moves, UTurn) is ignored by this program. It got the right answer with completely the wrong inputs.

 

[Ironmage]

Genesect at 37 is wild to me. That feels about correct; depending on how you count, there's 30-40 mons in Ubers currently, so Genesect being pretty low on ubers viability means there's 30ish mons that are better than it, which is close enough to 37. But literally everything that makes Genesect good(ability, coverage, ESpeed, boosting moves, UTurn) is ignored by this program. It got the right answer with completely the wrong inputs.

I think part of this is the algorithm overvaluing the Steel type. The 90 power used is stronger than Iron Head and Flash cannon, while the lack of super-effective coverage means that it doesn't care how common moves like Earthquake are, but it still reads the high number of resistances. On the high end, we have dialga at the top and on the low end it takes until the top of D tier before there are any steel-types (meltan at 644).

(If you're curious, the lowest-placing fully-evolved Steel type is Mawile at 408, and the top placing non-legendary steel is Excadrill at 53)