Pokémon GO came out July 6th and people have been training all over the place (and sometimes getting themselves into trouble). Catching and evolving Pokémon continues to be great fun for trainers of all ages.
One Pokémon in particular has been the subject of speculation. Eevee can evolve into one of three Pokémon, and people have been trying to figure out what makes it choose one over the other.
There’s growing rumors that either:
(1) the moves an Eevee has will bias it towards one evolution over another, or
(2) the trainer’s starting Pokémon determines the Eevee outcome, or
(3) that it depends on the faction you join.
To unravel this mystery, G0mega on Reddit started a thread for players to enter their evolution data: starting Eevee moveset, starting trainer Pokémon , and Eevee outcome.
There are 7335 responses to date and the results are illuminating:
Overall, Vaporeon seems to be the most common at 36%, followed by Flareon at 33%, and Jolteon at 31%. However, this distribution may vary by (1) Eevee moves, (2) starter Pokemon, or (3) trainer faction. To tackle these three questions, I’m going to use a series of logistic regression1 models.
First up, the Eevee’s moves:
The moves that an Eevee starts with seems to have no discernible effect on the likelihood of evolving into the various evolutions. Jolteon is still the rarest, followed by Flareon and then Vaporeon.
Next, starting Pokémon:
The same seems to hold for starting Pokémon. Your choice at the beginning doesn’t seem to affect the Eevee outcome.
And lastly, player faction:
No effect here either: the color of your flag doesn’t affect the color of your Eevee.
So that’s it–maybe not very satisfying but it seems that Niantic most likely determines Eevee evolution according to some weighted probability function. No matter your Eevee’s moves, your starting Pokémon, or your faction, you’re most likely going to get a Vaporeon.
Thanks for reading!
Want more data visualization? Check out my other posts at: https://vizthis.wordpress.com/
1 The database contained username, and users who evolved multiple Eevees were able to enter data multiple times. Because the data from the same users was potentially non-independent, I used mixed-effects models.