/Pokemon-Showdown-Win-Rate-Prediction

Primary LanguagePythonMIT LicenseMIT

Pokemon Showdown Win Rate Prediction

This is an simple model predicts win-rate for pokemon (showdown) matches.
WORK IN PROGRESS (including this readme)

Requirements ( latest should work )

  • python 3.6
  • tensorflow
  • tensorlayer
  • termcolor (used to print out pretty logs)

Usage (v2)

Get match data

python spider.py [tierName]

e.g.

python spider.py gen7vgc2019sunseries

Convert raw data to embedded data

python convertRaw.py [filename] (-max maxSpecies)

e.g.

python convertRaw.py gen7ou.txt

Train the model

python trainv2.py [formatPrefix (data file name without .txt)]

Please use -h option to print the help message for more options.
Note: -fmt & -load is not working now.

Notes

Currently accuracy is about 61.9% (best performance).
maxpool is better than summation (might for classification problems).