This is a document that describes an attempt at a target API for working with teams, setting up showdown bots that run using a specified configuration, and running genetic algorithms to optimize teams.
Thank you to github copilot for helping me write this document lmao
team = generate_team(format='gen8randombattle', seed=1234)
# a team object contains a list of 6 pokemon objects with their moves, items, etc.
# it's also tracking the (normalised) number of wins/losses/ties for this team.
# (would be nice to hard-code an auto-tie at 100 moves or something, but that's not a priority)
# here's an example env file for running a showdown bot:
# BATTLE_BOT=safest
# WEBSOCKET_URI=sim.psim.us:8000
# PS_USERNAME=whimsicaldreams
# PS_PASSWORD=password
# BOT_MODE=CHALLENGE_USER
# POKEMON_MODE=gen3ou
# RUN_COUNT=1
# USER_TO_CHALLENGE=...
# SAVE_REPLAY=False
# TEAM_NAME=gen3/my_teams
# here's an idealised Bot class that uses these same env variables to run a showdown bot, but more cleanly:
bot_1 = Bot(
team,
strategy='safest',
websocket_uri='sim.psim.us:8000',
user='whimsicaldreams',
password='password',
bot_mode='CHALLENGE_USER',
format='gen3ou',
num_battles=1,
save_replay=False,
)
# this will require hacking the bot code to remove dependence from environment variables, but it's doable
# note the absence of the user to challenge -- this is so we can be flexible about which bots are fighting at any given time (needs coding)
# the bot class will also need to be able to run a battle and return the results
# now we can run a battle between two bots:
bot_2 = Bot(
team,
strategy='safest',
websocket_uri='sim.psim.us:8000',
user='melonchomper',
password='password',
bot_mode='CHALLENGE_USER', # this would need to be ACCEPT_CHALLENGE
format='gen3ou',
num_battles=1,
save_replay=False,
)
# from a little google searching, it looks like we can run these bots in parallel using asyncio:
async def run_battle(bot_1, bot_2):
# individually run the showdown bots at the same time, and wait for the results
await asyncio.gather(bot_1.challenge(bot_2), bot_2.accept(bot_1))
# parse the results
# update the win/loss/tie counts for each team
...
run_battle(bot_1, bot_2)
# within this function, we'll need to:
# - run the bot_1 and bot_2 showdown bots
# - parse the results of the battle
# - update the win/loss/tie counts for each team
# here's a more generic example, where we specify a number of battles for each team, and then run them all:
bots = {f'bot{i}':Bot(team, strategy='safest', websocket_uri='sim.psim.us:8000', user=f'bot{i}', password='password', bot_mode='CHALLENGE_USER' if i % 2 == 0 else 'ACCEPT_CHALLENGE', format='gen3ou', num_battles=10, save_replay=False) for i in range(10)}
# num_battles = 10 would mean that bot would challenge the same bot 10 times, so we can make every bot fight every other bot 10 times
# now we can run all of these battles in parallel using a while loop that ensures every bot has fought every other bot:
# iterate over every combination of bots, ignoring equivalent combinations (e.g. bot_1 vs bot_2 is the same as bot_2 vs bot_1)
for bot_1, bot_2 in itertools.combinations(bots.values(), 2):
if bot_1 != bot_2:
try:
# make sure bot_1 is the challenger and bot_2 is the acceptor
bot_1.bot_mode = 'CHALLENGE_USER'
bot_2.bot_mode = 'ACCEPT_CHALLENGE'
# check if either bot is currently in a battle
if bot_1.in_battle or bot_2.in_battle:
print(f'Bot {bot_1} or {bot_2} is currently in a battle, waiting for them to finish...')
while bot_1.in_battle or bot_2.in_battle:
await asyncio.sleep(1) # hopefully this will wait for the battle to finish
# run the battle
bot_1.in_battle = True
bot_2.in_battle = True
run_battle(bot_1, bot_2) # this will do 10 battles between bot_1 and bot_2
bot_1.in_battle = False
bot_2.in_battle = False
except Exception as e:
print(e)
print(f'Error running battle between {bot_1} and {bot_2}')
# ensure we're only here once every bot has fought every other bot
while any(bot.in_battle for bot in bots.values()):
await asyncio.sleep(1)
# now we can (hopefully) print out the results of each bot:
for bot in bots:
print(bot.user, bot.team.wins, bot.team.losses, bot.team.ties)