"""
Functions for adding reward values
"""
import pandas as pd
# from utils import *
#####################################
### Opponent previous move counts ###
#####################################
def get_opponent_move(sub_df):
"""
fills in the `opponent_move` column
"""
for i in range(len(sub_df)):
if i%2 == 0:
sub_df.at[i, 'opponent_move'] = sub_df.at[i + 1, 'player_move']
else:
sub_df.at[i, 'opponent_move'] = sub_df.at[i - 1, 'player_move']
return sub_df
#########################
### human reward move ###
#########################
def human_reward_move(sub_df):
"""
tally human rewards on its rock, paper, or scissors moves.
3 points for a win, 0 for a tie, -1 for a loss
"""
# reward dictionary
dic_reward={'win':3,'tie':0,'loss':-1}
# initialization reward of each game
dic_move={'rock':0,'paper':0,'scissors':0}
# loop through the human rows
for i in range(0,len(sub_df),2):
# fetch player outcomes
outcome=sub_df.get('player_outcome').iloc[i]
# fetch player moves
move=sub_df.get('player_move').iloc[i]
# avoid all the nans and 'none's in player_move
if move != 'none'and not pd.isna(move):
# tally all the point associated with rewards
dic_move[move]+=dic_reward[outcome]
# get the correspounding move's reward column name
col_name=move+'_reward'
# store reward points in the reward column
sub_df.at[i,'rock_reward']=dic_move['rock']
sub_df.at[i,'paper_reward']=dic_move['paper']
sub_df.at[i,'scissors_reward']=dic_move['scissors']
return sub_df
#############################################
### human reward on past and current move ###
#############################################
def human_reward_past_cur_move(sub_df):
"""
tally human reward based on its past move and current move.
"""
dic_reward={'win':3,'tie':0,'loss':-1}
# initialize dic_move to store reward for each combination
dic_move={'rock':{'rock':0,'paper':0,'scissors':0},'paper':{'rock':0,'paper':0,'scissors':0},
'scissors':{'rock':0,'paper':0,'scissors':0}}
for i in range(2,len(sub_df),2):
# human current outcome
outcome=sub_df.get('player_outcome').iloc[i]
# human previous move
pre_move=sub_df.get('player_move').iloc[i-2]
# human current move
cur_move=sub_df.get('player_move').iloc[i]
# ignore Nan cells
if pre_move != 'none'and not pd.isna(pre_move) and\
cur_move != 'none'and not pd.isna(cur_move):
# get reward for move combination of current round
dic_move[pre_move][cur_move]+=dic_reward[outcome]
# fill in reward
sub_df.at[i,'rock_rock_reward']=dic_move['rock']['rock']
sub_df.at[i,'rock_paper_reward']=dic_move['rock']['paper']
sub_df.at[i,'rock_scissors_reward']=dic_move['rock']['scissors']
sub_df.at[i,'paper_rock_reward']=dic_move['paper']['rock']
sub_df.at[i,'paper_paper_reward']=dic_move['paper']['paper']
sub_df.at[i,'paper_scissors_reward']=dic_move['paper']['scissors']
sub_df.at[i,'scissors_rock_reward']=dic_move['scissors']['rock']
sub_df.at[i,'scissors_paper_reward']=dic_move['scissors']['paper']
sub_df.at[i,'scissors_scissors_reward']=dic_move['scissors']['scissors']
return sub_df
###############################################################
### human reward on opponent past move and own current move ###
###############################################################
def human_reward_oppo_past_cur_move(sub_df):
"""
tally rewards based on the combination of opponent's past move and
human current move
"""
dic_reward={'win':3,'tie':0,'loss':-1}
# initialize dic_move to store reward of move combination
dic_move={'rock':{'rock':0,'paper':0,'scissors':0},'paper':{'rock':0,'paper':0,'scissors':0},
'scissors':{'rock':0,'paper':0,'scissors':0}}
for i in range(2,len(sub_df),2):
# outcome of current round
outcome=sub_df.get('player_outcome').iloc[i]
# opponent previous move
oppo_pre_move=sub_df.get('player_move').iloc[i-1]
# human current move
cur_move=sub_df.get('player_move').iloc[i]
# get rid of Nan cells
if oppo_pre_move != 'none'and not pd.isna(oppo_pre_move) \
and cur_move != 'none'and not pd.isna(cur_move):
# get reward of move combination
dic_move[oppo_pre_move][cur_move]+=dic_reward[outcome]
# fill in reward value
sub_df.at[i,'opponent_rock_rock_reward']=dic_move['rock']['rock']
sub_df.at[i,'opponent_rock_paper_reward']=dic_move['rock']['paper']
sub_df.at[i,'opponent_rock_scissors_reward']=dic_move['rock']['scissors']
sub_df.at[i,'opponent_paper_rock_reward']=dic_move['paper']['rock']
sub_df.at[i,'opponent_paper_paper_reward']=dic_move['paper']['paper']
sub_df.at[i,'opponent_paper_scissors_reward']=dic_move['paper']['scissors']
sub_df.at[i,'opponent_scissors_rock_reward']=dic_move['scissors']['rock']
sub_df.at[i,'opponent_scissors_paper_reward']=dic_move['scissors']['paper']
sub_df.at[i,'opponent_scissors_scissors_reward']=dic_move['scissors']['scissors']
return sub_df
def oppo_past_human_past_cur_move(sub_df):
"""
reward based on opponent past move, human past move and human current move.
"""
name=['rock','paper','scissors']
cols=[f'opponent_{oppo}_{past}_{cur}_reward'for oppo in name for past in name for cur in name]
sub_df=add_col(sub_df, cols, value =0)
dic_reward={'win':3,'tie':0,'loss':-1}
for i in range(2,len(sub_df),2):
outcome=sub_df.get('player_outcome').iloc[i]
oppo_pre_move=sub_df.get('player_move').iloc[i-1]
agent_pre_move=sub_df.get('player_move').iloc[i-2]
cur_move=sub_df.get('player_move').iloc[i]
if not pd.isna(oppo_pre_move) and not pd.isna(agent_pre_move) and not pd.isna(cur_move):
sub_df.loc[i,cols]=sub_df.loc[i-2,cols]
col_name='opponent_'+oppo_pre_move+'_'+agent_pre_move+'_'+cur_move+'_reward'
sub_df.loc[i,col_name]+=dic_reward[outcome]
return sub_df
