def determine_cascade_profile(zip_suffix,
number_of_games,
numberWE,
numberWF,
supply,
demand,
reserve=None):
"""
Given an initial number of games, number_of_games,
cascades back to the first time the profile in
file_location is found. This function iteratively
halfs number_of_games until a profile is found
OR we get below 200, in which case an error is thrown.
"""
while (number_of_games >= 100):
zip_location = setup.get_zip_location(zip_suffix, number_of_games)
zf = zipfile.ZipFile(zip_location)
file_location = setup.get_agent_dir_location(
number_of_games, supply, demand) + setup.get_file_location(
numberWE, numberWF, reserve)
if file_location in zf.namelist():
print(
'\tFound! =', number_of_games, ', (WE, WF) = (', numberWE, ',',
numberWF, ')',
(' reserve = ' + str(reserve) if reserve is not None else ''))
print('\t', zip_location)
return number_of_games
else:
print('\tNot found for number_of_games = ', number_of_games,
', trying half')
number_of_games = int(number_of_games / 2)
raise ValueError(
"******************* Could not find data for file_location ",
file_location, ", in zip ", zip_location)
def plot_all_soft_expected_agents_pure_nash(number_of_games):
"""
Wrapper to plot all expected number of strategies.
"""
for (demand, supply) in setup.get_grid_demand_impressions():
print('Expected Pure Nash for (demand, supply) = (', demand, ',',
supply, ')')
dir_location = setup.get_agent_dir_location(number_of_games, supply,
demand)
plot_soft_expected_agents_pure_nash(number_of_games, dir_location,
demand, supply)
def plot_all_proportion_pure_nash(number_of_games):
"""
Save all the best response graphs to an image and the
image of the proportion of pure nash
"""
for (demand, supply) in setup.get_grid_demand_impressions():
print('Proportion Graphs Pure Nash for (demand, supply) = (', demand,
',', supply, ')')
dir_location = setup.get_agent_dir_location(number_of_games, supply,
demand)
dict_of_pure_nash = mean_best_response_graphs.get_dict_of_pure_nash(
number_of_games, demand, supply, dir_location)
plot_proportion_pure_nash(number_of_games, demand, supply,
dict_of_pure_nash)
def produce_agents_plots(number_of_games, demand_factor, impressions):
"""
Plots varying the composition of the game, i.e., type and number of agents.
"""
image_prefix = 'demand-factor-' + demand_factor.replace(
'.', '_') + '-' + impressions + 'impressions-'
agent_dir_location = setup.get_agent_dir_location(number_of_games,
impressions,
demand_factor)
print('\t(games, demand, supply) = (', number_of_games, ',', demand_factor,
',', impressions, ')')
for x in setup.get_two_agents_combinations():
for y in [True, False]:
print('\t\tAgents Graph: ' + str(x) +
(', Fix Agent ' + str(x[0]) if y else ', Fix Agent ' +
str(x[1])))
plot_agent_group(number_of_games, agent_dir_location, image_prefix,
demand_factor, impressions, x[0], x[1], y)
def get_profile_revenue(zip_suffix,
number_of_games,
numberWE,
numberWF,
supply,
demand,
reserve=None):
"""
Given the description of a profile, returns the mean revenue of the market maker.
"""
number_of_games = deviation_analysis.determine_cascade_profile(
zip_suffix, number_of_games, numberWE, numberWF, supply, demand,
reserve)
zip_location = setup.get_zip_location(zip_suffix, number_of_games)
file_location = setup.get_agent_dir_location(
number_of_games, supply, demand) + setup.get_file_location(
numberWE, numberWF, reserve)
data = get_data.get_agent_data(zip_location, file_location)
data = data.dropna()
return get_data.mean_confidence_interval(data.wincost)
def get_specific_profile_data(zip_suffix,
numberWE,
numberWF,
specific_number_of_games,
supply,
demand,
reserve=None):
"""
Given numberWE, numberWF, specific_number_of_games, supply, demand;
produces the profile data for numberWE playing numberWF specific_number_of_games
times with supply and demand.
"""
zip_location = setup.get_zip_location(zip_suffix, specific_number_of_games)
dir_location = setup.get_agent_dir_location(specific_number_of_games,
supply, demand)
file_location = setup.get_file_location(numberWE, numberWF, reserve)
mix = get_data.get_agent_data(zip_location, dir_location + file_location)
mix = mix.dropna()
WEdata = mix[mix.agent.str.contains('WEAgent')]
WFdata = mix[mix.agent.str.contains('WFAgent')]
"""print('get_specific_profile_data (1) = ', file_name)
print('get_specific_profile_data (2) = ', dir_location)
print('get_specific_profile_data (3) = ', reserve)
print('get_specific_profile_data (4) = ', file_location)"""
return ('WE' * numberWE + 'WF' * numberWF, {
'WE': {
'n': len(WEdata),
'mean': WEdata.profit.mean(),
'var': WEdata.profit.var()
},
'WF': {
'n': len(WFdata),
'mean': WFdata.profit.mean(),
'var': WFdata.profit.var()
}
})