def new_stage_data(action, portfolio, old_state, new_state, portfolio_value,
done, episode_data):
old_portfolio_value = portfolio_value
#low_price = new_state[2]
#changing code to use average price rather than normalized price
price = episodic_data.data_average_price(data_dict, episode_data)
next_price = episodic_data.data_average_price(data_dict, new_state)
#price = data_dict[episodic_data.list_md5_string_value(episode_data)][-1]
#next_price = data_dict[episodic_data.list_md5_string_value(new_state)][-1]
#buying
if action == 1:
#old_price = old_state[1]
portfolio_value -= price
portfolio += 1
#selling
elif action == 2:
#old_price = old_state[2]
portfolio_value += price
portfolio -= 1
elif action == 0:
portfolio = portfolio
#reward = 0
#if new_state:
new_state = new_state + [portfolio]
#if portfolio >= 0:
#low_price = new_state[2]
#else:
#low_price = new_state[1]
#reward system might need to change and require some good thinking
#if done:
reward = (portfolio_value + portfolio * next_price)
if reward > 0:
reward = 2 * reward #increasing reward
#pdb.set_trace();
return new_state, reward, done, portfolio, portfolio_value
def new_stage_data(action, portfolio, old_state, new_state, portfolio_value, done, episode_data):
old_portfolio_value = portfolio_value
#low_price = new_state[2]
#changing code to use average price rather than normalized price
price = episodic_data.data_average_price(data_dict, episode_data)
next_price = episodic_data.data_average_price(data_dict, new_state)
#price = data_dict[episodic_data.list_md5_string_value(episode_data)][-1]
#next_price = data_dict[episodic_data.list_md5_string_value(new_state)][-1]
#buying
if action == 1:
#old_price = old_state[1]
portfolio_value -= price
portfolio += 1
#selling
elif action == 2:
#old_price = old_state[2]
portfolio_value += price
portfolio -= 1
elif action == 0:
portfolio = portfolio
#reward = 0
#if new_state:
new_state = new_state + [portfolio]
#if portfolio >= 0:
#low_price = new_state[2]
#else:
#low_price = new_state[1]
#reward system might need to change and require some good thinking
#if done:
reward = (portfolio_value + portfolio * next_price)
if reward > 0:
reward = 2*reward #increasing reward
#pdb.set_trace();
return new_state, reward, done, portfolio, portfolio_value
def show_trader_path(actions, episode_data, portfolio_list, portfolio_value_list, reward_list):
i = 0
#print("Action, Average Price, Portfolio, Portfolio Value, Reward")
for index, action in enumerate(actions):
episode = episode_data[index]
action_name = action_map[actions[index]]
price = episodic_data.data_average_price(data_dict, episode)
portfolio = portfolio_list[index]
portfolio_value = portfolio_value_list[index]
i += 1
reward = reward_list[index]
#print(action_name, price, portfolio, portfolio_value, reward)
#print("last price:")
episode = episode_data[i]
last_price = episodic_data.data_average_price(data_dict, episode)
#print(last_price)
reward = (portfolio_value_list[-1] + portfolio_list[-1]*last_price)
return reward
def show_trader_path(actions, episode_data, portfolio_list,
portfolio_value_list, reward_list):
i = 0
#print("Action, Average Price, Portfolio, Portfolio Value, Reward")
for index, action in enumerate(actions):
episode = episode_data[index]
action_name = action_map[actions[index]]
price = episodic_data.data_average_price(data_dict, episode)
portfolio = portfolio_list[index]
portfolio_value = portfolio_value_list[index]
i += 1
reward = reward_list[index]
#print(action_name, price, portfolio, portfolio_value, reward)
#print("last price:")
episode = episode_data[i]
last_price = episodic_data.data_average_price(data_dict, episode)
#print(last_price)
reward = (portfolio_value_list[-1] + portfolio_list[-1] * last_price)
return reward