class TradingEnv(gym.Env):
def __init__(self,
data_source,
look_backward_window=1,
mode="rl",
starting_cash={"STOCK": 1000000},
commission_multiplier=1,
min_commission=5,
tax_multiplier=1) -> None:
self.look_backward_window = look_backward_window
self.mode = mode
self.starting_cash = starting_cash
self.commission_multiplier = commission_multiplier
self.min_commission = min_commission
self.tax_multiplier = tax_multiplier
self._context = Context(look_backward_window=look_backward_window)
#set data_source
self._context.set_data_source(data_source)
#set event_source
event_source = SimulationEventSource()
self._context.set_event_source(event_source)
#set broker
broker = SimulationBroker(self._context)
self._context.set_broker(broker)
# transaction_cost decider
transaction_cost_decider = CNStockTransactionCostDecider(
commission_multiplier=commission_multiplier,
min_commission=min_commission,
tax_multiplier=tax_multiplier)
self._context.set_transaction_cost_decider("CS",
transaction_cost_decider)
# setup account and portfolio
portfolio = Portfolio(starting_cash=starting_cash, init_positions={})
self._context.set_portfolio(portfolio)
# setup a tracker to record key info
tracker = Tracker(self._context)
self._context.set_tracker(tracker)
self._context.event_bus.publish_event(Event(EVENT.POST_SYSTEM_INIT))
#setUP executor
if mode == "rl":
self._executor = RLExecutor(self._context)
else:
self._executor = Executor(self._context)
# user strategy
user_strategy = Strategy(self._context)
def reset(self):
state = Context.get_instance().history_bars()
return state
def step(self, action):
reward, is_done, info = self._executor.send(action)
#pdb.set_trace()
next_state = Context.get_instance().history_bars()
return next_state, reward, is_done, info
@property
def trading_dt(self):
return self._context.trading_dt
def render(self):
if self.mode == "rl":
returns_list = self._context.tracker._portfolio_forward_bar_returns.copy(
)
returns_list.insert(0, 0)
else:
returns_list = self._context.tracker._portfolio_current_bar_returns.copy(
)
#pdb.set_trace()
index = self._context.availabel_trading_dts
returns = pd.DataFrame(returns_list,
index=index,
columns=["unit_net_value"])
unit_net_value = (returns + 1).cumprod()
plot_performance(unit_net_value)
info["profit_and_loss"] = tracker._portfolio_forward_bar_pnl[-1]
if self._context.trading_dt == self.available_trading_dts[-1]:
is_done = True
else:
is_done = False
return reward, is_done, info
if __name__ == "__main__":
from sharpe.utils.mock_data import create_toy_feature
from sharpe.data.data_source import DataSource
from sharpe.core.context import Context
feature_df, price_s = create_toy_feature(order_book_ids_number=2,
feature_number=3)
data_source = DataSource(feature_df=feature_df, price_s=price_s)
#
context = Context(look_backward_window=2)
context.set_data_source(data_source)
#
from sharpe.mod.event_source import SimulationEventSource
default_event_source = SimulationEventSource()
context.set_event_source(default_event_source)
executor = Executor(context)
executor.send(action=0)
class TradingEnv(gym.Env):
def __init__(self,
data_source,
look_backward_window=1,
mode="non-rl",
starting_cash={"STOCK": 1000000},
commission_multiplier=1,
min_commission=5,
tax_multiplier=1) -> None:
self.look_backward_window = look_backward_window
self.mode = mode
self.starting_cash = starting_cash
self.commission_multiplier = commission_multiplier
self.min_commission = min_commission
self.tax_multiplier = tax_multiplier
self._context = Context(look_backward_window=look_backward_window)
#set mode
self._context.set_mode(mode)
#set data_source
self._context.set_data_source(data_source)
#set event_source
event_source = SimulationEventSource()
self._context.set_event_source(event_source)
#set broker
broker = SimulationBroker(self._context)
self._context.set_broker(broker)
# transaction_cost decider
transaction_cost_decider = CNStockTransactionCostDecider(
commission_multiplier=commission_multiplier,
min_commission=min_commission,
tax_multiplier=tax_multiplier)
self._context.set_transaction_cost_decider("CS",
transaction_cost_decider)
# setup account and portfolio
portfolio = Portfolio(starting_cash=starting_cash, init_positions={})
self._context.set_portfolio(portfolio)
#setUP executor
if mode == "rl":
self._executor = RLExecutor(self._context)
else:
self._executor = Executor(self._context)
# user strategy
user_strategy = Strategy(self._context)
# setup a tracker to record key info
tracker = Tracker(self._context)
self._context.set_tracker(tracker)
self._context.event_bus.publish_event(Event(EVENT.POST_SYSTEM_INIT))
self._context.update_time(
calendar_dt=self._context.available_trading_dts[0],
trading_dt=self._context.available_trading_dts[0])
# action and observation space
self.action_space = gym.spaces.Box(
0,
1,
shape=(len(data_source.order_book_ids_index), ),
dtype=np.float32) # include cash
# get the state space from the data min and max
if look_backward_window == 1:
self.observation_space = gym.spaces.Box(
low=-np.inf,
high=np.inf,
shape=(len(data_source.order_book_ids_index),
len(data_source.feature_list)),
dtype=np.float32)
else:
self.observation_space = gym.spaces.Box(
low=-np.inf,
high=np.inf,
shape=(len(data_source.order_book_ids_index),
look_backward_window, len(data_source.feature_list)),
dtype=np.float32)
def reset(self):
self._context.update_time(
calendar_dt=self._context.available_trading_dts[0],
trading_dt=self._context.available_trading_dts[0])
state = Context.get_instance().history_bars()
return state
def step(self, action):
reward, is_done, info = self._executor.send(action)
#pdb.set_trace()
next_state = Context.get_instance().history_bars()
return next_state, reward, is_done, info
@property
def trading_dt(self):
return self._context.trading_dt
@property
def available_trading_dts(self):
return self._context.available_trading_dts
def render(self, auto_open=True):
if self.mode == "rl":
returns_list = self._context.tracker._portfolio_forward_bar_returns.copy(
)
returns_list.insert(0, 0)
else:
returns_list = self._context.tracker._portfolio_current_bar_returns.copy(
)
#pdb.set_trace()
index = self._context.available_trading_dts
returns = pd.DataFrame(returns_list,
index=index,
columns=["unit_net_value"])
unit_net_value = (returns + 1).cumprod()
plot_performance(unit_net_value, auto_open=auto_open)