An agent-environment based backtesting framework.
class BuyAndHold(BaseStrategy):
def __init__(self, engine, broker, account):
super().__init__(engine, broker, account)
def before_trading_end(self):
if self.current_date == self.trade_dates[0]:
self.account.order_target_pct_to(pd.Series(1, index=["000001"]))
class BuyAndHold(BaseAgent):
def __init__(self, data):
super().__init__(data)
self.i = -1
def take_action(self, state, date):
self.buffer.append(state)
self.i += 1
if self.i == 0:
target_positions = np.array([0.2 for i in range(5)])
return target_positions
else:
return NO_ACTION
while True:
try:
print(f"current date is {env.current_date}")
action = agent.take_action(state, env.current_date)
next_state, reward, truncated, terminated, info = env.step(action)
state = next_state
returns.append(reward)
except IndexError:
break