def _reset(self, t, i):
self.orderbook = copy.deepcopy(self.orderbookOriginal) # TODO: Slow but currently required to reset after every episode due to change of order book states during matching
orderbookState, orderbookIndex = self._get_random_orderbook_state()
bidAskFeature = self._makeFeature(orderbookIndex=orderbookIndex)
state = ActionState(t, i, {'bidask': bidAskFeature}) #np.array([[t, i]])
self.execution = None
self.orderbookIndex = orderbookIndex
self.actionState = state
return state.toArray()
def step(self, action):
action = self.levels[action]
if self.execution is None:
self.execution = self._create_execution(action)
else:
self.execution = self._update_execution(self.execution, action)
logging.info(
'Created/Updated execution.' +
'\nAction: ' + str(action) + ' (' + str(self.execution.getOrder().getType()) + ')' +
'\nt: ' + str(self.actionState.getT()) +
'\nruntime: ' + str(self.execution.getRuntime()) +
'\ni: ' + str(self.actionState.getI())
)
self.execution, counterTrades = self.execution.run(self.orderbook)
i_next = self._determine_next_inventory(self.execution)
t_next = self._determine_next_time(self.execution.getState().getT())
bidAskFeature = self._makeFeature(orderbookIndex=self.execution.getOrderbookIndex())
state_next = ActionState(t_next, i_next, {'bidask': bidAskFeature})
done = self.execution.isFilled() or state_next.getI() == 0
# if done == True:
# #reward = self.execution.getReward()
# #volumeRatio = 1.0
# else:
reward, volumeRatio = self.execution.calculateRewardWeighted(counterTrades, self.I[-1])
logging.info(
'Run execution.' +
'\nTrades: ' + str(len(counterTrades)) +
'\nReward: ' + str(reward) + ' (Ratio: ' + str(volumeRatio) + ')' +
'\nDone: ' + str(done)
)
self.orderbookIndex = self.execution.getOrderbookIndex()
self.actionState = state_next
return state_next.toArray(), reward, done, {}