def run(start_date, end_date, file_name):
global file_handle, model
try:
file_handle = open(file_name, "w")
if model is None:
model = prepare_model() # Creating an ML-model.
alg = TNG(start_date, end_date
) # Creating an instance of environment to run algorithm in.
alg.addInstrument(TICKER) # Adding an instrument.
alg.addTimeframe(TICKER, TIMEFRAME) # Adding a time frame.
alg.run_backtest(on_bar) # Backtesting...
del alg
except IOError:
sys.stderr.write("Error while running a backtest for %s.\n" %
(file_name))
else:
sys.stderr.write("File %s done.\n" % (file_name))
if file_handle is not None:
file_handle.close()
file_handle = None
return model
from datetime import datetime
from tradingene.algorithm_backtest.tng import TNG
from tradingene.backtest_statistics import backtest_statistics as bs
name = "example1"
regime = "SP"
start_date = datetime(2018, 1, 1)
end_date = datetime(2018, 5, 1)
timeframe = 1440
alg = TNG(name, regime, start_date, end_date)
alg.addInstrument("btcusd")
alg.addTimeframe("btcusd", timeframe)
positionId = None
def onBar(instrument):
global positionId
if positionId is None:
if instrument.macd().histogram[2] > instrument.macd().histogram[1]:
positionId = alg.openLong(1)
if instrument.macd().histogram[2] < instrument.macd().histogram[1]:
positionId = alg.openShort(1)
else:
if instrument.macd().histogram[2] > instrument.macd().histogram[1] and\
alg.getPositionSide(positionId) <=0:
alg.closePosition(positionId)
positionId = alg.openLong(1)
if instrument.macd().histogram[2] < instrument.macd().histogram[1] and \
alg.getPositionSide(positionId) >= 0:
alg.closePosition(positionId)
return 1
elif data['close'][lookforward - 1] * 1.01 < data['open'][0]:
return -1
else:
return 0
def onBar(instrument):
inp = calculate_input(
instrument.rates[1:lookback + 1]) # Calculating inputs
prediction = model.predict([inp])[0] # Making prediction
if prediction > 0:
alg.buy()
elif prediction < 0:
alg.sell()
# end of onBar()
model = prepare_model() # Creating an ML-model.
alg = TNG(end_train_date, end_test_date) # Creating an instance of environment to run algorithm in.
alg.addInstrument(ticker) # Adding an instrument.
alg.addTimeframe(ticker, timeframe) # Adding a time frame.
alg.run_backtest(onBar) # Backtesting...
stat = bs.BacktestStatistics(alg) # Retrieving statistics of the backtest
pnl = stat.calculate_PnL()
num_positions = stat.calculate_number_of_trades()
print("pnl=%f, num_positions=%d" % (pnl, num_positions))
stat.backtest_results() # Displaying the backtest statistics
'open'] * THRESHOLD # Setting the take profit as big as the price threshold
sl = tp # Setting the stop loss equal to the take profit.
alg.setTP(profit=tp)
alg.setSL(loss=sl)
elif prediction == 0: # Class "0" predicts price to fall down
alg.sell()
tp = instrument.rates[0][
'open'] * THRESHOLD # Setting the take profit as big as the price threshold
sl = tp # Setting the stop loss equal to the take profit.
alg.setTP(profit=tp)
alg.setSL(loss=sl)
# end of onBar()
model = prepare_model() # Creating an ML-model.
alg = TNG(
END_TRAIN_DATE,
END_TEST_DATE) # Creating an instance of environment to run algorithm in.
alg.addInstrument(TICKER) # Adding an instrument.
alg.addTimeframe(TICKER, TIMEFRAME) # Adding a time frame.
alg.run_backtest(onBar) # Backtesting...
stat = bs.BacktestStatistics(alg) # Retrieving statistics of the backtest
pnl = stat.calculate_PnL()
num_positions = stat.calculate_number_of_trades()
print("pnl=%f, num_positions=%d" % (pnl, num_positions))
stat.backtest_results() # Displaying the backtest statistics
def calculate_output(data):
return (data['close'][lookforward - 1] - data['open'][0]) / data['open'][0]
def onBar(instrument):
inp = calculate_input(instrument.rates[1:lookback + 1])
prediction1 = model1.predict([inp])
prediction2 = model2.predict([inp])
prediction3 = model3.predict([inp])
if prediction1 > 0 and prediction2 > 0 and prediction3 > 0: # If market rising is predicted...
alg.buy()
elif prediction1 < 0 and prediction2 < 0 and prediction3 < 0: # If market falling is predicted...
alg.sell()
model1, model2, model3 = prepare_model()
alg = TNG(end_train_date, end_test_date)
alg.addInstrument(ticker)
alg.addTimeframe(ticker, timeframe)
alg.run_backtest(onBar)
stat = bs.BacktestStatistics(alg) # Retrieving statistics of the backtest
pnl = stat.calculate_PnL() # Retrieving the PnL.
num_positions = stat.calculate_number_of_trades(
) # Retrieving the number of trades done.
print("pnl=%f, num_positions=%d" % (pnl, num_positions))
stat.backtest_results() # Displaying the backtest statistics
def calculate_output(data):
return (data['close'][lookforward - 1] - data['open'][0]) / data['open'][0]
def onBar(instrument):
inp = calculate_input(instrument.rates[1:lookback + 1])
inp = scaler.transform(inp)
prediction = model.predict([inp])
if prediction > 0: # If market rising is predicted...
alg.buy()
elif prediction < 0: # If market falling is predicted...
alg.sell()
model = prepare_model()
alg = TNG(end_train_date, end_test_date)
alg.addInstrument(ticker) # Adding an instrument.
alg.addTimeframe(ticker, timeframe) # Adding a time frame.
alg.run_backtest(onBar) # Backtesting...
stat = bs.BacktestStatistics(alg) # Retrieving statistics of the backtest
pnl = stat.calculate_PnL() # Retrieving the PnL.
num_positions = stat.calculate_number_of_trades(
) # Retrieving the number of trades done.
print("pnl=%f, num_positions=%d" % (pnl, num_positions))
stat.backtest_results() # Displaying the backtest statistics
end_date = datetime(2018, 2, 1)
ticker = "btcusd"
timeframe = 60
def onBar(instrument):
bollinger10 = instrument.bollinger(
10) # Retrieving the value of the Bollinger indicator of period 10.
if instrument.close[1] > bollinger10.top[
1]: # If the price soars above the Bollinger top line...
alg.buy()
elif instrument.close[1] < bollinger10.bottom[
1]: # If the price falls below the Bollinger bottom line...
alg.sell()
alg = TNG(
start_date, end_date
) # Creating an instance of the class (TNG) to run the algorithm within.
alg.addInstrument(ticker) # Adding an instrument.
alg.addTimeframe(ticker, timeframe) # Adding a time frame.
alg.run_backtest(onBar) # Backtesting...
stat = bs.BacktestStatistics(alg) # Retrieving statistics of the backtest
pnl = stat.calculate_PnL()
num_positions = stat.calculate_number_of_trades()
print("pnl=%f, num_positions=%d" % (pnl, num_positions))
stat.backtest_results() # Displaying the backtest statistics