Tutorial
Here we step-by-step test an MACD strategy with stop loss based on ATR and 2 stages take profit.
The strategy was found here:
Creating a new strategy file
cipher new macd
cat macd.py
from cipher import Cipher, Strategy
class MacdStrategy(Strategy):
pass
def main():
cipher = Cipher()
cipher.add_source('binance_spot_ohlc', symbol='BTCUSDT', interval='1h')
cipher.set_strategy(MacdStrategy())
cipher.run(start_ts='2020-01-01', stop_ts='2020-02-01')
print(cipher.sessions)
print(cipher.stats)
cipher.plot()
if __name__ == '__main__':
main()
Let's use hourly DOGEUSDT ohlc from GateIO:
cipher.add_source('gateio_spot_ohlc', symbol='DOGE_USDT', interval='1h')
And run the backtest:
python macd.py
Adding indicators
We have to add MACD and ATR indicators. To achieve this, the compose method has to be edited.
One way to experiment with indicators is to start an interactive session inside compose (create a breakpoint):
class MacdStrategy(Strategy):
def compose(self):
df = self.datas.df
import pdb; pdb.set_trace()
return df
You can also use your IDE to set breakpoints, see a PyCharm example.
Then try to type:
df.ta.macd()
# returns a dataframe with next columns: MACD_12_26_9, MACDh_12_26_9, MACDs_12_26_9
If you want to know the indicator parameters, use help in a Python REPL:
import pandas_ta as ta
help(ta.macd)
# macd(close, fast=None, slow=None, signal=None, talib=None, offset=None, **kwargs)
# Moving Average Convergence Divergence (MACD)
# ...
In order to see the MACD, we have to add it to the dataframe and add another row to plot:
class MacdStrategy(Strategy):
def compose(self):
df = self.datas.df
return df.merge(df.ta.macd(), left_index=True, right_index=True)
def main():
# ...
cipher.plot(rows=[['ohlc'], ['MACD_12_26_9', 'MACDh_12_26_9', 'MACDs_12_26_9']])
We don't need the histogram, so MACDh_12_26_9 can be excluded, and let's rename
the remaining MACD columns to macd and macds (signal), and add ATR.
class MacdStrategy(Strategy):
def compose(self):
df = self.datas.df
macd_df = df.ta.macd()
df["macd"] = macd_df["MACD_12_26_9"]
df["macds"] = macd_df["MACDs_12_26_9"]
df["atr"] = df.ta.atr()
return df
def main():
cipher = Cipher()
cipher.add_source("gateio_spot_ohlc", symbol="DOGE_USDT", interval="1h")
cipher.set_strategy(MacdStrategy())
cipher.run(start_ts="2020-01-01", stop_ts="2020-02-01")
cipher.plot(rows=[["ohlc"], ["macd", "macds"]])
Signals
The entry signal for long: macd crosses macds from below when macds < 0.
Add the entry signal and the signals row to plot:
class MacdStrategy(Strategy):
def compose(self):
df = self.datas.df
macd_df = df.ta.macd()
df["macd"] = macd_df["MACD_12_26_9"]
df["macds"] = macd_df["MACDs_12_26_9"]
difference = df["macds"] - df["macd"]
cross = np.sign(difference.shift(1)) != np.sign(difference)
df["entry"] = cross & (difference < 0) & (df["macds"] < 0)
df["atr"] = df.ta.atr()
return df
def main():
cipher = Cipher()
cipher.add_source("gateio_spot_ohlc", symbol="DOGE_USDT", interval="1h")
cipher.set_strategy(MacdStrategy())
cipher.run(start_ts="2020-01-01", stop_ts="2020-02-01")
cipher.plot(rows=[["ohlc"], ["signals"], ["macd", "macds"]])
Position manipulation
import numpy as np
from cipher import Cipher, Session, Strategy, quote
class MacdStrategy(Strategy):
def compose(self):
df = self.datas.df
macd_df = df.ta.macd()
df["macd"] = macd_df["MACD_12_26_9"]
df["macds"] = macd_df["MACDs_12_26_9"]
difference = df["macds"] - df["macd"]
cross = np.sign(difference.shift(1)) != np.sign(difference)
df["entry"] = cross & (difference < 0) & (df["macds"] < 0)
df["atr"] = df.ta.atr()
df["atr_stop_loss"] = df["close"] - (df["atr"] * 1.5)
df["atr_take_profit"] = df["close"] + (df["atr"] * 1.5)
return df
def on_entry(self, row: dict, session: Session):
# add DOGE worth 100USD
session.position = quote(100)
# set brackets
session.stop_loss = row["atr_stop_loss"]
session.take_profit = row["atr_take_profit"]
# remember the next_stop loss and take_profit for the session
session.meta["next_take_profit"] = (
row["close"] + (row["atr_take_profit"] - row["close"]) * 2
)
session.meta["next_stop_loss"] = row["close"]
def on_take_profit(self, row: dict, session: Session):
if session.meta["next_take_profit"]:
session.position *= 0.5
session.take_profit = session.meta["next_take_profit"]
session.meta["next_take_profit"] = None
session.stop_loss = session.meta["next_stop_loss"]
else:
session.position = 0
def main():
cipher = Cipher()
cipher.add_source("gateio_spot_ohlc", symbol="DOGE_USDT", interval="1h")
cipher.set_strategy(MacdStrategy())
cipher.run(start_ts="2020-01-01", stop_ts="2020-02-01")
cipher.plot(
rows=[
["ohlc", "atr_stop_loss", "atr_take_profit"],
["signals"],
["macd", "macds"],
["position"],
["balance"],
]
)
print(cipher.sessions)
print(cipher.stats)
if __name__ == "__main__":
main()
Session Period PnL
--------------------- -------- ---------
long 2020-01-03 00:00 1h -1.35068
long 2020-01-03 05:00 2d 21h 2.65973
long 2020-01-09 09:00 1d 1h 2.51076
long 2020-01-10 11:00 1d 13h 0.787893
long 2020-01-13 12:00 1h -0.678198
long 2020-01-14 01:00 1h -0.916931
long 2020-01-16 19:00 15h 2.21554
long 2020-01-20 02:00 1d 17h -1.82047
long 2020-01-21 03:00 16h -1.11205
long 2020-01-22 07:00 15h -1.10829
long 2020-01-22 13:00 10h -0.977476
long 2020-01-24 00:00 2h -1.09632
long 2020-01-24 04:00 20h 0.706303
long 2020-01-25 16:00 1d 1h 1.70895
long 2020-01-30 11:00 5h -1.09761
---------------- --------------------- ------
start 2020-01-02 09:00
stop 2020-01-31 23:00
period 29d 14h
trades 15
longs 15 100.0%
shorts 0
period median 15h
period max 2d 21h
success 6 40.0%
success median 1.9622442059571628645
success max 2.659732540861812779
success row 3
failure 9 60.0%
failure median 1.097612065354000837
failure max 1.820469798657718121
failure row 5
spf 0.6666666666666666
pnl 0.431161233126911900
commission 0E-18
balance min -1.8459700380155368
balance max 6.108997031624028
balance drawdown 7.844823888438853
romad 0.054961238041599234
---------------- --------------------- ------
Exercise
- Try different symbols, intervals, date ranges
- Set commission
- Implement shorting
- Add
EMA200and only long when price is aboveEMA200