def add_indicators(self):
#self.add_ema(9)
#self.add_ema(21)
#self.strategy = strategies.CustomStrategy(self.df['close'], self.df['ema9'], self.df['ema21'])
#return self.strategy
self.rsi = indicators.RSI(self.df['close'], 9)
self.macd = indicators.MACD(self.df['close'], 12, 26, 9)
def test_MACD(self):
macd = indicators.MACD(self.ohlc.close, 12, 26, 9)
strategy = strategies.MACDStrategy(self.ohlc, macd)
result = strategy.backtest(self.ohlc.close)
self.assertEqual(result.start, datetime(2020, 1, 8, 1, 30, 0))
self.assertEqual(result.end, datetime(2020, 1, 18, 11, 15, 0))
self.assertEqual(result.trades, 41)
self.assertTrue(math.isclose(result.netret, 22.39, rel_tol=1e-02))
self.assertTrue(math.isclose(result.sharpe, 0.1409, rel_tol=1e-03))
self.assertTrue(math.isclose(result.maxdrawdown, -2.97, rel_tol=1e-02))
self.assertEqual(result.maxdrawdownduration, 156)
def test_RSI_MACD(self):
rsi = indicators.RSI(self.ohlc.close, period=9)
macd = indicators.MACD(self.ohlc.close, 12, 26, 9)
strategy = strategies.RSIMACDStrategy(self.ohlc, rsi, macd)
result = strategy.backtest(self.ohlc.close)
self.assertEqual(result.start, datetime(2020, 1, 8, 1, 30, 0))
self.assertEqual(result.end, datetime(2020, 1, 18, 11, 15, 0))
self.assertEqual(result.trades, 41)
self.assertTrue(math.isclose(result.netret, 26.06, rel_tol=1e-02))
self.assertTrue(math.isclose(result.sharpe, 0.1297, rel_tol=1e-03))
self.assertTrue(math.isclose(result.maxdrawdown, -2.85, rel_tol=1e-02))
self.assertEqual(result.maxdrawdownduration, 82)
def drawMACD(ax, data, periods=[12, 26, 9]):
close = data['close']
if data.__contains__('dif') & data.__contains__('dea') & data.__contains__('hist'):
macd = data['dif']
macd_signal = data['dea']
macd_histogram = data['hist']
else:
macd, macd_signal, macd_histogram = ind.MACD(close, fast_period=periods[0], slow_period=periods[1],
signal_period=periods[2])
ax.plot(macd, label='macd%d' % periods[0])
ax.plot(macd_signal, label='macd_singal%d' % periods[1])
ax.bar(close.index, macd_histogram.clip(lower=0), facecolor='r')
ax.bar(close.index, macd_histogram.clip(upper=0), facecolor='g')
drawHline(ax, [0])
# ax.set_ylabel('MACD')
ax.set_yticks([])
def test_MACD(self):
macd = pd.read_csv('tests/data/BTCUSDT_15m_MACD_12_26_9.csv',
index_col='time',
parse_dates=True)
mymacd = indicators.MACD(self.ohlc.close,
short_period=12,
long_period=26,
period=9)
# consider first 250 values as warmup
ismacdclose = np.isclose(macd.macd.iloc[250:],
mymacd.df.MACD.iloc[250:],
equal_nan=True)
ismacdsignalclose = np.isclose(macd.macdsignal.iloc[250:],
mymacd.df.signal.iloc[250:],
equal_nan=True)
self.assertTrue(np.all(ismacdclose))
self.assertTrue(np.all(ismacdsignalclose))
def _indicators(self):
""" This will calculate our technical indicators based on the
parameters in our ti dict ... this can be ran in one bang OR
iteratively. It goes directly to the indicator structs. Not pretty,
but what the f**k.
"""
# TODO: update this if we ever want to add multiple OHLC frames
for time_str in self.ohlc:
for indicator in self.ti:
if indicator == "RSI":
self.ti[indicator]["data"] = ind.RSI(
self.ohlc[time_str], self.ti[indicator]["n"])
elif indicator == "ROC":
self.ti[indicator]["data"] = ind.ROC(
self.ohlc[time_str], self.ti[indicator]["n"])
elif indicator == "AMA":
self.ti[indicator]["data"] = ind.AMA(
self.ohlc[time_str].close, self.ti[indicator]["n"],
self.ti[indicator]["fn"], self.ti[indicator]["sn"])
elif indicator == "CCI":
self.ti[indicator]["data"] = ind.CCI(
self.ohlc[time_str], self.ti[indicator]["n"])
elif indicator == "FRAMA":
self.ti[indicator]["data"] = ind.FRAMA(
self.ohlc[time_str], self.ti[indicator]["n"])
elif indicator == "RVI2":
self.ti[indicator]["data"] = ind.RVI2(
self.ohlc[time_str], self.ti[indicator]["n"],
self.ti[indicator]["s"])
elif indicator == "MACD":
self.ti[indicator]["data"] = ind.MACD(
self.ohlc[time_str], self.ti[indicator]["f"],
self.ti[indicator]["s"], self.ti[indicator]["m"])
elif indicator == "ADX":
self.ti[indicator]["data"] = ind.ADX(
self.ohlc[time_str], self.ti[indicator]["n"])
elif indicator == "ELI":
self.ti[indicator]["data"] = ind.ELI(
self.ohlc[time_str], self.ti[indicator]["n"])
elif indicator == "TMI":
self.ti[indicator]["data"] = ind.TMI(
self.ohlc[time_str], self.ti[indicator]["nb"],
self.ti[indicator]["nf"])
def indicator_values(self, df_prices):
"""Function is used to generate X inputs based on indicator performance
Accepts a prices df from util.get_data, returns a df of indicator performance"""
# window = self.indicator_window
# simple moving average and bollinger bands
sma, upper, lower = ind.simple_ma(df_prices,
window=19,
bollinger=True,
threshold=1.45102)
# BB percentage
perc_upper = df_prices.iloc[:, 0] / upper.iloc[:, 0]
perc_upper = pd.DataFrame(perc_upper.rename('upper_BB%'))
perc_lower = lower.iloc[:, 0] / df_prices.iloc[:, 0]
perc_lower = pd.DataFrame(perc_lower.rename('lower_BB%'))
# moving average convergence divergence
macd, macd_s = ind.MACD(df_prices, ema1_days=12, ema2_days=24)
# below checks percentage crossover
md = macd_s['macd_signal'] / macd['macd']
md.rename('macd_sigal/macd')
# stoichastic oscillator
K, D = ind.stoc_osc(df_prices, k_window=12)
# below is to check for percentage of crossover for the stoich osc
k = D.iloc[:, 0] / K.iloc[:, 0]
k.replace([np.inf, -np.inf, np.nan], 0, inplace=True)
k.rename('D/K', inplace=True)
df_ind = pd.concat([perc_upper, perc_lower, md, D, k], axis=1)
self.window = df_ind.isnull().any(1).nonzero()[0].max(
) + 1 # highest index is mainteined to prevent training during this period
if self.verbose:
self.ind = df_ind
return (df_ind.iloc[self.window:, :])
def main():
parser = argparse.ArgumentParser()
parser.add_argument("file")
parser.add_argument("--sma", type=int, nargs='+', help='Adds SMA to the price graph')
parser.add_argument("--ema", type=int, nargs='+', help='Adds EMA to the price graph')
parser.add_argument("--rsi", type=int, help='Adds RSI in a subplot')
parser.add_argument("--macd", type=int, nargs=3, help='Adds MACD in a subplot')
parser.add_argument("--bbands", type=int, nargs=2, help='Adds Bollinger Bands to the price graph')
parser.add_argument("--strategy", type=int, nargs='+', help='Adds a strategy, 1 : EMA Cross')
args = parser.parse_args()
ohlc = pd.read_csv(args.file, index_col='time', parse_dates=True)
close = ohlc['close']
# calculates the number of sublots
rows = 1
if args.rsi != None:
rows += 1
if args.macd != None:
rows += 1
fig = make_subplots(rows=rows, cols=1, shared_xaxes=True)
# plots close bar graph with averages
row = 1
fig.add_trace(go.Scatter(x=ohlc.index, y=close, name="Close"), row=row, col=1)
if args.ema != None:
for period in args.ema:
ema = indicators.EMA(close, period)
fig.add_trace(go.Scatter(x=ohlc.index, y=ema.df['ema{}'.format(period)], name="EMA {}".format(period)), row=row, col=1)
if args.sma != None:
for period in args.sma:
sma = indicators.SMA(close, period)
fig.add_trace(go.Scatter(x=ohlc.index, y=sma.df['sma{}'.format(period)], name="SMA {}".format(period)), row=row, col=1)
# plots Bollinger Bands
if args.bbands != None:
bbands = indicators.BollingerBands(close, args.bbands[0])
fig.add_trace(go.Scatter(x=ohlc.index, y=bbands.df.ma), row=row, col=1)
fig.add_trace(go.Scatter(x=ohlc.index, y=bbands.df.upper), row=row, col=1)
fig.add_trace(go.Scatter(x=ohlc.index, y=bbands.df.lower), row=row, col=1)
row += 1
row += 1
# plots RSI
if args.rsi != None:
rsi = indicators.RSI(close, args.rsi)
fig.add_trace(go.Scatter(x=ohlc.index, y=rsi.df.rsi, name="RSI {}".format(args.rsi)), row=row, col=1)
row += 1
# plots MACD
if args.macd != None:
macd = indicators.MACD(close, args.macd[0], args.macd[1], args.macd[2])
fig.add_trace(go.Scatter(x=ohlc.index, y=macd.df.MACD, name="MACD {} {} {}".format(args.macd[0], args.macd[1], args.macd[2])), row=row, col=1)
fig.add_trace(go.Scatter(x=ohlc.index, y=macd.df.signal, name="MACD Signal"), row=row, col=1)
row += 1
# plots strategy
if args.strategy != None:
if args.strategy[0] == 1:
fast_ema = indicators.EMA(ohlc.close, period=args.strategy[1])
slow_ema = indicators.EMA(ohlc.close, period=args.strategy[2])
strategy = strategies.AvgCrossStrategy(ohlc.close, fast_ema.data(), slow_ema.data())
fig.add_trace(go.Scatter(x=close.loc[strategy.signals['positions'] == 1.0].index, y=close.loc[strategy.signals['positions'] == 1.0],
mode='markers', marker=dict(size=12, symbol='triangle-up', color='green'), name="Buy"), row=1, col=1)
fig.add_trace(go.Scatter(x=close.loc[strategy.signals['positions'] == -1.0].index, y=close.loc[strategy.signals['positions'] == -1.0],
mode='markers', marker=dict(size=12, symbol='triangle-down', color='red'), name="Sell"), row=1, col=1)
elif args.strategy[0] == 2:
macd = indicators.MACD(ohlc.close, args.strategy[1], args.strategy[2], args.strategy[3])
strategy = strategies.MACDStrategy(ohlc.close, macd)
fig.add_trace(go.Scatter(x=close.loc[strategy.signals['positions'] == 1.0].index, y=close.loc[strategy.signals['positions'] == 1.0],
mode='markers', marker=dict(size=12, symbol='triangle-up', color='green'), name="Buy"), row=1, col=1)
fig.add_trace(go.Scatter(x=close.loc[strategy.signals['positions'] == -1.0].index, y=close.loc[strategy.signals['positions'] == -1.0],
mode='markers', marker=dict(size=12, symbol='triangle-down', color='red'), name="Sell"), row=1, col=1)
elif args.strategy[0] == 3:
bb1 = indicators.BollingerBands(ohlc.close, 20, 1)
bb2 = indicators.BollingerBands(ohlc.close, 20, 2)
strategy = strategies.DBBStrategy(ohlc.close, bb1, bb2)
fig.add_trace(go.Scatter(x=close.loc[strategy.signals['positions'] == 1.0].index, y=close.loc[strategy.signals['positions'] == 1.0],
mode='markers', marker=dict(size=12, symbol='triangle-up', color='green'), name="Buy"), row=1, col=1)
fig.add_trace(go.Scatter(x=close.loc[strategy.signals['positions'] == -1.0].index, y=close.loc[strategy.signals['positions'] == -1.0],
mode='markers', marker=dict(size=12, symbol='triangle-down', color='red'), name="Sell"), row=1, col=1)
elif args.strategy[0] == 4:
rsi = indicators.RSI(ohlc.close, 9)
macd = indicators.MACD(ohlc.close, 12, 26, 9)
strategy = strategies.RSIMACDStrategy(ohlc.close, rsi, macd)
fig.add_trace(go.Scatter(x=close.loc[strategy.signals['positions'] == 1.0].index, y=close.loc[strategy.signals['positions'] == 1.0],
mode='markers', marker=dict(size=12, symbol='triangle-up', color='green'), name="Buy"), row=1, col=1)
fig.add_trace(go.Scatter(x=close.loc[strategy.signals['positions'] == -1.0].index, y=close.loc[strategy.signals['positions'] == -1.0],
mode='markers', marker=dict(size=12, symbol='triangle-down', color='red'), name="Sell"), row=1, col=1)
fig.show()
tp_style='hlc3')
BB20.batch_fit(candles)
# RSI:
print('Fitting RSI...')
RSI = ind.RSI(window_length=14, time_frame=time_frame)
RSI.batch_fit(candles)
# Stochastic RSI:
print('Fitting Stochastic RSI...')
SRSI = ind.StochasticRSI(stoch_length=14, time_frame=time_frame)
SRSI.batch_fit(candles)
# MACD:
print('Fitting MACD...')
MACD = ind.MACD(12, 26, 9, time_frame, 'close')
MACD.batch_fit(candles)
""" Making plots using Plotly """
print('\nPlotting...')
fig = make_subplots(
rows=5,
cols=1,
shared_xaxes=True,
specs=[[{
"rowspan": 2
}], [None], [{}], [{}], [{}]],
)
# Bollinger Bands on the background:
BB20.plot(fig)
# Then Candlesticks:
def main():
parser = argparse.ArgumentParser()
parser.add_argument("file")
args = parser.parse_args()
df = pd.read_csv(args.file, index_col='time', parse_dates=True)
#close = df.loc['20191225':,['close']].astype('float64')
close = df
print("Buy And Hold Strategy :")
strat1 = strategies.BuyAndHoldStrategy(close, 0.001)
res = strat1.backtest(close['close'])
print(res)
print(flush=True)
add_ema(close, 12)
add_ema(close, 26)
macd = indicators.MACD(close['close'], 12, 26, 9)
rsi = indicators.RSI(close['close'], 9)
bb1 = indicators.BollingerBands(close['close'], 20, 1)
bb2 = indicators.BollingerBands(close['close'], 20, 2)
print("MACD Strategy :")
strat3 = strategies.MACDStrategy(close, macd, 0.001)
res = strat3.backtest(close['close'])
print(res)
print(flush=True)
print("RSI Strategy :")
strat4 = strategies.RSIStrategy(close, rsi, 0.001)
res = strat4.backtest(close['close'])
print(res)
print(flush=True)
print("Double Bollinger Bands Strategy :")
strat5 = strategies.DBBStrategy(close['close'], bb1, bb2)
res = strat5.backtest(close['close'])
print(res)
print(flush=True)
print("Custom RSI + MACD Strategy :")
strat2 = strategies.RSIMACDStrategy(close, rsi, macd, 0.001)
res = strat2.backtest(close['close'])
print(res)
fig, (ax1, ax2, ax3, ax4) = plt.subplots(4, sharex=True)
fig.suptitle(args.file)
ax1.plot(close['close'])
ax1.plot(close['ema12'], label='EMA12')
ax1.plot(close['ema26'], label='EMA26')
ax1.plot(close['close'].loc[strat2.signals['positions'] == 1.0].index,
close['close'].loc[strat2.signals['positions'] == 1.0],
'^',
markersize=10,
color='g')
ax1.plot(close['close'].loc[strat2.signals['positions'] == -1.0].index,
close['close'].loc[strat2.signals['positions'] == -1.0],
'v',
markersize=10,
color='r')
ax1.legend()
macd, macd_signal = macd.data()
ax2.plot(macd, label='MACD 12 26 9')
ax2.plot(macd_signal)
#ax3.plot(strat1.signals['equity'])
ax3.plot(rsi.data(), label='RSI 9')
ax4.plot(strat2.signals['equity'])
plt.show()