def plot_pred(pred_df,
real_df,
real_col,
pred_col="fitted",
upper_col="upper",
lower_col="lower",
outpath=None,
show=True):
import datetime
from pathlib import Path
from plotly.subplots import make_subplots
# compute last date of training data
last_date = pred_df.index[0]
last_date -= datetime.timedelta(days=1)
y_annotation = pred_df[upper_col].max()
min_y_plot = real_df[-21:][real_col].min()
max_y_plot = real_df[-21:][real_col].max()
symbol_business = 'circle'
symbol_holidays = 'diamond'
sz_business = 7
sz_holidays = 10
marker_symbols_real = [
symbol_business if is_business_day(d.date()) else symbol_holidays
for d in real_df.index
]
marker_sizes_real = [
sz_business if is_business_day(d.date()) else sz_holidays
for d in real_df.index
]
marker_symbols_pred = [
symbol_business if is_business_day(d.date()) else symbol_holidays
for d in pred_df.index
]
marker_sizes_pred = [
sz_business if is_business_day(d.date()) else sz_holidays
for d in pred_df.index
]
fig = make_subplots(
rows=2,
cols=1,
# shared_xaxes=True,
vertical_spacing=0.2,
specs=[[{
"type": "scatter"
}], [{
"type": "table"
}]])
fig.add_trace(draw_scatter(real_df, real_col, 'Reali', marker_symbols_real,
marker_sizes_real),
row=1,
col=1)
fig.add_trace(draw_scatter(pred_df, pred_col, 'Previsioni',
marker_symbols_pred, marker_sizes_pred),
row=1,
col=1)
fig.add_trace(draw_confidence_bands(pred_df, upper_col, 'Upper Bound'),
row=1,
col=1)
fig.add_trace(draw_confidence_bands(pred_df, lower_col, 'Lower Bound'),
row=1,
col=1)
fig.add_shape(type="line",
x0=last_date,
y0=0,
x1=last_date,
y1=y_annotation,
name="Ultimi dati training",
line=dict(
color="firebrick",
width=4,
dash="dashdot",
)),
fig.add_annotation(x=last_date,
y=min_y_plot,
ax=-50,
ay=20,
xanchor="right",
yanchor="top",
text=f"Ultimi dati training:\n{last_date.date()}",
showarrow=True,
arrowhead=2),
fig.update_layout(xaxis_title='Data',
xaxis=dict(range=[
last_date - datetime.timedelta(14),
last_date + datetime.timedelta(8)
],
autorange=False),
yaxis_title='N. ospedalizzazioni',
yaxis=dict(range=[
min_y_plot - y_annotation / 10,
max_y_plot + y_annotation / 10
],
autorange=False),
title=f'Andamento {real_col}'.title(),
hovermode="x",
font=dict(
family="Courier New, monospace",
size=16,
))
fig.add_trace(make_table(pred_df, real_col), row=2, col=1)
fig.update_layout(width=1000,
height=800,
paper_bgcolor='rgba(0,0,0,0)',
plot_bgcolor='rgba(0,0,0,0)')
if show:
fig.show()
else:
if not outpath:
outpath = f"./results/{last_date.date()}/{real_col}.html"
Path(outpath).parent.mkdir(exist_ok=True, parents=True)
fig.write_html(outpath)
fig.write_image(outpath.replace(".html", ".png"))
return fig
def trade_signal(signal_date, commodity, months_ahead):
#Confirms signal_date is not a business day or holiday
if not is_business_day(signal_date) or (is_holiday(signal_date)):
return -2
#Finds the near and far commodity ticker
#Crude oil
if commodity == "WTI":
#WTI crude
near_commodity = "CHRIS/CME_CL1"
far_commodity = ''.join(["CHRIS/CME_CL", str(months_ahead + 1)])
elif commodity == "BRENT":
#Brent crude
near_commodity = "CHRIS/ICE_B1"
far_commodity = ''.join(["CHRIS/ICE_B", str(months_ahead + 1)])
#Natural gas
elif commodity == "NG":
#US natural gas
near_commodity = "CHRIS/CME_NG1"
far_commodity = ''.join(["CHRIS/CME_NG", str(months_ahead + 1)])
elif commodity == "UKNG":
#UK natural gas
near_commodity = "CHRIS/ICE_M1"
far_commodity = ''.join(["CHRIS/ICE_M", str(months_ahead + 1)])
#Cattle
elif commodity == "LC":
#Live cattle
near_commodity = "CHRIS/CME_LC1"
far_commodity = ''.join(["CHRIS/CME_LC", str(months_ahead + 1)])
elif commodity == "FC":
#Feeder cattle
near_commodity = "CHRIS/CME_FC1"
far_commodity = ''.join(["CHRIS/CME_FC", str(months_ahead + 1)])
#Distillates
elif commodity == "RBOBGAS":
#RBOB Gas
near_commodity = "CHRIS/CME_RB1"
far_commodity = ''.join(["CHRIS/CME_RB", str(months_ahead + 1)])
elif commodity == "GASOIL":
#Gas oil
near_commodity = "CHRIS/ICE_G1"
far_commodity = ''.join(["CHRIS/ICE_G", str(months_ahead + 1)])
elif commodity == "HEATOIL":
#NY Harbor heating oil (CME)
near_commodity = "CHRIS/CME_HO1"
far_commodity = ''.join(["CHRIS/CME_HO", str(months_ahead + 1)])
#Wheat
elif commodity == "WHEAT":
#Plain ol' wheat
near_commodity = "CHRIS/CME_W1"
far_commodity = ''.join(["CHRIS/CME_W", str(months_ahead + 1)])
elif commodity == "KWHEAT":
#Kansas City hard red wheat
#Only goes out to the fifth month (sixth future)
near_commodity = "CHRIS/CME_KW1"
far_commodity = ''.join(["CHRIS/CME_KW", str(months_ahead + 1)])
elif commodity == "MWHEAT":
#Milling wheat
#Only goes out to the fifth month (sixth future)
near_commodity = "CHRIS/LIFFE_EBM1"
far_commodity = ''.join(["CHRIS/LIFFE_EBM", str(months_ahead + 1)])
#Catch all for commodities not on the list
else:
return -3
#Converts the date to the right format and fetches the near and far future
if signal_date.month < 10:
signal_date_month_string = ''.join(["0", str(signal_date.month)])
else:
signal_date_month_string = str(signal_date.month)
if signal_date.day < 10:
signal_date_day_string = ''.join(["0", str(signal_date.day)])
else:
signal_date_day_string = str(signal_date.day)
signal_date_string = ''.join([
'"',
str(signal_date.year), '-', signal_date_month_string, '-',
signal_date_day_string, '"'
])
## print(signal_date_string)
try:
near = quandl.get(near_commodity,
start_date=signal_date_string,
end_date=signal_date_string,
returns="numpy")
for item in near:
near_price = float(item[4])
## print(near_price)
except:
return -4
try:
far = quandl.get(far_commodity,
start_date=signal_date_string,
end_date=signal_date_string,
returns="numpy")
for item in far:
far_price = float(item[4])
## print(far_price)
except:
return -5
#Figures out the dates of the the near and far futures, days between the two dates
near_date = expiration_date(signal_date.year, signal_date.month, commodity)
far_date = near_date + relativedelta(months=+months_ahead)
far_date = expiration_date(far_date.year, far_date.month, commodity)
## print(near_date)
## print(far_date)
if near_date.date() <= signal_date:
## print(True)
near_date = near_date + relativedelta(months=+1)
far_date = far_date + relativedelta(months=+1)
## print(near_date)
## print(far_date)
delta = far_date - near_date
## print(delta.days)
#Calculates the signal and returns it
try:
signal = (math.log(near_price) -
math.log(far_price)) * 365 / delta.days
except:
return -6
return signal
},
ignore_index=True)
for j in range(df1.shape[0]):
df_output['pct_chg_date_%d' % (j + 1)][i] = df1['pct_chg'][j]
i += 1
print(df_output)
logging.debug(df_output)
if __name__ == '__main__':
logging.debug('start...')
print('analyze daily data')
start_date = '20210514'
end_date = '20210514'
start = datetime.datetime.strptime(start_date, '%Y%m%d')
end = datetime.datetime.strptime(end_date, '%Y%m%d')
#end = datetime.datetime.now()
#start=end -datetime.timedelta(days = 2)
for i in range((end - start).days + 1):
date = start + datetime.timedelta(days=i)
date_str = date.strftime('%Y%m%d')
print(date_str)
if utils.is_business_day(date):
data = chaodiefantan(date_str)
#measuremnet(data)
else:
print('Not Business day!')
print('end')
# b2 = read_data('ICE', 'B', 2)
b7 = read_data(second_exchange, second_commodity_code, 7)
cb = pd.concat([cl1, b1], axis=1, join='inner')
# Initialization of Parameters
cl_pos = 0
b_pos = 0
pnl = 0
pnl_ls = []
day_count = 0 # The way the code structured now, only trading days are counted.
previous_price1 = 0
previous_price2 = 0
position = 0 # This is the indicator of whether we are holding positions, if
# we are during rolling period and not holding anything, this would also be 1
if is_holiday(start_date) or not is_business_day(start_date):
dt = next_business_nonholilday(start_date, 1)
else:
dt = start_date
# Executing the Roll Yield Strategy
while (dt < termination_date):
# Calculating PnL
if dt in cb.index:
if day_count != 0:
pnl += cl_pos * (cl1.loc[dt][0] - previous_price1)
+b_pos * (b1.loc[dt][0] - previous_price2)
pnl_ls.append([dt, pnl])
# Check if it's time to roll
# If both contracts have non-zero positions