std.append(var_t**0.5)
p = prices['2800'][0:len(mean)].values
df = pd.DataFrame({
'2800': p,
'mean': mean,
'std': std
},
index=prices.index[0:len(mean)])
df['mal'] = MA(df['2800'], 120)
df['mas'] = MA(df['2800'], 10)
df['diff'] = (df['mas'] / df['mal'] - 1) / 10
df['slope'] = df['diff'].pct_change()
df['mean_std'] = 2 * df['std'] - df['mean']
df['zero'] = 0.
df = df.dropna()
from pyetf.figure import plot_chart
plot_chart(df[['2800', 'mal']],
sub=df[['zero', 'mean_std']],
sub_fill=df['std'])
#
df1 = df[df['slope'] > 0]
df2 = df[df['slope'] <= 0]
x1 = df1['diff'].values
y1 = -df1['mean_std'].values
b1, t1 = liner_regression(x1, y1)
x2 = df2['diff'].values
y2 = -df2['mean_std'].values
b2, t2 = liner_regression(x2, y2)
# -*- coding: utf-8 -*-
"""
Created on Sat Jul 27 10:01:54 2019
@author: w
"""
import ffn
from pyetf.data import eod
from pyetf.data import tushare
from pyetf.figure import plot_chart
# retrieve data from eod and combine
etf_tickers = [
'IVV', 'GLD', 'VNQ', 'AGG', 'TLT', 'IEF', 'IEI', 'OIL', 'USO', 'DBC'
]
etf_tickers = ['IVV', 'GLD', 'TLT', 'IEF', 'DBC']
prices = ffn.get(tickers=etf_tickers,
market='US',
provider=eod,
start='2000-01-01')
stats = prices.calc_stats()
stats.display()
pl = prices.copy()
pl.dropna()
pl = pl.rebase()
plot_chart(pl)
mc_budget = [0.05, 0.55]
# retrieve data from eod and combine
start_date_str = '2013-01-01'
prices = ffn.get(tickers=etf_tickers,
market='US',
provider=eod,
start=start_date_str)
# calc portfolio weights
#w1 = prices.to_weights(func_weighting=rpw_standard, risk_weights=mc_budget).dropna()
#w2 = prices.to_weights(func_weighting=rpw_garch, risk_weights=mc_budget).dropna()
w3 = prices.to_weights(func_weighting=rpw_future,
hist_length=-30,
risk_weights=mc_budget).dropna()
# calc portfolio performance
#pf1 = prices.to_NAV(w1)
#pf2 = prices.to_NAV(w2)
pf3 = prices.to_NAV(w3)
#pf1.calc_stats().display()
#pf2.calc_stats().display()
pf3.calc_stats().display()
# plot portfolio
pl = pf3.copy()
w = w3.copy()
pl.dropna()
pl = pl.rebase()
w.dropna()
plot_chart(pl[['NAV', 'spy']], sub_fill=w)
pdvar_lstm[i] = var_lstm[i]
pdvar_lstm.columns = prices.columns
var_tmp = pdvar_lstm.loc[pdvar_lstm.index.year>=start_year].copy()
var_lstm = var_tmp.loc[var_tmp.index<=end_date].copy() # half end year
# GARCH
from pyetf.algos import forecast_var_from_garch
var_garch = []
for t in range(0, len(prices)-m+1):
p = prices.iloc[t:t+m].copy()
r = p.to_returns().dropna()
v = []
for i in range(len(r.columns)):
v_tmp, _ = forecast_var_from_garch(100.0*r[r.columns[i]])
v.append(v_tmp)
var_garch.append(v)
pdvar_garch = pd.DataFrame(data=var_garch, index=prices.index[len(prices)-len(var_garch):len(prices)], columns=prices.columns)
var_tmp = pdvar_garch.loc[pdvar_garch.index.year>=start_year].copy()
var_garch = var_tmp.loc[var_tmp.index<=end_date].copy() # half end year
'''
# plot
e = benchmark
varl = pd.DataFrame(data=var[e], index=var.index)
#varl['lstm']=var_lstm[e]/10000
#varl['garch']=var_garch[e]/10000
varl['risk'] = mv[e]
#varl['mv']=varl[e]*(1+mv_t[e]*10000)
#varl['mv'].loc[varl['mv']<=0]=0.00001
plot_chart(pl[e], sub_fill=varl[['risk']])
#plot_chart(pl[e], sub=varl)
w = prices.to_weights(func_weighting=rpw_ledoit_wolf,
risk_weights=mc_budget,
min_assets_number=minn,
max_assets_number=maxn).dropna()
elif method == 'mean_var':
#historical
w = prices.to_weights(func_weighting=rpw_mean_var,
model="mean_var",
risk_weights=mc_budget,
min_assets_number=minn,
max_assets_number=maxn).dropna()
# calc portfolio performance
pf = prices.to_NAV(w)
# plot portfolio
#pl = pf.copy() # all data
pl_tmp = pf.loc[pf.index.year >= start_year].copy()
pl = pl_tmp.loc[pl_tmp.index <= end_date].copy() # half end year
#pl = pl_tmp.loc[pl_tmp.index.year<=end_year].copy() # whole end year
pl.dropna()
w_tmp = w.loc[w.index.year >= start_year].copy()
wl = w_tmp.loc[w_tmp.index <= end_date].copy() # half end year
#wl = w_tmp.loc[w_tmp.index.year<=end_year].copy() # whole end year
wl.dropna()
pl = pl.rebase()
plot_chart(pl[['NAV', benchmark]], sub_fill=wl)
# performance
pl.calc_stats().display()
#for y in range(start_year, end_year+1):
# pl[str(y)].calc_stats().display()
# hist
hts = ts.loc['2013'].values
var1 = []
# rolling estimate var
while (len(hts) < len(ts)):
f_var1 = hts[-hln:].var()
var1.append(f_var1)
hts = np.append(hts, ts.iloc[len(hts)])
var1 = np.append(np.zeros([len(prices) - len(var1), 1]), var1)
# garch method
hts = ts.loc['2013'].values
var2 = []
# rolling estimate var
while (len(hts) < len(ts)):
f_var2, r2 = forecast_var_from_garch(hts[-hln:])
var2.append(f_var2)
hts = np.append(hts, ts.iloc[len(hts)])
var2 = np.append(np.zeros([len(prices) - len(var2), 1]), var2)
#lstm
var3 = forecast_var_from_lstm(prices[e])
prices['var1'] = var1
prices['var2'] = var2
prices['var3'] = var3 * 10
from pyetf.figure import plot_chart
plot_chart(prices[['spy']], sub=prices[['var1', 'var2', 'var3']])