def _analyze(self):
# Summary output
ret = self.ret
ret = ret[ret.index >= self.s_date]
ret.index = pd.to_datetime(ret.index)
ret['net_close'] = ret['close'] / ret['close'].tolist()[0]
ret['net'] = ret['net'] / ret['net'].tolist()[0]
net_lst = ret['net'].tolist()
annR = annROR(net_lst, 1)
sharp = yearsharpRatio(net_lst, 1)
max_retrace = maxRetrace(net_lst, 1)
self.annR = annR
self.sharp = sharp
self.max_retrace = max_retrace
f = plt.figure(figsize=(15, 8))
plt.rcParams['font.sans-serif'] = ['SimHei']
# Plot return
ax1 = f.add_subplot(111)
ax1.plot(ret.net, 'blue', label='net_hmm')
ax1.plot(ret.net_close, 'red', label='net_close')
ax1.legend()
ax1.set_title(
'%s特征因子%s训练窗口%s测试窗口%s净值曲线' %
(self.asset, self.cols_features, self.train_days, self.test_days))
ax1.set_xlabel('Time')
ax1.set_ylabel('Value')
def _analyze(self):
# Summary output
ret = self.ret
ret = ret[ret.index >= self.s_date]
ret.index = pd.to_datetime(ret.index)
ret['net_close'] = ret['close'] / ret['close'].tolist()[0]
ret['net'] = ret['net'] / ret['net'].tolist()[0]
self.ret = ret
net_lst = ret['net'].tolist()
annR = annROR(net_lst, 1)
sharp = yearsharpRatio(net_lst, 1)
max_retrace = maxRetrace(net_lst, 1)
self.annR = annR
self.sharp = sharp
self.max_retrace = max_retrace
def get_account_state(fold_path, file_lst, calen):
net_df = []
for date in calen:
for i in file_lst:
if date in i and 'account' in i:
df_ = pd.read_csv(fold_path + i, encoding='gbk') \
.loc[:, ['交易日', '总资产']]
net_df.append(df_)
break
net_df = pd.concat(net_df).sort_values(['交易日'])
print(net_df)
state_sdate = calen[0]
state_edate = calen[-1]
asset_lst = net_df['总资产'].tolist()
net_lst = [i / asset_lst[0] for i in asset_lst]
annR = annROR(net_lst, 1)
sharp = yearsharpRatio(net_lst, 1)
max_retrace = maxRetrace(net_lst, 1)
total_ret = net_lst[-1] - 1
today_ret = net_lst[-1] / net_lst[-2] - 1
print('sharp:%s' % sharp)
print('annR:%s' % annR)
print('max_retrace:%s' % max_retrace)
state_value_lst = []
state_value_lst.append([
state_sdate, state_edate, asset_lst[0], asset_lst[-1], total_ret,
today_ret, annR, sharp, max_retrace
])
df_today = pd.DataFrame(state_value_lst,
columns=[
'开始日期', '结束日期', '初始资产', '当前资产', '总收益', '当日收益',
'年化收益', '夏普', '最大回撤'
])
df_today['初始资产'] = df_today['初始资产'].apply(lambda x: '%.0f' % x)
df_today['当前资产'] = df_today['当前资产'].apply(lambda x: '%.0f' % x)
df_today['夏普'] = df_today['夏普'].apply(lambda x: '%.2f' % x)
df_today['年化收益'] = df_today['年化收益'].apply(lambda x: '%.2f%%' % (x * 100))
df_today['总收益'] = df_today['总收益'].apply(lambda x: '%.2f%%' % (x * 100))
df_today['当日收益'] = df_today['当日收益'].apply(lambda x: '%.2f%%' % (x * 100))
df_today['最大回撤'] = df_today['最大回撤'].apply(lambda x: '%.2f%%' % (x * 100))
return df_today
# profit_df = profit_df.fillna(0)
profit_df['chg'] = (
profit_df['profit'] - profit_df['close'].shift(1) *
profit_df['count'] * VolumeMultiple * fee
) * level / profit_df['close'].shift(1) / (VolumeMultiple *
2)
profit_df = profit_df.fillna(0)
if method == 'sum':
profit_df['net'] = 1 + profit_df['chg'].cumsum()
else:
profit_df['net'] = (1 + profit_df['chg']).cumprod()
print(profit_df)
net_lst = profit_df.net.tolist()
if level == 1:
sharpe_ratio = yearsharpRatio(net_lst, 1)
if method == 'sum':
ann_return = annROR_signal(net_lst, 1)
else:
ann_return = annROR(net_lst, 1)
max_drawdown = maxRetrace(net_lst, 1)
ret['symbol'].append(state.iloc[0, 2].split('.')[0])
ret['tm'].append(state.iloc[1, 2].split()[0] +
state.iloc[1, 2].split()[1])
ret['start_time'].append(s_date)
ret['end_time'].append(e_date)
ret['复盘模型'].append(state.iloc[2, 2])
ret['K线数量'].append(state.iloc[5, 1])
ret['盈利比'].append(state.iloc[10, 1])
ret['trading_times'].append(int(state.iloc[13, 1]))
def calculate_strategy_net(calen, next_tradeday, trade_fold_path,
trade_file_lst, strategy_fold_path, account_hold,
strategy_name_lst):
trade_detail_all = []
for date in calen:
trade_detail = get_trade_detail(trade_fold_path, trade_file_lst, date)
trade_detail_all.append(trade_detail)
trade_detail_all = pd.concat(trade_detail_all)
trade_state_all = get_trade_state(trade_detail_all)[[
'策略名称', '证券代码', '证券名称', '操作', '成交均价', '成交数量', '成交金额', '手续费', '日期'
]]
trade_state_all['bs'] = trade_state_all['操作'].apply(
lambda x: trans_to_bs(x))
trade_state_all['持仓'] = trade_state_all['成交数量'] * trade_state_all['bs']
trade_state_all['成交金额'] = -trade_state_all['成交金额'] * trade_state_all['bs']
win_r_lst = []
trade_base_date_all = []
for strategy_name in strategy_name_lst:
temp = trade_state_all[trade_state_all['策略名称'] == strategy_name]
print(temp)
ret_lst = []
if len(temp) == 0:
win_r_lst.append([strategy_name, 1, 1])
continue
win_time = 0
trad_time = 0
cost_lst = []
buy_code_lst = []
for date, group in temp.groupby(['日期']):
if len(buy_code_lst) > 0:
for i in range(len(buy_code_lst)):
cost_i = cost_lst[i]
code = buy_code_lst[i]
sell_df = group[(group['bs'] == -1)
& (group['证券代码'] == code)]
if len(sell_df) > 0:
sell_value = abs(sell_df['成交金额'].sum()) - abs(
sell_df['手续费'].sum())
ret = sell_value - cost_i
ret_lst.append(ret)
if ret > 0:
win_time += 1
trad_time += 1
else:
trad_time += 1
trade_base_date_all.append([strategy_name, date, ret])
cost_lst = []
buy_code_lst = []
buy_df = group[group['bs'] == 1]
if len(buy_df) == 0:
buy_code_lst = []
else:
for buy_code, buy_group in buy_df.groupby(['证券代码']):
buy_code_lst.append(buy_code)
cost = abs(buy_group['成交金额'].sum()) + abs(
buy_group['手续费'].sum())
cost_lst.append(cost)
win_r = 1
odd = 1
if trad_time > 0:
win_r = win_time / trad_time
ret_pos = [i for i in ret_lst if i > 0]
ret_nev = [i for i in ret_lst if i < 0]
if len(ret_nev) > 0:
pos_ave = 0
if len(ret_pos) > 0:
pos_ave = abs(np.mean(ret_pos))
odd = pos_ave / abs(np.mean(ret_nev))
win_r_lst.append([strategy_name, win_r, odd])
trade_base_date_all_df = pd.DataFrame(trade_base_date_all,
columns=['策略名称', '日期', '收益'])
win_df = pd.DataFrame(win_r_lst, columns=['策略名称', '胜率', '盈亏比'])
trade_state_today = trade_state_all[trade_state_all['日期'] == calen[-1]]
hold_df_firstday = trade_state_all[(trade_state_all['日期'] == calen[0])
& (trade_state_all['bs'] == 1)]
hold_df_back = trade_state_all[(trade_state_all['日期'] > calen[0])]
hold_df = pd.concat([hold_df_firstday, hold_df_back])[['策略名称', '证券代码', '证券名称', '持仓']]\
.groupby(['策略名称', '证券代码', '证券名称']).sum().reset_index(drop=False)
print('==============hold_df')
print(hold_df)
hold_today = hold_df[hold_df['持仓'] > 0]
hold_today = trade_state_all[['策略名称', '证券代码', '证券名称', '日期', '成交均价', 'bs']]\
.merge(hold_today, on=['策略名称', '证券代码', '证券名称'])
hold_today = hold_today[(hold_today['日期'] == hold_today['日期'].max())
& (hold_today['bs'] == 1)]
account_hold_ = account_hold[['证券代码', '持股', '最新价']]
account_hold_lst = account_hold['证券代码'].tolist()
account_hold_ = account_hold_.set_index(['证券代码'])
print('==============hold_today')
print(hold_today)
print('==============account_hold_')
print(account_hold_)
hold_ret_lst = []
for strategy_name, group in hold_today.groupby(['策略名称']):
for id, row_ in group.iterrows():
code = row_['证券代码']
if code not in account_hold_lst:
continue
hold = row_['持仓']
price = account_hold_.loc[code]['最新价']
cost__ = row_['成交均价']
hold = min(hold, account_hold_.loc[code]['持股'])
hold_ret_lst.append(
[strategy_name, code, row_['证券名称'], hold, cost__, price])
hold_strategy_today = pd.DataFrame(
hold_ret_lst, columns=['策略名称', '证券代码', '证券名称', '持仓', '成本价', '最新价'])
print('==============hold_strategy_today')
print(hold_strategy_today)
strategy_account_all_today = []
strategy_account_all_dict = {}
for strategy_name in strategy_name_lst:
name = strategy_dict[strategy_name]
hold = hold_strategy_today[hold_strategy_today['策略名称'] ==
strategy_name]
hold.index = hold['证券代码']
trade_state = trade_state_today[trade_state_today['策略名称'] ==
strategy_name]
last_strategy_account = pd.read_csv(
strategy_fold_path + name + '_' + calen[-1] + '.csv',
encoding='gbk')[['策略名称', 'date', 'cash', 'position_value']]
last_strategy_account['date'] = last_strategy_account['date'].apply(
lambda x: str(x))
last_strategy_account = last_strategy_account.set_index(['date'])
#print(type(last_strategy_account['date'].tolist()[0]))
last_cash = last_strategy_account.loc[calen[-2]]['cash']
today_cash = last_cash + trade_state['成交金额'].sum(
) - trade_state['手续费'].sum()
today_hold = 0
print(hold)
if len(hold) > 0:
for code in hold['证券代码'].tolist():
price = hold.loc[code]['最新价']
position = hold.loc[code]['持仓']
today_hold = today_hold + price * position
today_strategy_account = pd.DataFrame(
[[strategy_name,
str(calen[-1]), today_cash, today_hold]],
columns=['策略名称', 'date', 'cash', 'position_value'])
print(today_strategy_account)
last_strategy_account = last_strategy_account.reset_index(drop=False)[[
'策略名称', 'date', 'cash', 'position_value'
]]
print(last_strategy_account)
strategy_account_new = pd.concat(
[last_strategy_account, today_strategy_account])
print(strategy_account_new)
next_tradeday = str(next_tradeday).replace('-', '')
strategy_account_new.to_csv(strategy_fold_path + name + '_' +
next_tradeday + '.csv',
encoding='gbk')
strategy_account_all_today.append(today_strategy_account)
strategy_account_all_dict[strategy_name] = strategy_account_new
strategy_account_all_today = pd.concat(
strategy_account_all_today).sort_values(['策略名称', 'date'])
strategy_state = []
strategy_net_all_dict = {}
profolio_asset_dict = {}
for strategy_name in strategy_account_all_dict:
df = strategy_account_all_dict[strategy_name]
df['asset'] = df['cash'] + df['position_value']
df['net'] = df['asset'] / df['asset'].tolist()[0]
net_lst = df['net'].tolist()
strategy_net_all_dict[strategy_name] = net_lst
profolio_asset_dict[strategy_name] = df['asset'].tolist()
today_profit = df['asset'].tolist()[-1] - df['asset'].tolist()[-2]
today_ret = net_lst[-1] / net_lst[-2] - 1
today_net = net_lst[-1]
total_ret = today_net - 1
annR = annROR(net_lst, 1)
sharp = yearsharpRatio(net_lst, 1)
max_retrace = maxRetrace(net_lst, 1)
strategy_state.append([
strategy_name, calen[0], calen[-1], today_net, total_ret, annR,
sharp, max_retrace, today_profit, today_ret
])
profolio_asset_dict['date'] = calen
profolio_asset_df = pd.DataFrame(profolio_asset_dict).set_index(['date'])
profolio_asset_df['组合资产'] = profolio_asset_df.sum(axis=1)
profolio_asset_df['组合'] = profolio_asset_df['组合资产'] / profolio_asset_df[
'组合资产'].tolist()[0]
net_lst = profolio_asset_df['组合'].tolist()
strategy_net_all_dict['组合'] = net_lst
today_profit = profolio_asset_df['组合资产'].tolist(
)[-1] - profolio_asset_df['组合资产'].tolist()[-2]
today_ret = net_lst[-1] / net_lst[-2] - 1
today_net = net_lst[-1]
total_ret = today_net - 1
annR = annROR(net_lst, 1)
sharp = yearsharpRatio(net_lst, 1)
max_retrace = maxRetrace(net_lst, 1)
strategy_state.append([
'组合', calen[0], calen[-1], today_net, total_ret, annR, sharp,
max_retrace, today_profit, today_ret
])
strategy_state_df = pd.DataFrame(strategy_state,
columns=[
'策略名称', '开始日期', '结束日期', '净值', '累计收益',
'年化收益', '夏普', '最大回撤', '当日盈亏', '当日收益'
])
strategy_state_df = strategy_state_df.merge(win_df, on=['策略名称'])
asset_df = pd.DataFrame(strategy_net_all_dict)
asset_df['date'] = calen
asset_df['date'] = pd.to_datetime(asset_df['date'])
asset_df['沪深300'] = index_hq_net_lst
title_str = '策略净值曲线'
asset_df.set_index(['date']).plot()
plt.rcParams['font.sans-serif'] = ['SimHei']
plt.title(title_str)
plt.savefig(fold_path + 'fig/' + 'net_' + calen[-1] + '.png')
return strategy_account_all_today, hold_strategy_today, strategy_state_df, trade_base_date_all_df
def get_account_state(fold_path, file_lst, calen):
net_df = []
for date in calen:
for i in file_lst:
if date in i and 'account' in i:
df_ = pd.read_csv(fold_path + i, encoding='gbk') \
.loc[:, ['交易日', '总资产']]
net_df.append(df_)
break
net_df = pd.concat(net_df).sort_values(['交易日'])
print(net_df)
asset_lst = net_df['总资产'].tolist()
net_lst = [i / asset_lst[0] for i in asset_lst]
data_daily = copy.deepcopy(net_df).rename(columns={'交易日': '日期'})
data_daily['账户净值'] = net_lst
data_daily['time'] = data_daily['日期'].apply(
lambda x: pd.to_datetime(str(x)))
data_daily.index = data_daily['time']
data_daily = data_daily.drop(['time'], axis=1)
data_week = data_daily.resample('W').last()
data_week['开始日期'] = data_daily['日期'].resample('W').first()
data_week['结束日期'] = data_daily['日期'].resample('W').last()
data_week['总资产'] = data_daily['总资产'].resample('W').last()
data_week['账户净值'] = data_daily['账户净值'].resample('W').last()
data_week['周度利润'] = data_week['总资产'] - data_week['总资产'].shift(1)
data_week['周度收益'] = data_week['周度利润'] / data_week['总资产'].shift(1)
data_week['周度利润'] = data_week['周度利润'].fillna(
value=data_week['总资产'].tolist()[0] - account_init)
data_week['周度收益'] = data_week['周度收益'].fillna(
value=data_week['周度利润'].tolist()[0] / account_init)
data_month = data_daily.resample('M').last()
data_month['日期'] = data_daily['日期'].resample('M').last()
data_month['总资产'] = data_daily['总资产'].resample('M').last()
data_month['账户净值'] = data_daily['账户净值'].resample('M').last()
data_month['月度利润'] = data_month['总资产'] - data_month['总资产'].shift(1)
data_month['月度收益'] = data_month['月度利润'] / data_month['总资产'].shift(1)
data_month['月度利润'] = data_month['月度利润'].fillna(
value=data_month['总资产'].tolist()[0] - account_init)
data_month['月度收益'] = data_month['月度收益'].fillna(
value=data_month['总资产'].tolist()[0] / account_init - 1)
account_month_week_state = pd.DataFrame(
[[
'月度统计',
len(data_month), data_month['月度利润'].max(),
data_month['月度利润'].min(), data_month['月度收益'].max(),
data_month['月度收益'].min()
],
[
'周度统计',
len(data_week), data_week['周度利润'].max(), data_week['周度利润'].min(),
data_week['周度收益'].max(), data_week['周度收益'].min()
]],
columns=['账户统计', '周期数', '最高利润', '最低利润', '最高收益率', '最低收益率'])
account_month_week_state['最高利润'] = account_month_week_state['最高利润'].apply(
lambda x: '%.0f' % x)
account_month_week_state['最低利润'] = account_month_week_state['最低利润'].apply(
lambda x: '%.0f' % x)
account_month_week_state['最高收益率'] = account_month_week_state[
'最高收益率'].apply(lambda x: '%.2f%%' % (x * 100))
account_month_week_state['最低收益率'] = account_month_week_state[
'最低收益率'].apply(lambda x: '%.2f%%' % (x * 100))
data_week['开始日期'] = data_week['开始日期'].apply(lambda x: str(x)[:10])
data_week['结束日期'] = data_week['结束日期'].apply(lambda x: str(x)[:10])
data_week['总资产'] = data_week['总资产'].apply(lambda x: '%.0f' % x)
data_week['账户净值'] = data_week['账户净值'].apply(lambda x: '%.4f' % x)
data_week['周度利润'] = data_week['周度利润'].apply(lambda x: '%.0f' % x)
data_week['周度收益'] = data_week['周度收益'].apply(lambda x: '%.2f%%' % (x * 100))
data_week = data_week[['开始日期', '结束日期', '总资产', '账户净值', '周度利润', '周度收益']]
data_month['总资产'] = data_month['总资产'].apply(lambda x: '%.0f' % x)
data_month['账户净值'] = data_month['账户净值'].apply(lambda x: '%.4f' % x)
data_month['月度利润'] = data_month['月度利润'].apply(lambda x: '%.0f' % x)
data_month['月度收益'] = data_month['月度收益'].apply(lambda x: '%.2f%%' %
(x * 100))
data_month['月份'] = data_month['日期'].apply(lambda x: str(x)[:6])
data_month = data_month[['月份', '总资产', '账户净值', '月度利润', '月度收益']]
state_sdate = calen[0]
state_edate = calen[-1]
annR = annROR(net_lst, 1)
sharp = yearsharpRatio(net_lst, 1)
max_retrace = maxRetrace(net_lst, 1)
total_ret = net_lst[-1] - 1
today_ret = net_lst[-1] / net_lst[-2] - 1
print('sharp:%s' % sharp)
print('annR:%s' % annR)
print('max_retrace:%s' % max_retrace)
state_value_lst = []
state_value_lst.append([
state_sdate, state_edate, asset_lst[0], asset_lst[-1], total_ret,
today_ret, annR, sharp, max_retrace
])
df_today = pd.DataFrame(state_value_lst,
columns=[
'开始日期', '结束日期', '初始资产', '当前资产', '总收益', '当日收益',
'年化收益', '夏普', '最大回撤'
])
df_today['初始资产'] = df_today['初始资产'].apply(lambda x: '%.0f' % x)
df_today['当前资产'] = df_today['当前资产'].apply(lambda x: '%.0f' % x)
df_today['夏普'] = df_today['夏普'].apply(lambda x: '%.2f' % x)
df_today['年化收益'] = df_today['年化收益'].apply(lambda x: '%.2f%%' % (x * 100))
df_today['总收益'] = df_today['总收益'].apply(lambda x: '%.2f%%' % (x * 100))
df_today['当日收益'] = df_today['当日收益'].apply(lambda x: '%.2f%%' % (x * 100))
df_today['最大回撤'] = df_today['最大回撤'].apply(lambda x: '%.2f%%' % (x * 100))
return df_today, data_week, data_month, account_month_week_state
def get_strategy_account(foldPath, file_lst, end_date, strategy_name_lst,
calen, init_asset):
file_name_today = get_file_name(file_lst, end_date, 'all_today')
all_today = pd.read_excel(foldPath+file_name_today, encoding='gbk') \
.loc[:, ['策略名称', '总买额', '总卖额', '净买额', '手续费', '盈亏']].dropna().set_index(['策略名称'])
account = []
asset_df = pd.DataFrame(calen, columns=['date'])
for strategy_name in strategy_name_lst:
file_name_history = get_file_name(file_lst, end_date,
strategy_name + '_history')
df_today = pd.read_excel(foldPath+file_name_history, encoding='gbk') \
.loc[:, ['策略名称', '日期', '当日成交额', '当日盈亏', '累计盈亏']]
china_name = df_today['策略名称'].tolist()[0]
df_today['日期'] = df_today['日期'].apply(lambda x: str(x)[:8])
df_today = df_today[df_today['日期'] > calen[0]]
asset_lst = [init_asset]
asset_upgrate = init_asset
for i in range(1, len(calen) - 1):
date = calen[i]
df_ = df_today[df_today['日期'] == date]
if len(df_) > 0:
asset_upgrate = asset_upgrate + df_['当日盈亏'].tolist()[0]
asset_lst.append(asset_upgrate)
today_asset = all_today.loc[china_name]['盈亏'] + asset_upgrate
asset_lst.append(today_asset)
net_lst = [i / asset_lst[0] for i in asset_lst]
annR = annROR(net_lst, 1)
sharp = yearsharpRatio(net_lst, 1)
max_retrace = maxRetrace(net_lst, 1)
total_ret = net_lst[-1] - 1
today_ret = net_lst[-1] / net_lst[-2] - 1
account.append([
china_name, init_asset, today_asset, total_ret, today_ret, annR,
sharp, max_retrace
])
asset_df[china_name] = net_lst
asset_df[strategy_name] = asset_lst
asset_df['date'] = pd.to_datetime(asset_df['date'])
asset_df_value = asset_df.set_index(['date']).ix[:, strategy_name_lst]
asset_df_value['组合'] = asset_df_value.sum(axis=1)
asset_df_value[
'组合净值'] = asset_df_value['组合'] / asset_df_value['组合'].tolist()[0]
asset_df['组合'] = asset_df_value['组合净值'].tolist()
title_str = '策略净值曲线'
name_lst = copy.deepcopy(strategy_china_name_lst)
name_lst.append('组合')
asset_df.set_index(['date']).ix[:, name_lst].plot()
plt.rcParams['font.sans-serif'] = ['SimHei']
plt.title(title_str)
plt.savefig(fold_path + 'fig/' + 'net_' + end_date + '.png')
net_lst = asset_df_value['组合净值'].tolist()
today_asset = asset_df_value['组合'].tolist()[-1]
total_ret = net_lst[-1] - 1
today_ret = net_lst[-1] / net_lst[-2] - 1
annR = annROR(net_lst, 1)
sharp = yearsharpRatio(net_lst, 1)
max_retrace = maxRetrace(net_lst, 1)
account.append([
'组合', asset_df_value['组合'].tolist()[0], today_asset, total_ret,
today_ret, annR, sharp, max_retrace
])
account_df = pd.DataFrame(
account,
columns=['策略名称', '初始资产', '当前资产', '总收益', '当日收益', '年化收益', '夏普', '最大回撤'])
all_today['总买额'] = all_today['总买额'].apply(lambda x: '%.0f' % x)
all_today['总卖额'] = all_today['总卖额'].apply(lambda x: '%.0f' % x)
all_today['净买额'] = all_today['净买额'].apply(lambda x: '%.0f' % x)
all_today['手续费'] = all_today['手续费'].apply(lambda x: '%.2f' % x)
all_today['当日盈亏'] = all_today['盈亏'].apply(lambda x: '%.0f' % x)
all_today['策略名称'] = all_today.index
# account_df['开始日期'] = calen[0]
# account_df['结束日期'] = calen[-1]
account_df['当前资产'] = account_df['当前资产'].apply(lambda x: '%.0f' % x)
account_df['夏普'] = account_df['夏普'].apply(lambda x: '%.2f' % x)
account_df['年化收益'] = account_df['年化收益'].apply(lambda x: '%.2f%%' %
(x * 100))
account_df['总收益'] = account_df['总收益'].apply(lambda x: '%.2f%%' % (x * 100))
account_df['当日收益'] = account_df['当日收益'].apply(lambda x: '%.2f%%' %
(x * 100))
account_df['最大回撤'] = account_df['最大回撤'].apply(lambda x: '%.2f%%' %
(x * 100))
print(account_df)
return all_today[['策略名称', '总买额', '总卖额', '净买额', '手续费',
'当日盈亏']], account_df[[
'策略名称', '初始资产', '当前资产', '总收益', '当日收益', '年化收益', '夏普',
'最大回撤'
]]
except:
data_ori = stock_price(asset, 'daily', '2010-01-01', '2020-12-30')
try:
dataFactory = DataFactory(asset, factor_lst, future_period, s_data_date, e_data_date, data_ori)
dataset = dataFactory.get_dataset_close()
data_set = dataset[dataset.index >= train_s_date]
print(asset, '特征因子：', feature_lst, '训练窗口:', train_days, '测试窗口:', test_days)
strategy_obj = HmmStrategy(
asset=asset, s_date=s_date, max_states=max_states, leverage=1, fee=0.001, data_set=data_set,
cols_features=feature_lst, price_name='close', type='stock', train_days=train_days, test_days=test_days,
insample=False)
annR, sharp, max_retrace, ret_df = strategy_obj.run_outsample()
f = plt.figure(figsize=(15, 8))
plt.rcParams['font.sans-serif'] = ['SimHei']
sharp_close = yearsharpRatio(ret_df.net_close.tolist(), 1)
# Plot return
# ax1 = f.add_subplot(111)
# ax1.plot(ret_df.net, 'blue', label='net_hmm')
# ax1.plot(ret_df.net_close, 'red', label='net_close')
# ax1.legend()
# ax1.set_title(
# '%s特征因子%s训练窗口%s测试窗口%s净值曲线' % (asset, feature_lst, train_days, test_days))
# ax1.set_xlabel('Time')
# ax1.set_ylabel('Value')
# plt.show()
print('夏普比率: ' + str(np.around(sharp, 2))+'行情夏普比率: ' + str(np.around(sharp_close, 2))+
'年化收益: ' + str(np.around(annR, 2))+'最大回撤: ' + str(np.around(max_retrace, 2)))
period_param_state.append([
asset, feature_lst, train_days, test_days, annR, sharp, max_retrace, sharp_close])
def calculate_strategy_net(calen, next_tradeday, trade_fold_path, trade_file_lst, strategy_fold_path, account_hold, strategy_name_lst):
trade_detail_all = []
for date in calen:
trade_detail = get_trade_detail(trade_fold_path, trade_file_lst, date)
trade_detail_all.append(trade_detail)
trade_detail_all = pd.concat(trade_detail_all)
trade_state_all = get_trade_state(trade_detail_all)[
['策略名称', '证券代码', '证券名称', '操作', '成交均价', '成交数量', '成交金额', '手续费', '日期']]
trade_state_all['bs'] = trade_state_all['操作'].apply(lambda x: trans_to_bs(x))
trade_state_all['持仓'] = trade_state_all['成交数量'] * trade_state_all['bs']
trade_state_all['成交金额'] = -trade_state_all['成交金额'] * trade_state_all['bs']
trade_state_today = trade_state_all[trade_state_all['日期'] == calen[-1]]
hold_df_firstday = trade_state_all[(trade_state_all['日期'] == calen[0]) & (trade_state_all['bs'] == 1)]
hold_df_back = trade_state_all[(trade_state_all['日期'] > calen[0])]
hold_df = pd.concat([hold_df_firstday, hold_df_back])[['策略名称', '证券代码', '证券名称', '持仓']]\
.groupby(['策略名称', '证券代码', '证券名称']).sum().reset_index(drop=False)
print('==============hold_df')
print(hold_df)
hold_today = hold_df[hold_df['持仓'] > 0]
account_hold_ = account_hold[['证券代码', '持股', '最新价']]
account_hold_lst = account_hold['证券代码'].tolist()
account_hold_ = account_hold_.set_index(['证券代码'])
print('==============hold_today')
print(hold_today)
print('==============account_hold_')
print(account_hold_)
hold_ret_lst = []
for strategy_name, group in hold_today.groupby(['策略名称']):
for id, row_ in group.iterrows():
code = row_['证券代码']
if code not in account_hold_lst:
continue
hold = row_['持仓']
price = account_hold_.loc[code]['最新价']
hold = min(hold, account_hold_.loc[code]['持股'])
hold_ret_lst.append([strategy_name, code, row_['证券名称'], hold, price])
hold_strategy_today = pd.DataFrame(hold_ret_lst, columns=['策略名称', '证券代码', '证券名称', '持仓', '最新价'])
print('==============hold_strategy_today')
print(hold_strategy_today)
strategy_account_all_today = []
strategy_account_all_dict = {}
for strategy_name in strategy_name_lst:
name = strategy_dict[strategy_name]
hold = hold_strategy_today[hold_strategy_today['策略名称'] == strategy_name]
hold.index = hold['证券代码']
trade_state = trade_state_today[trade_state_today['策略名称'] == strategy_name]
last_strategy_account = pd.read_csv(
strategy_fold_path + name + '_' + calen[-1] + '.csv', encoding='gbk')[
['策略名称', 'date', 'cash', 'position_value']]
last_strategy_account['date'] = last_strategy_account['date'].apply(lambda x: str(x))
last_strategy_account = last_strategy_account.set_index(['date'])
#print(type(last_strategy_account['date'].tolist()[0]))
last_cash = last_strategy_account.loc[calen[-2]]['cash']
today_cash = last_cash + trade_state['成交金额'].sum() - trade_state['手续费'].sum()
today_hold = 0
print(hold)
if len(hold) > 0:
for code in hold['证券代码'].tolist():
price = hold.loc[code]['最新价']
position = hold.loc[code]['持仓']
today_hold = today_hold + price * position
today_strategy_account = pd.DataFrame([[strategy_name, str(calen[-1]), today_cash, today_hold]],
columns=['策略名称', 'date', 'cash', 'position_value'])
print(today_strategy_account)
last_strategy_account = last_strategy_account.reset_index(drop=False)[['策略名称', 'date', 'cash', 'position_value']]
print(last_strategy_account)
strategy_account_new = pd.concat([last_strategy_account, today_strategy_account])
print(strategy_account_new)
next_tradeday = str(next_tradeday).replace('-', '')
strategy_account_new.to_csv(strategy_fold_path + name + '_' + next_tradeday + '.csv', encoding='gbk')
strategy_account_all_today.append(today_strategy_account)
strategy_account_all_dict[strategy_name] = strategy_account_new
strategy_account_all_today = pd.concat(strategy_account_all_today).sort_values(['策略名称', 'date'])
strategy_state = []
strategy_net_all_dict = {}
profolio_asset_dict = {}
for strategy_name in strategy_account_all_dict:
df = strategy_account_all_dict[strategy_name]
df['asset'] = df['cash'] + df['position_value']
df['net'] = df['asset'] / df['asset'].tolist()[0]
net_lst = df['net'].tolist()
strategy_net_all_dict[strategy_name] = net_lst
profolio_asset_dict[strategy_name] = df['asset'].tolist()
today_profit = df['asset'].tolist()[-1] - df['asset'].tolist()[-2]
today_ret = net_lst[-1] / net_lst[-2] - 1
today_net = net_lst[-1]
total_ret = today_net - 1
annR = annROR(net_lst, 1)
sharp = yearsharpRatio(net_lst, 1)
max_retrace = maxRetrace(net_lst, 1)
strategy_state.append([strategy_name, calen[0], calen[-1], today_net, total_ret, annR, sharp, max_retrace, today_profit, today_ret])
profolio_asset_dict['date'] = calen
profolio_asset_df = pd.DataFrame(profolio_asset_dict).set_index(['date'])
profolio_asset_df['组合资产'] = profolio_asset_df.sum(axis=1)
profolio_asset_df['组合'] = profolio_asset_df['组合资产'] / profolio_asset_df['组合资产'].tolist()[0]
net_lst = profolio_asset_df['组合'].tolist()
strategy_net_all_dict['组合'] = net_lst
today_profit = profolio_asset_df['组合资产'].tolist()[-1] - profolio_asset_df['组合资产'].tolist()[-2]
today_ret = net_lst[-1] / net_lst[-2] - 1
today_net = net_lst[-1]
total_ret = today_net - 1
annR = annROR(net_lst, 1)
sharp = yearsharpRatio(net_lst, 1)
max_retrace = maxRetrace(net_lst, 1)
strategy_state.append(
['组合', calen[0], calen[-1], today_net, total_ret, annR, sharp, max_retrace, today_profit, today_ret])
strategy_state_df = pd.DataFrame(strategy_state, columns=['策略名称', '开始日期', '结束日期', '净值', '累计收益', '年化收益', '夏普', '最大回撤', '当日盈亏', '当日收益'])
asset_df = pd.DataFrame(strategy_net_all_dict)
asset_df['date'] = calen
asset_df['date'] = pd.to_datetime(asset_df['date'])
title_str = '策略净值曲线'
asset_df.set_index(['date']).plot()
plt.rcParams['font.sans-serif'] = ['SimHei']
plt.title(title_str)
plt.savefig(fold_path + 'fig/' + 'net_' + calen[-1] + '.png')
return strategy_account_all_today, hold_strategy_today, strategy_state_df
net_df_all = net_df_all.fillna(0)
net_df_all = net_df_all.set_index(['date_time'])
net_df_all['chg'] = net_df_all.mean(axis=1)
# print(net_df_all)
net_df_all['net'] = (1 + net_df_all['chg']).cumprod()
# print(net_df_all)
# net_df_all['date_time'] = net_df_all.index
close_df = pd.read_csv(resualt_path + 'net_param/%s_%s_%s.csv' %
(asset, 240, 60),
encoding='gbk')[['date_time', 'close']]
net_df_all = net_df_all.merge(close_df, on=['date_time'])
net_df_all['close_net'] = net_df_all['close'] / net_df_all[
'close'].tolist()[0]
net_lst = net_df_all['net'].tolist()
annR = annROR(net_lst, 1)
sharp = yearsharpRatio(net_lst, 1)
max_retrace = maxRetrace(net_lst, 1)
print('%s: 夏普：%s 年化收益：%s 最大回撤：%s' % (asset, sharp, annR, max_retrace))
net_df_all['date_time'] = pd.to_datetime(net_df_all['date_time'])
net_df_all = net_df_all.set_index(['date_time'])
title_str = '%s 夏普 %.2f 年化收益 %.2f 最大回撤 %.2f' % (
asset, sharp, 100 * annR, 100 * max_retrace)
net_df_all.ix[:, ['net', 'close_net']].plot()
plt.rcParams['font.sans-serif'] = ['SimHei']
plt.title(title_str)
# plt.savefig(fold + 'fig/' + start_day + '_profolio_' + pos_method + '.png')
plt.show()
# train_days_lst = [i for i in range(480, 30, -40)]
# test_days_lst = [i for i in range(20, 250, 20)]
# period_param_state = []
