def is_stop(self, code, date):
"""
判断是否需要止盈，盈利超过指定值，则止盈
:param code: 股票代码
:param date: 日期
:return: True - 止盈，False - 不止盈
"""
holding_stock = self.account.get_holding(code)
print('止盈判断：%s' % code, flush=True)
print(holding_stock, flush=True)
dm = DataModule()
if holding_stock is not None:
df_daily = dm.get_k_data(code,
autype=None,
begin_date=date,
end_date=date)
if df_daily.index.size > 0:
df_daily.set_index(['date'], 1, inplace=True)
profit = (holding_stock['volume'] * df_daily.loc[date]['close'] - holding_stock['cost']) \
* 100 / holding_stock['cost']
return profit >= self.max_profit
return False
class TrackingStopLoss(BaseStopLoss):
def __init__(self, account, max_loss):
BaseStopLoss.__init__(self, account)
self.max_loss = max_loss
self.dm = DataModule()
def update_holding(self, code, date):
"""
更新持仓股的最高价
:param code:
:param date:
:return:
"""
df_daily = self.dm.get_k_data(code,
autype='hfq',
begin_date=date,
end_date=date)
if df_daily.index.size > 0:
df_daily.set_index(['date'], 1, inplace=True)
close = df_daily.loc[date]['close']
holding = self.account.get_holding(code)
if 'highest' not in holding or holding['highest'] < close:
holding['highest'] = close
self.account.update_holding(code, holding)
def is_stop(self, code, date):
"""
判断股票在当前日期是否需要止损
:param code: 股票代码
:param date: 日期
:return: True - 止损， False - 不止损
"""
holding_stock = self.account.get_holding(code)
print('止损判断：%s' % code, flush=True)
print(holding_stock, flush=True)
if holding_stock is not None:
df_daily = self.dm.get_k_data(code,
autype='hfq',
begin_date=date,
end_date=date)
if df_daily.index.size > 0:
df_daily.set_index(['date'], 1, inplace=True)
close = df_daily.loc[date]['close']
# 计算回落的百分比
if 'highest' in holding_stock and close < holding_stock[
'highest']:
profit = (close - holding_stock['highest']) \
* 100 / holding_stock['highest']
print('止损判断：收盘价：%6.2f, 历史最高：%6.2f， 回撤：%5.2f' %
(close, holding_stock['highest'], profit),
flush=True)
return (profit < 0) & (abs(profit) >= abs(self.max_loss))
return False
def compute(self, begin_date, end_date):
"""
计算指定时间段内所有股票的该因子的值，并保存到数据库中
:param begin_date: 开始时间
:param end_date: 结束时间
"""
dm = DataModule()
frc = FinanceReportCrawler()
code_report_dict = frc.get_code_reports()
codes = set(code_report_dict.keys())
for code in codes:
dailies = dm.get_k_data(code, autype=None, begin_date=begin_date, end_date=end_date)
# 如果没有合适的数据
if dailies.index.size == 0:
continue
# 业绩报告列表
reports = code_report_dict[code]
dailies.set_index(['date'], inplace=True)
update_requests = []
for current_date in dailies.index:
# 用来保存最后一个公告日期小于等于当前日期的财报
last_report = None
for report in reports:
announced_date = report['announced_date']
# 如果公告日期大于当前调整日，则结束循环
if announced_date > current_date:
break
last_report = report
# 如果找到了正确时间范围的年报， 则计算PE
if last_report is not None:
pe = dailies.loc[current_date]['close'] / last_report['eps']
pe = round(pe, 3)
print('%s, %s, %s, eps: %5.2f, pe: %6.2f' %
(code, current_date, last_report['announced_date'], last_report['eps'], pe),
flush=True)
update_requests.append(
UpdateOne(
{'code': code, 'date': current_date},
{'$set': {'code': code, 'date': current_date, 'pe': pe}}, upsert=True))
if len(update_requests) > 0:
save_result = self.collection.bulk_write(update_requests, ordered=False)
print('股票代码: %s, 因子: %s, 插入：%4d, 更新: %4d' %
(code, self.name, save_result.upserted_count, save_result.modified_count), flush=True)
def __init__(self, strategy_option, begin_date=None, end_date=None):
self.strategy_option = strategy_option
if begin_date is None:
self.begin_date = self.strategy_option.begin_date()
else:
self.begin_date = begin_date
if end_date is None:
self.end_date = self.strategy_option.end_date()
else:
self.end_date = end_date
self.dm = DataModule()
self.code_daily_cache = dict()
def __init__(self, strategy_option, begin_date=None, end_date=None):
self.strategy_option = strategy_option
if begin_date is None:
self.begin_date = self.strategy_option.begin_date()
else:
self.begin_date = begin_date
if end_date is None:
self.end_date = self.strategy_option.end_date()
else:
self.end_date = end_date
self.dm = DataModule()
self.code_daily_cache = dict()
self.logger = QuantLogger('backtest')
self.candidatesLogger = QuantLogger('candidates')
def train_model(user_image: Image, style_image: Image):
"""Trains a Deep Learning model to extract the stylings
from `style_image` and applies them onto `user_image`
then returns `user_image`.
Args:
user_image (Image): Image you want to apply styles onto.
style_image (Image): Image you want to extract styles from.
Returns:
user_image (Image): `user_image` with styling applied.
"""
image_processor = ImageProcessor(maximum_image_size=(512, 512))
print(f"user_image.size: {user_image.size} | style_image.size: {style_image.size}")
image_size = image_processor.get_common_image_size(user_image, style_image)
# device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
device = "cpu"
user_image = image_processor.prepare_images(user_image, image_size)
style_image = image_processor.prepare_images(style_image, image_size)
normalization_mean = torch.tensor([0.485, 0.456, 0.406]).to(image_processor.device)
normalization_std = torch.tensor([0.229, 0.224, 0.225]).to(image_processor.device)
datamodule = DataModule(user_image, style_image)
model = StyleTransferModel(
user_image=user_image,
style_image=style_image,
normalization_mean=normalization_mean,
normalization_std=normalization_std,
)
cnn = models.vgg19(pretrained=True).features.to(device).eval()
model._build_model_and_loss_functions(base_cnn_model=cnn)
print("Training...")
tb_logger = pl_loggers.TensorBoardLogger("logs/")
trainer_params = {"max_epochs": 6000}
if device != "cpu":
print("Using a gpu!")
trainer_params["gpus"] = 1
trainer = pl.Trainer(**trainer_params)
trainer.fit(model, datamodule)
sample = user_image.to(device)
model.to(device)
image = model(sample)
image = image_processor.save_image(
image, "output.png", "Sample Output", display_image=True
)
return image
def compute(self, begin_date, end_date):
codes = get_all_codes()
dm = DataModule()
for code in codes:
df_daily = dm.get_k_data(code, begin_date=begin_date, end_date=end_date)
if df_daily.index.size == 0:
continue
# 当日放量下跌
df_daily['change'] = df_daily['close'] - df_daily['pre_close']
df_daily = df_daily[df_daily['change'] < 0]
df_daily['last_volume'] = df_daily['volume'].shift(1)
df_daily.dropna(inplace=True)
df_daily['volume_change'] = round(df_daily['volume']/df_daily['last_volume'], 2)
df_daily = df_daily[df_daily['volume_change'] > 1.5]
# 收的阴线实体不大
df_daily['entity'] = round(abs((df_daily['open'] -df_daily['close'])) * 100/df_daily['close'], 2)
df_daily = df_daily[df_daily['entity'] < 3]
# 大部分时间在昨日之上运行
df_daily.set_index(['date'], 1, inplace=True)
update_requests = []
for date in df_daily.index:
# 大部分时间在昨日之上运行
pre_close = df_daily.loc[date]['pre_close']
df_minute = dm.get_k_data(code, period='M1', begin_date=date, end_date=date)
df_minute = df_minute[df_minute['close'] > pre_close]
if df_minute.index.size > 150:
update_requests.append(UpdateOne({
'code': code, 'date': date},
{'$set': {'code': code, 'date': date}},
upsert=True))
if len(update_requests) > 0:
save_result = self.collection.bulk_write(update_requests, ordered=False)
print('股票代码: %s, 因子: %s, 插入：%4d, 更新: %4d' %
(code, self.name, save_result.upserted_count, save_result.modified_count), flush=True)
def is_stop(self, code, date):
"""
判断股票在当前日期是否需要止损
:param code: 股票代码
:param date: 日期
:return: True - 止损， False - 不止损
"""
holding_stock = self.account.get_holding(code)
print('止损判断：%s' % code, flush=True)
print(holding_stock, flush=True)
dm = DataModule()
if holding_stock is not None:
df_daily = dm.get_k_data(code, autype=None, begin_date=date, end_date=date)
if df_daily.index.size > 0:
df_daily.set_index(['date'], 1, inplace=True)
profit = (holding_stock['volume'] * df_daily.loc[date]['close'] - holding_stock['cost']) \
* 100 / holding_stock['cost']
return (profit < 0) & (abs(profit) >= abs(self.max_loss))
return False
class Backtest:
def __init__(self, strategy_option, begin_date=None, end_date=None):
self.strategy_option = strategy_option
if begin_date is None:
self.begin_date = self.strategy_option.begin_date()
else:
self.begin_date = begin_date
if end_date is None:
self.end_date = self.strategy_option.end_date()
else:
self.end_date = end_date
self.dm = DataModule()
self.code_daily_cache = dict()
self.logger = QuantLogger('backtest')
self.candidatesLogger = QuantLogger('candidates')
def start(self):
"""
策略回测。结束后打印出收益曲线(沪深300基准)、年化收益、最大回撤
"""
# 初始总资金
initial_capital = self.strategy_option.capital()
# 初始现金
cash = initial_capital
# 单只股票仓位上限
single_position = self.strategy_option.single_position()
# 从获取策略配置中获取股票池
stock_pool = self.strategy_option.stock_pool()
# 保存持仓股的日期
account = Account()
# 获取卖出信号
sell_signal = self.strategy_option.sell_signal(account)
# 获取买入信号
buy_signal = self.strategy_option.buy_signal(account)
# 时间为key的净值、收益和同期沪深基准
df_profit = DataFrame(columns=['net_value', 'profit', 'hs300'])
# 获取交易日历，
all_dates = get_trading_dates(begin_date=self.begin_date,
end_date=self.end_date)
# 获取沪深300在统计周期内的第一天的值
hs300_k = self.dm.get_k_data('000300',
index=True,
begin_date=all_dates[0],
end_date=all_dates[0])
hs300_begin_value = hs300_k.loc[hs300_k.index[0]]['close']
# 获取股票池数据
rebalance_dates, date_codes_dict = stock_pool.get_option_stocks()
# 获取止损策略
stop_loss_policy = self.strategy_option.get_stop_loss(account)
# 获取止盈策略
stop_profit_policy = self.strategy_option.get_stop_profit(account)
# 获取加仓策略
add_position_policy = self.strategy_option.get_add_position(account)
# 获取回测周期内股票池内所有股票的收盘价和前收价
all_option_code_set = set()
for rebalance_date in rebalance_dates:
for code in date_codes_dict[rebalance_date]:
all_option_code_set.add(code)
# 缓存股票的日线数据
for code in all_option_code_set:
dailies_df = self.dm.get_k_data(code,
autype=None,
begin_date=self.begin_date,
end_date=self.end_date)
dailies_df.set_index(['date'], inplace=True)
self.code_daily_cache[code] = dailies_df
last_phase_codes = None
this_phase_codes = None
to_be_sold_codes = set()
to_be_bought_codes = set()
last_date = None
# 加仓
to_be_added_signals = dict()
to_be_added_codes = set()
# 按照日期一步步回测
for _date in all_dates:
self.logger.info('开始回测，日期：' + _date)
# 处理复权
account.adjust_holding_volume_at_open(last_date, _date)
# 卖出
if len(to_be_sold_codes) > 0:
sold_codes_tmp = set(to_be_sold_codes)
for code in sold_codes_tmp:
try:
if code in account.holding_codes:
holding_stock = account.get_holding(code)
holding_volume = holding_stock['volume']
sell_price = self.code_daily_cache[code].loc[
_date]['open']
low_limit = self.code_daily_cache[code].loc[_date][
'low_limit']
if sell_price > low_limit:
sell_amount = holding_volume * sell_price
cash += sell_amount
cost = holding_stock['cost']
single_profit = (sell_amount -
cost) * 100 / cost
print('卖出 %s, %6d, %6.2f, %8.2f, %4.2f' %
(code, holding_volume, sell_price,
sell_amount, single_profit))
else:
print(
'当日跌停，无法卖出，股票代码：%s, 日期： %s，价格：%7.2f，跌停价：%7.2f'
% (code, _date, sell_price, low_limit),
flush=True)
# 从持仓股中卖出
account.sell_out(code)
# 从代码列表中删除
to_be_sold_codes.remove(code)
except:
print('卖出时，发生异常：%s, %s' % (code, _date), flush=True)
print('卖出后，现金: %10.2f' % cash)
# 加仓逻辑
add_codes_tmp = set(to_be_added_codes)
for code in add_codes_tmp:
add_signal = to_be_added_signals[code]
try:
if cash > add_signal['position']:
daily = self.code_daily_cache[code].loc[_date]
buy_price = daily['open']
high_limit = daily['high_limit']
if buy_price < high_limit:
volume = int(
int(add_signal['position'] / buy_price) /
100) * 100
buy_amount = buy_price * volume
cash -= buy_amount
print('加仓 %s, %6d, %6.2f, %8.2f' %
(code, volume, buy_price, buy_amount),
flush=True)
# 更新加仓后的持仓股
holding = account.get_holding(code)
holding['cost'] += buy_amount
holding['last_value'] += buy_amount
holding['volume'] += volume
holding['add_times'] += 1
holding['last_buy_hfq_price'] = buy_price * daily[
'au_factor']
account.update_holding(code, holding)
# 从待加仓列表中删除
to_be_added_codes.remove(code)
del to_be_added_signals[code]
else:
print(
'当日涨停，无法加仓，股票代码：%s, 日期： %s，价格：%7.2f，涨停价：%7.2f'
% (code, _date, buy_price, high_limit),
flush=True)
except:
print('加仓时，发生错误：%s, %s' % (code, _date), flush=True)
# 买入
if len(to_be_bought_codes) > 0:
sorted_to_be_bought_list = list(to_be_bought_codes)
sorted_to_be_bought_list.sort()
for code in sorted_to_be_bought_list:
try:
if cash > single_position:
daily = self.code_daily_cache[code].loc[_date]
buy_price = daily['open']
high_limit = daily['high_limit']
if buy_price < high_limit:
volume = int(
int(single_position / buy_price) /
100) * 100
buy_amount = buy_price * volume
cash -= buy_amount
print('买入 %s, %6d, %6.2f, %8.2f' %
(code, volume, buy_price, buy_amount),
flush=True)
# 维护账户的持仓股
account.buy_in(code,
volume=volume,
cost=buy_amount)
# 如果加仓策略不为空，则更新持仓股
if add_position_policy is not None:
holding = account.get_holding(code)
holding[
'last_buy_hfq_price'] = buy_price * daily[
'au_factor']
add_position_policy.update_holding(
code, _date, holding)
else:
print(
'当日涨停，无法买入，股票代码：%s, 日期： %s，价格：%7.2f，涨停价：%7.2f'
% (code, _date, buy_price, high_limit),
flush=True)
except:
print('买入时，发生错误：%s, %s' % (code, _date), flush=True)
print('买入后，现金: %10.2f' % cash)
# 持仓股代码列表
holding_codes = account.holding_codes
# 如果调整日，则获取新一期的股票列表
if _date in rebalance_dates:
# 暂存为上期的日期
if this_phase_codes is not None:
last_phase_codes = this_phase_codes
this_phase_codes = date_codes_dict[_date]
# 找到所有调出股票代码，在第二日开盘时卖出
if last_phase_codes is not None:
out_codes = self.find_out_stocks(last_phase_codes,
this_phase_codes)
for out_code in out_codes:
if out_code in holding_codes:
to_be_sold_codes.add(out_code)
# 检查是否有需要第二天卖出的股票
for holding_code in holding_codes:
if sell_signal.is_match(holding_code, _date):
to_be_sold_codes.add(holding_code)
# 检测止损信号
if stop_loss_policy is not None:
for holding_code in holding_codes:
if stop_loss_policy.is_stop(holding_code, _date):
to_be_sold_codes.add(holding_code)
print('止损，股票：%s' % holding_code, flush=True)
else:
stop_loss_policy.update_holding(holding_code, _date)
# 检测止盈信号
if stop_profit_policy is not None:
for holding_code in holding_codes:
if stop_profit_policy.is_stop(holding_code, _date):
to_be_sold_codes.add(holding_code)
print('止盈，股票：%s' % holding_code, flush=True)
else:
stop_profit_policy.update_holding(holding_code, _date)
print('待卖股票，日期：%s，代码列表：' % _date, to_be_sold_codes, flush=True)
# 检测是否有需要建仓的股票
if add_position_policy is not None:
for holding_code in holding_codes:
add_signal = add_position_policy.get_add_signal(
holding_code, _date)
if add_signal is not None:
to_be_added_signals[holding_code] = add_signal
to_be_added_codes.add(holding_code)
# 检查是否有需要第二天买入的股票
to_be_bought_codes.clear()
if this_phase_codes is not None:
for _code in this_phase_codes:
if _code not in holding_codes and \
buy_signal.is_match(_code, _date):
to_be_bought_codes.add(_code)
self.logger.info('待买股票，日期：%s，代码列表：%s', _date, to_be_bought_codes)
self.candidatesLogger.info('待买股票，日期：%s，代码列表：%s', _date,
to_be_bought_codes)
# 计算总市值
total_value = account.get_total_value(_date)
# 计算总资产
total_capital = total_value + cash
print('收盘后，现金: %10.2f, 总资产: %10.2f' % (cash, total_capital))
# 计算沪深300的增长
hs300_k_current = self.dm.get_k_data('000300',
index=True,
begin_date=_date,
end_date=_date)
hs300_current_value = hs300_k_current.loc[
hs300_k_current.index[0]]['close']
last_date = _date
df_profit.loc[_date] = {
'net_value':
round(total_capital / initial_capital, 2),
'profit':
round(
100 * (total_capital - initial_capital) / initial_capital,
2),
'hs300':
round(
100 * (hs300_current_value - hs300_begin_value) /
hs300_begin_value, 2)
}
# 打印回测收益曲线数值
print('Profit history start')
for index_date in df_profit.index:
print('%s, %6.2f, %6.2f' %
(index_date, df_profit.loc[index_date]['profit'],
df_profit.loc[index_date]['hs300']),
flush=True)
print('Profit history end')
drawdown = self.compute_drawdown(df_profit['net_value'])
annual_profit, sharpe_ratio = self.compute_sharpe_ratio(
df_profit['net_value'])
print('回测结果 %s - %s，年化收益： %7.3f, 最大回撤：%7.3f, 夏普比率：%4.2f' %
(self.begin_date, self.end_date, annual_profit, drawdown,
sharpe_ratio))
df_profit.plot(title='Backtest Result',
y=['profit', 'hs300'],
kind='line')
plt.show()
def compute_sharpe_ratio(self, net_values):
"""
计算夏普比率
:param net_values: 净值列表
"""
# 总交易日数
trading_days = len(net_values)
# 所有收益的DataFrame
profit_df = DataFrame(columns={'profit'})
# 收益之后，初始化为第一天的收益
profit_df.loc[0] = {'profit': round((net_values[0] - 1) * 100, 2)}
# 计算每天的收益
for index in range(1, trading_days):
# 计算每日的收益变化
profit = (net_values[index] -
net_values[index - 1]) / net_values[index - 1]
profit = round(profit * 100, 2)
profit_df.loc[index] = {'profit': profit}
# 计算标准差
profit_std = pow(profit_df.var()['profit'], 1 / 2)
# 年化收益
annual_profit = self.compute_annual_profit(trading_days,
net_values[-1])
# 夏普比率
sharpe_ratio = (annual_profit - 4.75) / profit_std
return annual_profit, sharpe_ratio
def compute_drawdown(self, net_values):
"""
计算最大回撤
:param net_values: 净值列表
"""
# 最大回撤初始值设为0
max_drawdown = 0
index = 0
# 双层循环找出最大回撤
for net_value in net_values:
for sub_net_value in net_values[index:]:
drawdown = 1 - sub_net_value / net_value
if drawdown > max_drawdown:
max_drawdown = drawdown
index += 1
return max_drawdown
def compute_annual_profit(self, trading_days, net_value):
"""
计算年化收益
"""
annual_profit = 0
if trading_days > 0:
# 计算年数
years = trading_days / 245
# 计算年化收益
annual_profit = pow(net_value, 1 / years) - 1
annual_profit = round(annual_profit * 100, 2)
return annual_profit
def find_out_stocks(self, last_phase_codes, this_phase_codes):
"""
找到上期入选本期被调出的股票，这些股票将必须卖出
:param last_phase_codes: 上期的股票列表
:param this_phase_codes: 本期的股票列表
:return: 被调出的股票列表
"""
out_stocks = []
for code in last_phase_codes:
if code not in this_phase_codes:
out_stocks.append(code)
return out_stocks
def compute(self, begin_date=None, end_date=None):
"""
计算指定日期内的信号
:param begin_date: 开始日期
:param end_date: 结束日期
"""
codes = get_all_codes()
dm = DataModule()
for code in codes:
try:
df_dailies = dm.get_k_data(code,
autype='hfq',
begin_date=begin_date,
end_date=end_date)
if df_dailies.index.size == 0:
continue
# 计算MA10
df_dailies['ma'] = df_dailies['close'].rolling(10).mean()
# 计算当日收盘和MA10的差值
df_dailies['delta'] = df_dailies['close'] - df_dailies['ma']
# 删除不再使用的ma和close列
df_dailies.drop(['ma', 'close'], 1, inplace=True)
# 判断突破类型
index_size = df_dailies.index.size
breaks = [0]
for index in range(1, index_size):
# 如果当前日期为停牌状态，则后面连续11日不参与计算
if df_dailies.loc[
df_dailies.index[index]]['is_trading'] is False:
count = 10
while count > 0:
index += 1
count -= 1
breaks.append(0)
index += 1
last = df_dailies.loc[df_dailies.index[index - 1]]['delta']
current = df_dailies.loc[df_dailies.index[index]]['delta']
# 向上突破设为1，向下突破设为-1，不是突破设为0
break_direction = 1 if last <= 0 < current else -1 if last >= 0 > current else 0
breaks.append(break_direction)
# 设置突破信号
df_dailies['break'] = breaks
# 将日期作为索引
df_dailies.set_index(['date'], 1, inplace=True)
# 删除不再使用的trade_status和delta数据列
df_dailies.drop(['is_trading', 'delta'], 1, inplace=True)
# 只保留突破的日期
df_dailies = df_dailies[df_dailies['break'] != 0]
# 将信号保存到数据库
update_requests = []
for index in df_dailies.index:
doc = {
'code':
code,
'date':
index,
# 方向，向上突破 up，向下突破 down
'direction':
'up' if df_dailies.loc[index]['break'] == 1 else 'down'
}
update_requests.append(
UpdateOne(doc, {'$set': doc}, upsert=True))
if len(update_requests) > 0:
update_result = self.collection.bulk_write(update_requests,
ordered=False)
print('%s, upserted: %4d, modified: %4d' %
(code, update_result.upserted_count,
update_result.modified_count),
flush=True)
except:
traceback.print_exc()
def compute(self, begin_date, end_date):
print(self.name, flush=True)
dm = DataModule()
df_daily = dm.get_k_data()
print(df_daily)
def get_option_stocks(self):
"""
实现股票池选股逻辑，找到指定日期范围的候选股票
条件：0 < PE < 30, 按从小到大排序，剔除停牌后，取前100个；再平衡周期：7个交易日
:return: tuple，再平衡的日期列表，以及一个dict(key: 再平衡日, value: 当期的股票列表)
"""
factor_module = FactorModule()
dm = DataModule()
# 因为上证指数没有停牌不会缺数，所以用它作为交易日历，
szzz_hq_df = dm.get_k_data('000001',
index=True,
begin_date=self.begin_date,
end_date=self.end_date)
all_dates = list(szzz_hq_df['date'])
# 缓存股票和其对应有交易的日期
code_dates_cache = dict()
# 调整日和其对应的股票
rebalance_date_codes_dict = dict()
rebalance_dates = []
# 保存上一期的股票池
last_phase_codes = []
# 所有的交易日数
dates_count = len(all_dates)
# 用再平衡周期作为步长循环
for index in range(0, dates_count, self.interval):
# 当前的调整日
rebalance_date = all_dates[index]
# 获取本期符合条件的备选股票
df_pe = factor_module.get_single_date_factors('pe', rebalance_date)
df_pe.sort_values('pe', ascending=True, inplace=True)
# 只保留小于30的数据
df_pe = df_pe[(0 < df_pe['pe']) & (df_pe['pe'] < 30)]
df_pe.set_index(['code'], inplace=True)
this_phase_option_codes = list(df_pe.index)[0:100]
print(this_phase_option_codes, flush=True)
# 本期入选的股票代码列表
this_phase_codes = []
# 找到在上一期的股票池，但是当前停牌的股票，保留在当期股票池中
if len(last_phase_codes) > 0:
for code in last_phase_codes:
if code not in list(code_dates_cache.keys()):
daily_ks = dm.get_k_data(code,
autype=None,
begin_date=self.begin_date,
end_date=self.end_date)
daily_ks.set_index(['date'], inplace=True)
code_dates_cache[code] = list(daily_ks.index)
if rebalance_date not in code_dates_cache[code]:
this_phase_codes.append(code)
print('上期停牌的股票：', flush=True)
print(this_phase_codes, flush=True)
# 剩余的位置用当前备选股票的
option_size = len(this_phase_option_codes)
if option_size > (100 - len(this_phase_codes)):
this_phase_codes += this_phase_option_codes[
0:100 - len(this_phase_codes)]
else:
this_phase_codes += this_phase_option_codes
# 当期股票池作为下次循环的上期股票池
last_phase_codes = this_phase_codes
# 保存到返回结果中
rebalance_date_codes_dict[rebalance_date] = this_phase_codes
rebalance_dates.append(rebalance_date)
print('当前最终的备选票：%s' % rebalance_date, flush=True)
print(this_phase_codes, flush=True)
return rebalance_dates, rebalance_date_codes_dict
def compute(self, begin_date, end_date):
"""
计算指定时间段内所有股票的该因子的值，并保存到数据库中
:param begin_date: 开始时间
:param end_date: 结束时间
"""
dm = DataModule()
# 获取所有股票
codes = get_all_codes()
for code in codes:
print('计算市盈率, %s' % code)
df_daily = dm.get_k_data(code,
autype=None,
begin_date=begin_date,
end_date=end_date)
if df_daily.index.size > 0:
df_daily.set_index(['date'], 1, inplace=True)
update_requests = []
for date in df_daily.index:
finance_report = DB_CONN['finance_report'].find_one(
{
'code': code,
'report_date': {
'$regex': '\d{4}-12-31'
},
'announced_date': {
'$lte': date
}
},
sort=[('announced_date', DESCENDING)])
if finance_report is None:
continue
# 计算滚动市盈率并保存到daily_k中
eps = 0
if finance_report['eps'] != '-':
eps = finance_report['eps']
# 计算PE
if eps != 0:
pe = round(df_daily.loc[date]['close'] / eps, 3)
print('%s, %s, %s, eps: %5.2f, pe: %6.2f' %
(code, date, finance_report['announced_date'],
finance_report['eps'], pe),
flush=True)
update_requests.append(
UpdateOne({
'code': code,
'date': date
}, {
'$set': {
'code': code,
'date': date,
'pe': pe
}
},
upsert=True))
if len(update_requests) > 0:
save_result = self.collection.bulk_write(update_requests,
ordered=False)
print('股票代码: %s, 因子: %s, 插入：%4d, 更新: %4d' %
(code, self.name, save_result.upserted_count,
save_result.modified_count),
flush=True)
def analyze(self):
# 初始化对数据管理子系统接口的调用
dm = DataModule()
# 初始化对因子管理子系统接口的调用
fm = FactorModule()
# 获取分析周期内的
all_dates = get_trading_dates(self.begin_date, self.end_date)
# 首档和末档，股票代码和后复权价格的Dictionary
top_dailies = dict()
bottom_dailies = dict()
# 暂存上一个调整
last_adjust_date = None
# 设置沪深300的首日值
hs300_k = dm.get_k_data('000300',
index=True,
begin_date=all_dates[0],
end_date=all_dates[0])
self.hs300_first_value = hs300_k.loc[0]['close']
# 计算每日收益
for index in range(0, len(all_dates), self.interval):
adjust_date = all_dates[index]
# 获取因子值，按照指定的顺序排序
df_factor = fm.get_single_date_factors(self.factor, adjust_date)
if df_factor.index.size == 0:
continue
df_factor.sort_values(self.factor,
ascending=self.ascending,
inplace=True)
# 将股票代码设为index
df_factor.set_index(['code'], inplace=True)
# 获取当日所有股票的行情
df_dailies = dm.get_one_day_k_data(autype='hfq', date=adjust_date)
# 将code设为index
df_dailies.set_index(['code'], inplace=True)
# 计算收益
self.compute_profit(last_adjust_date, df_dailies, top_dailies,
bottom_dailies, adjust_date)
# 删除停牌股票
df_dailies = df_dailies[df_dailies['is_trading']]
# 计算每当包含的股票数
total_size = df_dailies.index.size
single_position_count = int(total_size / self.position)
# 调整首档组合
self.adjust_top_position(top_dailies, df_factor, df_dailies,
single_position_count)
# 调整末档组合
self.adjust_bottom_position(bottom_dailies, df_factor, df_dailies,
single_position_count)
# 保存上一个调整日
last_adjust_date = adjust_date
# 生成零投资组合的组合收益
self.profit_df[
'portfolio'] = self.profit_df['top'] - self.profit_df['bottom']
self.draw()
class TrackingStopProfit(BaseStopProfit):
def __init__(self, account, max_profit, profit_drawdown):
BaseStopProfit.__init__(self, account)
self.max_profit = max_profit
self.profit_drawdown = profit_drawdown
self.dm = DataModule()
def update_holding(self, code, date):
"""
更新持仓股的最高市值
:param code:
:param date:
:return:
"""
df_daily = self.dm.get_k_data(code,
autype=None,
begin_date=date,
end_date=date)
if df_daily.index.size > 0:
df_daily.set_index(['date'], 1, inplace=True)
close = df_daily.loc[date]['close']
holding = self.account.get_holding(code)
current_value = holding['volume'] * close
if 'highest_value' in holding.keys():
if current_value > holding['highest_value']:
holding['highest_value'] = current_value
self.account.update_holding(code, holding)
else:
# 判断是否已经达到了最高的收益线
profit = (current_value -
holding['cost']) * 100 / holding['cost']
if profit > self.max_profit:
holding['highest_value'] = current_value
self.account.update_holding(code, holding)
def is_stop(self, code, date):
"""
判断股票在当前日期是否需要止盈，如果当前收益相对于最高收益的回撤已经达到了指定幅度，则止盈
:param code: 股票代码
:param date: 日期
:return: True - 止盈， False - 不止盈
"""
holding_stock = self.account.get_holding(code)
print('止盈判断：%s' % code, flush=True)
print(holding_stock, flush=True)
if holding_stock is not None:
df_daily = self.dm.get_k_data(code,
autype=None,
begin_date=date,
end_date=date)
if df_daily.index.size > 0:
df_daily.set_index(['date'], 1, inplace=True)
current_value = df_daily.loc[date]['close'] * holding_stock[
'volume']
# 计算回落的百分比
if 'highest_value' in holding_stock and current_value < holding_stock[
'highest_value']:
profit = (current_value - holding_stock['highest_value']) \
* 100 / holding_stock['highest_value']
print('止盈判断, profit: %5.2f, drawdown: %5.2f' %
(profit, self.profit_drawdown),
flush=True)
return (profit < 0) & (abs(profit) >= abs(
self.profit_drawdown))
return False
def compute(self, begin_date=None, end_date=None):
"""
计算指定时间段内所有股票的该因子的值，并保存到数据库中
:param begin_date: 开始时间
:param end_date: 结束时间
"""
dm = DataModule()
# 如果没有指定日期范围，则默认为计算当前交易日的数据
if begin_date is None:
begin_date = datetime.now().strftime('%Y-%m-%d')
if end_date is None:
end_date = datetime.now().strftime('%Y-%m-%d')
dates = get_trading_dates(begin_date, end_date)
for date in dates:
# 查询出股票在某一交易日的总股本
df_basics = dm.get_stock_basic_at(date)
if df_basics.index.size == 0:
continue
# 将索引改为code
df_basics.set_index(['code'], 1, inplace=True)
# 查询出股票在某一个交易日的收盘价
df_dailies = dm.get_one_day_k_data(autype=None, date=date)
if df_dailies.index.size == 0:
continue
# 将索引设为code
df_dailies.set_index(['code'], 1, inplace=True)
update_requests = []
for code in df_dailies.index:
try:
# 股价
close = df_dailies.loc[code]['close']
# 总股本
total_shares = df_basics.loc[code]['totals']
# 总市值 = 股价 * 总股本
total_capital = round(close * total_shares, 2)
print('%s, %s, mkt_cap: %15.2f' %
(code, date, total_capital),
flush=True)
update_requests.append(
UpdateOne(
{'code': code, 'date': date},
{'$set': {'code': code, 'date': date, self.name: total_capital}}, upsert=True))
except:
print('计算规模因子时发生异常，股票代码：%s，日期：%s'
% (code, date),
flush=True)
if len(update_requests) > 0:
save_result = self.collection.bulk_write(update_requests, ordered=False)
print('股票代码: %s, 因子: %s, 插入：%4d, 更新: %4d' %
(code, self.name, save_result.upserted_count, save_result.modified_count), flush=True)
def __init__(self):
self.holding = dict()
self.holding_codes = set()
self.dm = DataModule()
def __init__(self, account, strategy_option):
BasePosition.__init__(self, account, strategy_option)
self.dm = DataModule()
class ATRPosition(BasePosition):
def __init__(self, account, strategy_option):
BasePosition.__init__(self, account, strategy_option)
self.dm = DataModule()
def update_holding(self, code, date, updated_holding):
"""
更新持仓股，为新的持仓股增加ATR、加仓次数
:param code: 股票代码
:param date: 日期
:param updated_holding: 更新后的持仓股
"""
try:
existing_holding = self.account.get_holding(code)
print(existing_holding, flush=True)
if existing_holding is not None and 'atr' not in existing_holding:
atr = self.compute_atr(code, date)
print('Computed atr: %s' % str(atr), flush=True)
if atr is not None:
updated_holding['atr'] = atr
updated_holding['add_times'] = 0
self.account.update_holding(code, updated_holding)
except:
print('加仓策略，更新持仓股信息时，发生错误，股票代码：%s，日期：%s，' % (code, date),
flush=True)
traceback.print_exc()
def get_add_signal(self, code, date):
"""
如果符合加仓条件，则计算加仓信号
:param code: 股票代码
:param date: 日期
:return: 加仓信号，如果没有则返回None，正常信号包括：code - 股票代码，position - 仓位
"""
add_signal = None
try:
df_daily = self.dm.get_k_data(code,
autype=None,
begin_date=date,
end_date=date)
if df_daily.index.size > 0:
holding_stock = self.account.get_holding(code)
a = 5 / 0
df_daily.set_index(['date'], 1, inplace=True)
daily = df_daily.loc[date]
if 'atr' in holding_stock and daily['is_trading']:
au_factor = daily['au_factor']
# ATR
atr = holding_stock['atr']
# 已加仓次数
add_times = holding_stock['add_times']
# 末次加仓价格，比较时用的实际价格
last_buy_price = holding_stock['last_buy_hfq_price']
hfq_close = daily['close'] * au_factor
# 最多加仓4次 价格超过上一个加仓点+atr
if add_times < 4 and hfq_close - last_buy_price > atr:
position = self.compute_position(code, date)
add_signal = {'code': code, 'position': position}
except:
self.logger.error('计算加仓信号时，发生错误，股票代码：%s, 日期：%s' % (code, date))
return add_signal
def compute_atr(self, code, date):
"""
计算ATR
:param code: 股票代码
:param date: 日期
:return: ATR的值
"""
minute_cursor = DB_CONN['minute'].find(
{
'code': code,
'time': {
'$lte': date + ' 15:00'
}
},
sort=[('time', DESCENDING)],
projection={
'date': True,
'time': True,
'close': True,
'high': True,
'low': True,
'_id': False
},
limit=61)
minutes = [minute for minute in minute_cursor]
minutes.reverse()
if len(minutes) >= 60:
before_date = (datetime.strptime(date, '%Y-%m-%d') -
timedelta(days=20)).strftime('%Y-%m-%d')
df_daily = self.dm.get_k_data(code,
autype=None,
begin_date=before_date,
end_date=date)
if df_daily.index.size == 0:
return None
df_daily.set_index(['date'], 1, inplace=True)
atr_df = DataFrame(columns=['tr'])
for index in range(1, len(minutes)):
minute = minutes[index]
minute_date = minute['date']
try:
au_factor = df_daily.loc[minute_date]['au_factor']
high = minute['high'] * au_factor
low = minute['low'] * au_factor
pre_close = minutes[index - 1]['close'] * au_factor
atr_df.loc[index] = {
'tr':
max([
high - low,
abs(high - pre_close),
abs(pre_close - low)
])
}
except Exception as e:
self.logger.error(
'计算ATR发生异常，股票代码：%s， 分钟线日期：%s' % (code, minute_date), e)
print(atr_df)
atrs = atr_df['tr'].rolling(60).mean()
self.logger.debug(atrs)
return atrs[60]
else:
print('加仓策略（ATR），分钟线数据不足，股票代码：%s，日期：%s，分钟线数：%2d' %
(code, date, len(minutes)),
flush=True)
return None
def compute_position(self, code, date):
"""
计算应该分配的仓位， 这里采用固定仓位，总资金的0.015
:param code: 股票代码
:param date: 日期
:return:
"""
amount = self.strategy_option.capital() * 0.015
return amount
def statistic_profit(self):
"""
统计股票池的收益
"""
# 设定评测周期
rebalance_dates, codes_dict = self.get_option_stocks()
dm = DataModule()
# 用DataFrame保存收益
df_profit = DataFrame(columns=['profit', 'hs300'])
df_profit.loc[rebalance_dates[0]] = {'profit': 0, 'hs300': 0}
# 获取沪深300在统计周期内的第一天的值
hs300_k = dm.get_k_data('000300',
index=True,
begin_date=rebalance_dates[0],
end_date=rebalance_dates[0])
hs300_begin_value = hs300_k.loc[hs300_k.index[0]]['close']
# 通过净值计算累计收益
net_value = 1
for _index in range(1, len(rebalance_dates) - 1):
last_rebalance_date = rebalance_dates[_index - 1]
current_rebalance_date = rebalance_dates[_index]
# 获取上一期的股票池
codes = codes_dict[last_rebalance_date]
# 统计当前的收益
profit_sum = 0
# 参与统计收益的股票个数
profit_code_count = 0
for code in codes:
daily_ks = dm.get_k_data(code,
autype='hfq',
begin_date=last_rebalance_date,
end_date=current_rebalance_date)
index_size = daily_ks.index.size
# 如果没有数据，则跳过，长期停牌
if index_size == 0:
continue
# 买入价
in_price = daily_ks.loc[daily_ks.index[0]]['close']
# 卖出价
out_price = daily_ks.loc[daily_ks.index[index_size -
1]]['close']
# 股票池内所有股票的收益
profit_sum += (out_price - in_price) / in_price
profit_code_count += 1
profit = round(profit_sum / profit_code_count, 4)
hs300_k_current = dm.get_k_data('000300',
index=True,
begin_date=current_rebalance_date,
end_date=current_rebalance_date)
hs300_close = hs300_k_current.loc[
hs300_k_current.index[0]]['close']
# 计算净值和累积收益
net_value = net_value * (1 + profit)
df_profit.loc[current_rebalance_date] = {
'profit':
round((net_value - 1) * 100, 4),
'hs300':
round((hs300_close - hs300_begin_value) * 100 /
hs300_begin_value, 4)
}
print(df_profit)
# 绘制曲线
df_profit.plot(title='Stock Pool Profit Statistic', kind='line')
# 显示图像
plt.show()
class DailyKBreakMA10Signal:
def __init__(self):
self.dm = DataModule()
def is_k_up_break_ma10(self, code, begin_date, end_date):
"""
判断某只股票在某日是否满足K线上穿10日均线
:param code: 股票代码
:param begin_date: 开始日期
:param end_date: 结束日期
:return: True/False
"""
# 如果没有指定开始日期和结束日期，则直接返回False
if begin_date is None or end_date is None:
return False
daily_ks = self.dm.get_k_data(code,
autype='hfq',
begin_date=begin_date,
end_date=end_date)
# 需要判断两日的K线和10日均线的相对位置，所以如果K线数不满足11个，
# 也就是无法计算两个MA10，则直接返回False
index_size = daily_ks.index.size
if index_size < 11:
return False
# 比较收盘价和MA10的关系
daily_ks['ma'] = daily_ks['close'].rolling(10).mean()
daily_ks['delta'] = daily_ks['close'] - daily_ks['ma']
is_break_up = (daily_ks.loc[daily_ks.index[9]]['delta'] <= 0 <
daily_ks.loc[daily_ks.index[10]]['delta'])
return is_break_up
def is_k_down_break_ma10(self, code, begin_date, end_date):
"""
判断某只股票在某日是否满足K线下穿10日均线
:param code: 股票代码
:param begin_date: 开始日期
:param end_date: 结束日期
:return: True/False
"""
# 如果没有指定开始日期和结束日期，则直接返回False
if begin_date is None or end_date is None:
return False
daily_ks = self.dm.get_k_data(code,
autype='hfq',
begin_date=begin_date,
end_date=end_date)
# 需要判断两日的K线和10日均线的相对位置，所以如果K线数不满足11个，
# 也就是无法计算两个MA10，则直接返回False
if daily_ks.index.size < 11:
return False
# 比较收盘价和MA10的关系
daily_ks['ma'] = daily_ks['close'].rolling(10).mean()
daily_ks['delta'] = daily_ks['close'] - daily_ks['ma']
return daily_ks.loc[daily_ks.index[9]]['delta'] >= 0 > daily_ks.loc[
daily_ks.index[10]]['delta']
def __init__(self):
self.dm = DataModule()
def compute_profit(self, last_adjust_date, df_dailies, top_dailies,
bottom_dailies, adjust_date):
"""
计算收益
:param last_adjust_date: 上一个调整日
:param df_dailies:
:param top_dailies:
:param bottom_dailies:
:param adjust_date: 当前调整日
:return:
"""
# 只有存在上一个调整日，才计算上期的收益
if last_adjust_date is not None:
# 计算首档收益
top_profit = self.compute_average_profit(df_dailies, top_dailies)
# 计算末档收益
bottom_profit = self.compute_average_profit(
df_dailies, bottom_dailies)
# 计算组合收益
portfolio_profit = top_profit[0] - bottom_profit[0]
# 添加结果的DataFrame中
self.profit_df.loc[last_adjust_date] = {
'top': top_profit[0],
'bottom': bottom_profit[0],
'portfolio': portfolio_profit
}
# 计算累积收益（复利方式）
# 首档
top_cumulative_profit = round(
(self.last_top_net_value *
(1 + top_profit[0] / 100) - 1) * 100, 2)
self.last_top_net_value *= (1 + top_profit[0] / 100)
# 末档
bottom_cumulative_profit = round(
(self.last_bottom_net_value *
(1 + bottom_profit[0] / 100) - 1) * 100, 2)
self.last_bottom_net_value *= (1 + bottom_profit[0] / 100)
# 组合
portfolio_cumulative_profit = round(
(self.last_portfolio_net_value *
(1 + portfolio_profit / 100) - 1) * 100, 2)
self.last_portfolio_net_value *= (1 + portfolio_profit / 100)
# 计算沪深300的累计收益
dm = DataModule()
hs300_k = dm.get_k_data('000300',
index=True,
begin_date=adjust_date,
end_date=adjust_date)
hs300_k.set_index(['date'], 1, inplace=True)
hs300_profit = (hs300_k.loc[adjust_date]['close'] -
self.hs300_first_value) / self.hs300_first_value
self.cumulative_profit.loc[last_adjust_date] = {
'top': top_cumulative_profit,
'bottom': bottom_cumulative_profit,
'portfolio': portfolio_cumulative_profit,
'hs300': hs300_profit
}
self.count_df.loc[last_adjust_date] = {
'top': top_profit[1],
'bottom': bottom_profit[1]
}
class Account:
def __init__(self):
self.holding = dict()
self.holding_codes = set()
self.dm = DataModule()
def buy_in(self, code, volume, cost):
self.holding[code] = {
'volume': volume,
'cost': cost,
'last_value': cost
}
self.holding_codes.add(code)
def sell_out(self, code):
del self.holding[code]
self.holding_codes.remove(code)
def get_holding(self, code):
"""
通过股票代码获取该股票的持仓情况
:param code: 股票代码
:return: 持仓对象，如果没有该股票的持仓，则返回None
"""
if code in self.holding_codes:
return self.holding[code]
else:
return None
def update_holding(self, code, updated_holding):
"""
更新该持仓股的字段
:param code: 股票代码
:param updated_holding: 新的持仓股信息
"""
if code in self.holding_codes:
self.holding[code] = updated_holding
def adjust_holding_volume_at_open(self, last_date=None, current_date=None):
"""
开盘时，处理持仓股的复权
:param last_date: 上一个交易日
:param current_date: 当前交易日
"""
if last_date is not None and len(self.holding_codes) > 0:
for code in self.holding_codes:
try:
dailies = self.dm.get_k_data(code=code,
begin_date=last_date,
end_date=current_date)
if dailies.index.size == 2:
dailies.set_index(['date'], 1, inplace=True)
current_au_factor = dailies.loc[current_date][
'au_factor']
before_volume = self.holding[code]['volume']
last_au_factor = dailies.loc[last_date]['au_factor']
after_volume = int(
before_volume *
(current_au_factor / last_au_factor))
self.holding[code]['volume'] = after_volume
print('持仓量调整：%s, %6d, %10.6f, %6d, %10.6f' %
(code, before_volume, last_au_factor,
after_volume, current_au_factor),
flush=True)
except:
print('持仓量调整时，发生错误：%s, %s' % (code, current_date),
flush=True)
traceback.print_exc()
def get_total_value(self, date):
"""
计算当期那持仓股在某一天的总市值，并更新持仓股的上一个市值
:param date: 日期
"""
total_value = 0
dailies = self.dm.get_stocks_one_day_k_data(list(self.holding_codes),
date=date)
if dailies.index.size > 0:
dailies.set_index(['code'], 1, inplace=True)
for code in self.holding_codes:
try:
holding_stock = self.holding[code]
value = dailies.loc[code]['close'] * holding_stock['volume']
total_value += value
# 计算总收益
profit = (value - holding_stock['cost']
) * 100 / holding_stock['cost']
# 计算单日收益
last_value = holding_stock['last_value']
one_day_profit = (value - last_value) * 100 / last_value
# 暂存当日市值
self.holding[code]['last_value'] = value
print('持仓: %s, %10.2f, %10.2f, %4.2f, %4.2f' %
(code, value, last_value, profit, one_day_profit))
except:
print('计算收益时发生错误：%s, %s' % (code, date), flush=True)
return total_value
def __init__(self, account, max_loss):
BaseStopLoss.__init__(self, account)
self.max_loss = max_loss
self.dm = DataModule()
class Backtest:
def __init__(self, begin_date, end_date):
self.begin_date = begin_date
self.end_date = end_date
self.dm = DataModule()
self.code_daily_cache = dict()
def start(self):
"""
策略回测。结束后打印出收益曲线(沪深300基准)、年化收益、最大回撤、
:param begin_date: 回测开始日期
:param end_date: 回测结束日期
"""
total_capital = 1E7
cash = 1E7
single_position = 2E5
# 初始化信号对象
daily_k_break_ma10 = DailyKBreakMA10Signal()
low_pe_stock_pool = LowPeStockPool(self.begin_date, self.end_date, 7)
# 保存持仓股的日期
code_date_volume_dict = dict()
# 时间为key的净值、收益和同期沪深基准
df_profit = DataFrame(columns=['net_value', 'profit', 'hs300'])
# 因为上证指数没有停牌不会缺数，所以用它作为交易日历，
szzz_hq_df = self.dm.get_k_data('000001',
index=True,
begin_date=self.begin_date,
end_date=self.end_date)
all_dates = list(szzz_hq_df['date'])
# 获取沪深300在统计周期内的第一天的值
hs300_k = self.dm.get_k_data('000300',
index=True,
begin_date=all_dates[0],
end_date=all_dates[0])
hs300_begin_value = hs300_k.loc[hs300_k.index[0]]['close']
# 获取股票池数据
rebalance_dates, date_codes_dict = low_pe_stock_pool.get_option_stocks(
)
# 获取回测周期内股票池内所有股票的收盘价和前收价
all_option_code_set = set()
for rebalance_date in rebalance_dates:
for code in date_codes_dict[rebalance_date]:
all_option_code_set.add(code)
# 缓存股票的日线数据
for code in all_option_code_set:
dailies_df = self.dm.get_k_data(code,
autype=None,
begin_date=self.begin_date,
end_date=self.end_date)
dailies_hfq_df = self.dm.get_k_data(code,
autype='hfq',
begin_date=self.begin_date,
end_date=self.end_date)
# 计算复权因子
dailies_df[
'au_factor'] = dailies_hfq_df['close'] / dailies_df['close']
dailies_df.set_index(['date'], inplace=True)
self.code_daily_cache[code] = dailies_df
last_phase_codes = None
this_phase_codes = None
to_be_sold_codes = set()
to_be_bought_codes = set()
holding_code_dict = dict()
last_date = None
# 按照日期一步步回测
for _date in all_dates:
print('Backtest at %s.' % _date)
# 当期持仓股票列表
before_sell_holding_codes = list(holding_code_dict.keys())
# 处理复权
if last_date is not None and len(before_sell_holding_codes) > 0:
for code in before_sell_holding_codes:
try:
dailies = self.code_daily_cache[code]
current_au_factor = dailies.loc[_date]['au_factor']
before_volume = holding_code_dict[code]['volume']
last_au_factor = dailies.loc[last_date]['au_factor']
after_volume = int(
before_volume *
(current_au_factor / last_au_factor))
holding_code_dict[code]['volume'] = after_volume
print('持仓量调整：%s, %6d, %10.6f, %6d, %10.6f' %
(code, before_volume, last_au_factor,
after_volume, current_au_factor),
flush=True)
except:
print('持仓量调整时，发生错误：%s, %s' % (code, _date), flush=True)
# 卖出
if len(to_be_sold_codes) > 0:
code_set_tmp = set(to_be_sold_codes)
for code in code_set_tmp:
try:
if code in before_sell_holding_codes:
holding_stock = holding_code_dict[code]
holding_volume = holding_stock['volume']
sell_price = self.code_daily_cache[code].loc[
_date]['open']
sell_amount = holding_volume * sell_price
cash += sell_amount
cost = holding_stock['cost']
single_profit = (sell_amount - cost) * 100 / cost
print('卖出 %s, %6d, %6.2f, %8.2f, %4.2f' %
(code, holding_volume, sell_price,
sell_amount, single_profit))
del holding_code_dict[code]
to_be_sold_codes.remove(code)
except:
print('卖出时，发生异常：%s, %s' % (code, _date), flush=True)
print('卖出后，现金: %10.2f' % cash)
# 买入
if len(to_be_bought_codes) > 0:
sorted_to_be_bought_list = list(to_be_bought_codes)
sorted_to_be_bought_list.sort()
for code in sorted_to_be_bought_list:
try:
if cash > single_position:
buy_price = self.code_daily_cache[code].loc[_date][
'open']
volume = int(
int(single_position / buy_price) / 100) * 100
buy_amount = buy_price * volume
cash -= buy_amount
holding_code_dict[code] = {
'volume': volume,
'cost': buy_amount,
'last_value': buy_amount
}
print('买入 %s, %6d, %6.2f, %8.2f' %
(code, volume, buy_price, buy_amount),
flush=True)
except:
print('买入时，发生错误：%s, %s' % (code, _date), flush=True)
print('买入后，现金: %10.2f' % cash)
# 持仓股代码列表
holding_codes = list(holding_code_dict.keys())
# 如果调整日，则获取新一期的股票列表
if _date in rebalance_dates:
# 暂存为上期的日期
if this_phase_codes is not None:
last_phase_codes = this_phase_codes
this_phase_codes = date_codes_dict[_date]
# 找到所有调出股票代码，在第二日开盘时卖出
if last_phase_codes is not None:
out_codes = self.find_out_stocks(last_phase_codes,
this_phase_codes)
for out_code in out_codes:
if out_code in holding_code_dict:
to_be_sold_codes.add(out_code)
# 获取检测信号的开始日期和结束日期
current_date_index = all_dates.index(_date)
signal_begin_date = None
if current_date_index >= 10:
signal_begin_date = all_dates[current_date_index - 10]
# 检查是否有需要第二天卖出的股票
for holding_code in holding_codes:
if daily_k_break_ma10.is_k_down_break_ma10(
holding_code, begin_date=signal_begin_date,
end_date=_date):
to_be_sold_codes.add(holding_code)
# 检查是否有需要第二天买入的股票
to_be_bought_codes.clear()
if this_phase_codes is not None:
for _code in this_phase_codes:
if _code not in holding_codes and \
daily_k_break_ma10.is_k_up_break_ma10(_code, begin_date=signal_begin_date, end_date=_date):
to_be_bought_codes.add(_code)
# 计算总资产
total_value = 0
for code in holding_codes:
try:
holding_stock = holding_code_dict[code]
value = self.code_daily_cache[code].loc[_date][
'close'] * holding_stock['volume']
total_value += value
# 计算总收益
profit = (value - holding_stock['cost']
) * 100 / holding_stock['cost']
# 计算单日收益
one_day_profit = (value - holding_stock['last_value']
) * 100 / holding_stock['last_value']
# 暂存当日市值
holding_stock['last_value'] = value
print('持仓: %s, %10.2f, %4.2f, %4.2f' %
(code, value, profit, one_day_profit))
# 保存每一日股票的持仓数
code_date_volume_dict[code + '_' +
_date] = holding_stock['volume']
except:
print('计算收益时发生错误：%s, %s' % (code, _date), flush=True)
total_capital = total_value + cash
hs300_k_current = self.dm.get_k_data('000300',
index=True,
begin_date=_date,
end_date=_date)
hs300_current_value = hs300_k_current.loc[
hs300_k_current.index[0]]['close']
print('收盘后，现金: %10.2f, 总资产: %10.2f' % (cash, total_capital))
last_date = _date
df_profit.loc[_date] = {
'net_value':
round(total_capital / 1e7, 2),
'profit':
round(100 * (total_capital - 1e7) / 1e7, 2),
'hs300':
round(
100 * (hs300_current_value - hs300_begin_value) /
hs300_begin_value, 2)
}
# 打印回测收益曲线数值
print('Profit history start')
for index_date in df_profit.index:
print('%s, %6.2f, %6.2f' %
(index_date, df_profit.loc[index_date]['profit'],
df_profit.loc[index_date]['hs300']),
flush=True)
print('Profit history end')
drawdown = self.compute_drawdown(df_profit['net_value'])
annual_profit, sharpe_ratio = self.compute_sharpe_ratio(
df_profit['net_value'])
print('回测结果 %s - %s，年化收益： %7.3f, 最大回撤：%7.3f, 夏普比率：%4.2f' %
(self.begin_date, self.end_date, annual_profit, drawdown,
sharpe_ratio))
df_profit.plot(title='Backtest Result',
y=['profit', 'hs300'],
kind='line')
plt.show()
def compute_sharpe_ratio(self, net_values):
"""
计算夏普比率
:param net_values: 净值列表
"""
# 总交易日数
trading_days = len(net_values)
# 所有收益的DataFrame
profit_df = DataFrame(columns={'profit'})
# 收益之后，初始化为第一天的收益
profit_df.loc[0] = {'profit': round((net_values[0] - 1) * 100, 2)}
# 计算每天的收益
for index in range(1, trading_days):
# 计算每日的收益变化
profit = (net_values[index] -
net_values[index - 1]) / net_values[index - 1]
profit = round(profit * 100, 2)
profit_df.loc[index] = {'profit': profit}
# 计算标准差
profit_std = pow(profit_df.var()['profit'], 1 / 2)
# 年化收益
annual_profit = self.compute_annual_profit(trading_days,
net_values[-1])
# 夏普比率
sharpe_ratio = (annual_profit - 4.75) / profit_std
return annual_profit, sharpe_ratio
def compute_drawdown(self, net_values):
"""
计算最大回撤
:param net_values: 净值列表
"""
# 最大回撤初始值设为0
max_drawdown = 0
index = 0
# 双层循环找出最大回撤
for net_value in net_values:
for sub_net_value in net_values[index:]:
drawdown = 1 - sub_net_value / net_value
if drawdown > max_drawdown:
max_drawdown = drawdown
index += 1
return max_drawdown
def compute_annual_profit(self, trading_days, net_value):
"""
计算年化收益
"""
annual_profit = 0
if trading_days > 0:
# 计算年数
years = trading_days / 245
# 计算年化收益
annual_profit = pow(net_value, 1 / years) - 1
annual_profit = round(annual_profit * 100, 2)
return annual_profit
def find_out_stocks(self, last_phase_codes, this_phase_codes):
"""
找到上期入选本期被调出的股票，这些股票将必须卖出
:param last_phase_codes: 上期的股票列表
:param this_phase_codes: 本期的股票列表
:return: 被调出的股票列表
"""
out_stocks = []
for code in last_phase_codes:
if code not in this_phase_codes:
out_stocks.append(code)
return out_stocks
def __init__(self, begin_date, end_date):
self.begin_date = begin_date
self.end_date = end_date
self.dm = DataModule()
self.code_daily_cache = dict()
def __init__(self, account, max_profit, profit_drawdown):
BaseStopProfit.__init__(self, account)
self.max_profit = max_profit
self.profit_drawdown = profit_drawdown
self.dm = DataModule()
def compute(self, begin_date, end_date):
"""
计算指定日期内的信号
:param begin_date: 开始日期
:param end_date: 结束日期
"""
all_codes = get_all_codes()
dm = DataModule()
N = 20
k = 2
for code in all_codes:
try:
df_daily = dm.get_k_data(code,
autype='hfq',
begin_date=begin_date,
end_date=end_date)
# 计算MB，盘后计算，这里用当日的Close
df_daily['MID'] = df_daily['close'].rolling(N).mean()
# 计算STD20
df_daily['std'] = df_daily['close'].rolling(N).std()
# 计算UP
df_daily['UP'] = df_daily['MID'] + k * df_daily['std']
# 计算down
df_daily['DOWN'] = df_daily['MID'] - k * df_daily['std']
# 将日期作为索引
df_daily.set_index(['date'], inplace=True)
# 上轨和中轨右移一位
shifted_up = df_daily['UP'].shift(1)
shifted_middle = df_daily['MID'].shift(1)
# 收盘价突破或者跌破中轨的幅度占上轨和中轨宽度的比例
ref_line = (df_daily['close'] -
shifted_middle) / (shifted_up - shifted_middle)
ref_prev = ref_line.shift(1)
# 找到时间窗口内的最小值
min_val = ref_line.rolling(10).min()
# 找到时间窗口内最低点前的最大值
max_leading_value = ref_line.rolling(10).apply(
lambda vec: vec[:np.argmin(vec) + 1].max().astype(float),
raw=True)
# 中轨支撑的作用的范围
delta = 0.15
# 判断是否存在中轨支撑反弹的信号，要求：
# 时间窗口的最低点之前的最大值大于delta，最小值的绝对值小于delta，就有一个穿越阈值分界线的动作；
# 当前日期在也在阈值之上，表示又从最低点穿越到阈值分界线之上；
# 而判断前一日在阈值分界线之下，表示穿越是在当前交易日完成
m_rebound_mask = (abs(min_val) <= delta) & (ref_line > delta) & (ref_prev <= delta) & \
(max_leading_value > delta)
# 将信号保存到数据库
update_requests = []
df_daily['m_rebound_mask'] = m_rebound_mask
df_daily = df_daily[df_daily['m_rebound_mask']]
for date in df_daily.index:
doc = {'code': code, 'date': date, 'signal': 'mid_rebound'}
update_requests.append(
UpdateOne(doc, {'$set': doc}, upsert=True))
if len(update_requests) > 0:
update_result = self.collection.bulk_write(update_requests,
ordered=False)
print('%s, upserted: %4d, modified: %4d' %
(code, update_result.upserted_count,
update_result.modified_count),
flush=True)
except:
traceback.print_exc()