# 根据情况指定xtquant的路径
import sys
import numpy as np
import pandas as pd
from xtquant import xtdata
import xtquant
print(xtquant)
print(xtdata.data_dir)
# 指定获取投研端数据(可不指定,默认优先连接投研)
# xtdata.reconnect(port=58613)
# xtdata.download_sector_data()
class G():
pass
g = G()
def init(C):
# ------------------------参数设定-----------------------------
g.his_st = {}
# g.s = C.get_stock_list_in_sector("沪深A股") # 获取沪深A股股票列表
g.s = C.get_stock_list_in_sector("沪深300") # 获取沪深300股票列表
# g.s = ['000001.SZ']
g.day = 0
g.holdings = {i: 0 for i in g.s}
g.weight = [0.1] * 10
g.buypoint = {}
g.money = 1000000 # C.capital
g.accid = 'test'
g.profit = 0
# 因子权重
g.buy_num = 10 # 买排名前5的股票,在过滤中会用到
g.per_money = g.money / g.buy_num * 0.95
def after_init(C):
# ------------------------量价数据获取-----------------------------
data = xtdata.get_market_data_ex([], g.s, period='1d', dividend_type='front_ratio',
fill_data=True)
close_df = get_df_ex(data,"close")
open_df = get_df_ex(data,"open")
low_df = get_df_ex(data,"low")
high_df = get_df_ex(data,"high")
volume_df = get_df_ex(data,"volume")
amount_df = get_df_ex(data,"amount")
preclose_df = get_df_ex(data,"preClose")
# ------------------------基础数据获取-----------------------------
# 将 g.s 中的全部股票的 TotalVolume 都获取出来,组合成一个 DataFrame
# 例如 g.s 中有 10 个股票,那么下面的代码就会返回一个 1 行 10 列的 DataFrame
# 该 DataFrame 的 index 是股票代码,columns 是 TotalVolume
# 该 DataFrame 的数据是每个股票的 TotalVolume,请给出代码:
# C.get_instrumentdetail('600000.SH')['TotalVolume']
# 使用字典推导来获取每个股票的TotalVolume
total_volumes = {stock: C.get_instrumentdetail(stock)['TotalVolume'] for stock in g.s}
# 将字典转换为DataFrame,但先转化为一个嵌套字典
df_total_volume = pd.DataFrame({k: v for k, v in total_volumes.items()}, index=['TotalVolume'])
# ------------------------财务数据获取-----------------------------
# ------------------------因子1计算及处理--------------------------------
# 1. 市值因子: 用市值因子 = 股票收盘价 * 股票总股本,要利用好,要求对应列名相乘
factor = close_df * df_total_volume.loc['TotalVolume']
# ------------------------因子2计算及处理--------------------------------
# 判断 close_df 中的 code 上市时间大于120天
stock_opendate_filter = filter_opendate_qmt(C, close_df, 120)
# ------------------------上市日期过滤处理--------------------------------
# 两个布尔值的 DataFrame 对应相乘过滤掉每个交易日上市不足120天的
factor *= stock_opendate_filter.astype(int).replace(0, np.nan)
# ------------------------排序处理-----------------------------------
# 对 factor 每行在一行内进行排序
factor_sorted = rank_filter(factor, 10, ascending=True, method='min', na_option='keep')
# ------------------------因子组合得到布尔值信号--------------------------------
# 确保没有未来数据的影响,将因子数据向后移动一天
g.factor_df = factor_sorted.shift(1) #
g.close_df = close_df.shift(1) # 为了计算收益率,将收盘价向后移动一天
g.open_df = open_df
g.stock_opendate_filter = stock_opendate_filter
def handlebar(C):
# 获取当前 K 线位置
d = C.barpos
# 获取当前 K 线时间
backtest_time = timetag_to_datetime(C.get_bar_timetag(C.barpos), "%Y%m%d")
factor_series = g.factor_df.loc[backtest_time]
buy_list = daily_filter(factor_series, backtest_time)
print(backtest_time, buy_list)
# 获取持仓
hold = get_holdings(g.accid, 'stock')
need_sell = [s for s in hold if s not in buy_list]
print('\t\t\t\t\t\t\t', backtest_time, 'sell list', need_sell)
# 卖出
for s in need_sell:
price = g.open_df.loc[backtest_time, s]
vol = hold[s]['持仓数量']
passorder(24, 1101, g.accid, s, 11, price, vol, 1,"backtest","小市值",C)
# 获取持仓
hold = get_holdings(g.accid, 'stock')
asset = get_trade_detail_data(g.accid, 'stock', 'account')
cash = asset[0].m_dAvailable
buy_num = g.buy_num - len(hold)
buy_list = [s for s in buy_list if s not in hold]
# 买入
if buy_num > 0 and buy_list:
buy_list = buy_list[:buy_num]
# money = cash/buy_num
print(backtest_time, 'buy list', buy_list)
for s in buy_list:
price = g.open_df.loc[backtest_time, s]
if price > 0:
passorder(23, 1102, g.accid, s, 11, float(price), g.per_money,1,"backtest","小市值",C)
def daily_filter(factor_series, backtest_time):
# 将 factor_series 中值 True 的index,转化成列表
print(len(factor_series))
sl = factor_series[factor_series].index.tolist()
print(len(sl))
# exit()
# st过滤
sl = [s for s in sl if not is_st(s, backtest_time)]
sl = sorted(sl, key=lambda k: factor_series.loc[k])
return sl[:g.buy_num]
def is_st(s, date):
# 判断某日在历史上是不是st *st
st_dict = g.his_st.get(s, {})
if not st_dict:
return False
else:
st = st_dict.get('ST', []) + st_dict.get('*ST', [])
for start, end in st:
if start <= date <= end:
return True
def get_df(dt: dict, df: pd.DataFrame, values_name: str) -> pd.DataFrame:
'''
循环从字典里赋值矩阵
values_name可选字段: ['time', 'stime', 'open', 'high', 'low', 'close', 'volume','amount', 'settelementPrice', 'openInterest', 'preClose', 'suspendFlag']
'''
df1 = df.copy()
df1 = df1.apply(lambda x: dt[x.name][values_name])
return df1
def get_df_ex(data:dict,field:str) -> pd.DataFrame:
'''
ToDo:用于在使用get_market_data_ex的情况下,取到标准df
Args:
data: get_market_data_ex返回的dict
field: ['time', 'open', 'high', 'low', 'close', 'volume','amount', 'settelementPrice', 'openInterest', 'preClose', 'suspendFlag']
Return:
一个以时间为index,标的为columns的df
'''
_index = data[list(data.keys())[0]].index.tolist()
_columns = list(data.keys())
df = pd.DataFrame(index=_index,columns=_columns)
for i in _columns:
df[i] = data[i][field]
return df
def rank_filter(df: pd.DataFrame, N: int, axis=1, ascending=False, method="max", na_option="keep") -> pd.DataFrame:
"""
Args:
df: 标准数据的df
N: 判断是否是前N名
axis: 默认是横向排序
ascending : 默认是降序排序
na_option : 默认保留nan值,但不参与排名
Return:
pd.DataFrame:一个全是bool值的df
"""
_df = df.copy()
_df = _df.rank(axis=axis, ascending=ascending, method=method, na_option=na_option)
return _df <= N
def filter_opendate_qmt(C, df: pd.DataFrame, n: int) -> pd.DataFrame:
'''
ToDo: 判断传入的df.columns中,上市天数是否大于N日,返回的值是一个全是bool值的df
Args:
C:contextinfo类
df:index为时间,columns为stock_code的df,目的是为了和策略中的其他df对齐
n:用于判断上市天数的参数,如要判断是否上市120天,则填写
Return:pd.DataFrame
'''
local_df = pd.DataFrame(index=df.index, columns=df.columns)
stock_list = df.columns
stock_opendate = {i: C.get_instrument_detail(i)["OpenDate"] for i in stock_list}
# print(type(stock_opendate["000001.SZ"]), stock_opendate["000001.SZ"])
for stock, date in stock_opendate.items():
local_df.at[date, stock] = 1
df_fill = local_df.fillna(method="ffill")
result = df_fill.expanding().sum() >= n
return result
def filter_opendate_xt(df: pd.DataFrame, n: int) -> pd.DataFrame:
'''
ToDo: 判断传入的df.columns中,上市天数是否大于N日,返回的值是一个全是bool值的df
Args:
C:contextinfo类
df:index为时间,columns为stock_code的df,目的是为了和策略中的其他df对齐
n:用于判断上市天数的参数,如要判断是否上市120天,则填写
Return:pd.DataFrame
'''
local_df = pd.DataFrame(index=df.index, columns=df.columns)
stock_list = df.columns
stock_opendate = {i: xtdata.get_instrument_detail(i)["OpenDate"] for i in stock_list}
for stock, date in stock_opendate.items():
local_df.at[date, stock] = 1
df_fill = local_df.fillna(method="ffill")
result = df_fill.expanding().sum() >= n
return result
def get_holdings(accid, datatype):
'''
Arg:
accondid:账户id
datatype:
'FUTURE':期货
'STOCK':股票
......
return:
{股票名:{'手数':int,"持仓成本":float,'浮动盈亏':float,"可用余额":int}}
'''
PositionInfo_dict = {}
resultlist = get_trade_detail_data(accid, datatype, 'POSITION')
for obj in resultlist:
PositionInfo_dict[obj.m_strInstrumentID + "." + obj.m_strExchangeID] = {
"持仓数量": obj.m_nVolume,
"持仓成本": obj.m_dOpenPrice,
"浮动盈亏": obj.m_dFloatProfit,
"可用余额": obj.m_nCanUseVolume
}
return PositionInfo_dict
if __name__ == '__main__':
import sys
from xtquant.qmttools import run_strategy_file
# 参数定义方法一,如果使用方法二定义参数,run_strategy_file的param参数可不传
param = {
'stock_code': '000300.SH', # 驱动handlebar的代码,
'period': '1d', # 策略执行周期 即主图周期
'start_time': '2022-01-01 00:00:00', # 注意格式,不要写错
'end_time': '2024-03-01 00:00:00', # 注意格式,不要写错
'trade_mode': 'backtest', # 'backtest':回测
'quote_mode': 'history',
# handlebar模式,'realtime':仅实时行情(不调用历史行情的handlebar),'history':仅历史行情, 'all':所有,即history+realtime
}
# user_script = os.path.basename(__file__) # 当前脚本路径,相对路径,绝对路径均可,此处为相对路径的方法
user_script = sys.argv[0] # 当前脚本路径,相对路径,绝对路径均可,此处为绝对路径的方法
print(user_script)
result = run_strategy_file(user_script, param=param)
if result:
print(result.get_backtest_index())
print(result.get_group_result())
xtdata.run()