ROE策略

# 导入函数库
import jqdata
import pandas as pd
from datetime import datetime

# 定义ROE计算函数
def roe_ttm(stocklist,scandate,num = 10,period = 'Q',cut = True,tb = True):
    if isinstance(scandate,datetime):
        scandate = datetime.strftime(scandate,'%Y-%m-%d')
    # 定义相关参数
    q = query(
        income.statDate,#财报统计的季度的最后一天
        income.pubDate,#公司发布财报日期
        income.code,  #股票代码
        income.np_parent_company_owners,#归属于母公司股东的净利润(元)
        indicator.adjusted_profit,#扣除非经常损益后的净利润(元)
        balance.equities_parent_company_owners,#归属于母公司股东权益合计(元)
    ).filter(
        income.code.in_(stocklist)#query是pandas里面的函数
    )
    # 季末时间表
    if period.upper() == 'Q':
        datelist = [str(dt) for dt in pd.date_range(start = '2000-01-01', end = scandate,freq = 'Q').to_period('Q')]
    # 年末时间表
    elif period.upper() == 'A':
        datelist = [str(dt) for dt in pd.date_range(start = '2000-01-01', end = scandate,freq ='A').to_period('A')]
    # 提取近期数据    
    if num + 5 < len(datelist) and period.upper() == 'Q':
        datelist = datelist[-num-5:]
    elif num + 1 < len(datelist) and period.upper == 'A':
        datelist = datelist[-num-1:]
    # 构建数据字典
    df = pd.DataFrame()
    for dt in datelist:
        df = pd.concat([df,get_fundamentals(q,statDate = dt)],axis = 0)
    df.columns = ['statDate','pubDate','code','net_profit','adjust_profit','equities']
    df.set_index(['statDate','code'],inplace = True)
    df = df.unstack('code')
    # 季末ROE滚动计算
    if period.upper() == 'Q':
        net_profit_ttm_df = pd.rolling_sum(df['net_profit'],4)
        adjusted_profit_ttm_df = pd.rolling_sum(df['adjust_profit'],4)
        equities_avg_df = pd.rolling_apply(df['equities'],4,lambda x: x[0]/2 + x[-1]/2)
    # 四种情况，摊薄或扣非
        if cut and tb:
            roe_df = adjusted_profit_ttm_df/df['equities']
        elif cut and (not tb):
            roe_df = adjusted_profit_ttm_df/equities_avg_df
        elif (not cut) and tb:
            roe_df = net_profit_ttm_df/df['equities']
        else:
            roe_df = net_profit_ttm_df/equities_avg_df      
    # 年末ROE滚动计算
    elif period.upper() == 'A':
        equities_avg_df = pd.rolling_apply(df['equities'],2,lambda x:x[0]/2+x[-1]/2)
        if cut and tb:
            roe_df = df['adjust_profit']/df['equities']
        elif cut and (not tb):
            roe_df = df['adjust_profit']/equities_avg_df
        elif (not cut) and tb:
            roe_df = df['net_profit']/df['equities']
        else:
            roe_df = df['net_profit']/equities_avg_df
    
    roe_df = roe_df.stack()
    df = df.stack()['pubDate']
    df = pd.concat([df,roe_df],axis = 1)
    
    if cut and tb:
        field = 'roe_cut_tb'
    elif cut and (not tb):
        field = 'roe_cut'
    elif (not cut) and tb:
        field = 'roe_tb'
    else:
        field = 'roe'
    df.columns = ['pubDate',field]
    df.reset_index(inplace = True)
    # 规范日期格式
    df['statDate'] = df['statDate'].map(lambda x:datetime.strptime(x,'%Y-%m-%d'))
    df['pubDate']  = df['pubDate'].map(lambda x:datetime.strptime(x,'%Y-%m-%d'))
    
    return df.dropna()

def roe_filter(roe_df):
    df = pd.DataFrame()
    field = roe_df.columns[-1]
    df['new']  = roe_df.groupby('code')[field].apply(lambda x: x.iloc[-1]) # 按股票代码进行分组
    df['min']  = roe_df.groupby('code')[field].apply(lambda x: min(x))
    df['max']  = roe_df.groupby('code')[field].apply(lambda x: max(x))
    df['mean'] = roe_df.groupby('code')[field].apply(lambda x: mean(x))
    df['qdesc_1'] = roe_df.groupby('code')[field].apply(lambda x: x.iloc[-1] < x.iloc[-2])
    df['qdesc_2'] = roe_df.groupby('code')[field].apply(lambda x: x.iloc[-1] < x.iloc[-2] and x.iloc[-2] < x.iloc[-3])
    df = df[ (df['min'] > 0.15) & (df['new'] > df['mean'] ) & (df['qdesc_2'] == False)]
    # 最近一年的滚动 roe 大于 15,最新 roe 大于均值，roe未连续两季度下降
    return df.index.tolist()
    
def pe_ratio_median(context): # #pe中位数控制仓位
    scandate = context.previous_date
    q = query(valuation.code,valuation.pe_ratio)
    df = get_fundamentals(q, scandate).dropna()
    df['ep_ratio'] = 1/df['pe_ratio']
    context.pe_ratio_median = 1/df['ep_ratio'].quantile(0.5)
    
def momentum_filter(stock):  #过滤月涨幅超过20%的股票,避免追高
    price = attribute_history(stock, 21, '1d', ['close'], skip_paused=True, df=True, fq='pre')
    condition = price['close'][-1]/price['close'][0] < 1.2
    return condition

def capital_allocation(context,buylist_tradable,method = 2): # PE加权或波动率加权
    df = pd.DataFrame()
    scandate = context.previous_date
    if method == 1:
        q = query(valuation.code,valuation.pe_ratio).filter(
            valuation.code.in_(buylist_tradable))
        df =get_fundamentals(q, scanDate).dropna()
        df['ep_ratio'] = 1/df['pe_ratio']
        df['weight'] = df['ep_ratio']/df['ep_ratio'].sum()
        return df
    elif method == 2:
        prices = history(250, '1d', 'close', security_list = buylist_tradable)
        df['std'] = (prices.pct_change().std(axis = 0)*sqrt(250))
        df['std2'] = 1/df['std']**3
        df['weight'] = df['std2']/df['std2'].sum()
        df.reset_index(inplace = True) 
        df.columns = ['code','std','std2','weight']
        return df

def selection(context):
    scandate = context.current_dt
    stocklist = get_index_stocks('000906.XSHG')
    roe_df = roe_ttm(stocklist,scandate,num = 12,period ='Q',tb = False)
    roe_df = roe_df[roe_df['pubDate'] < scandate]
    stocklist_1 = roe_filter(roe_df)
    context.buylist = [stock for stock in stocklist_1 if momentum_filter(stock)]
    context.weights0 = capital_allocation(context,context.buylist) #
#    print '每月调仓信息'
#    print  context.weights0
        
def rebalance(context):
    current_data = get_current_data()
    print context.buylist
    buylist_tradable = [stock for stock in context.buylist if not (current_data[stock].paused or current_data[stock].day_open 
                        > current_data[stock].high_limit*0.985)] # 筛选出没有停牌和开盘价没有大于涨停价的 0.985 的股票（防止买入高开的股票）
    if len(buylist_tradable) != 0:
        weights = capital_allocation(context,buylist_tradable) # 默认执行method2，即根据强化的波动率来计算股票权重
    
    for stock in context.portfolio.positions.keys():
        if stock not in context.buylist:
            order_target(stock,0)
            
    if  len(buylist_tradable) == 0:
        pass
    elif context.pe_ratio_median < 80:
        t_value = 0.95*context.portfolio.total_value
    elif context.pe_ratio_median > 80: 
        t_value = 0.30*context.portfolio.total_value
    for stock in buylist_tradable:
        weight = weights[weights['code']== stock]['weight']
        order_target_value(stock,weight*t_value)

# 初始化函数
def initialize(context):
    set_benchmark('000906.XSHG')
    set_option('use_real_price', True)
    set_order_cost(OrderCost(close_tax=0.001, open_commission=0.0003, 
    close_commission=0.0003, min_commission=5), type='stock')
    
    run_monthly(pe_ratio_median, 1,time='before_open')
    run_monthly(selection, 1,time='open-3h')
    run_monthly(rebalance, 1,time='open')    
    
def after_market_close(context):
    Pass