# 开始时间

start_date ='2013-01-01'
# 结束时间
end_date = '2017-06-01'

# 15，14，13三年的年终财报时间  
quarters = ['20161231', '20151231', '20141231'];

instruments = D.instruments(start_date, end_date)
# 代码，日期，净资产收益率, 归属母公司股东的净利润同比增长率，财报季度时间，财报对应的季度取值 1/2/3/4
# 参见 https://bigquant.com/docs/data_financial_statements.html
fields = ['instrument', 'date', 'fs_roe_ttm', 'fs_net_profit_yoy', 'fs_quarter' ,'fs_quarter_index']

# 获取以上字段的财报数据
financial_data = D.financial_statements(instruments, start_date, end_date, fields)

# 筛选出第四季度财报数据
financial_data = financial_data[financial_data['fs_quarter_index'] == 4]

# 获取市盈率历史数据
pe_data = D.history_data(instruments, start_date, end_date, ['name', 'pe_ttm'])

financial_data = financial_data.drop_duplicates(['instrument', 'fs_quarter']);
financial_data.sort_values(by='date', ascending=False, inplace=True)


stocks = pd.DataFrame()

def good_stock(df):
    global stocks
    instrument = df.iloc[0]['instrument']
    for q in quarters:
     # 找出该股票年报的数据，q in ['20151231', '20141231', '20131231']
     cur_df = df[df['fs_quarter'] == q]
     # 没有该年的财报数据
     if cur_df.empty:
        return
     # 连续三年净资产收益率 > 15
     if np.isnan(cur_df.iloc[0]['fs_roe_ttm']) or cur_df.iloc[0]['fs_roe_ttm'] <= 15:
        return
    # 最近一年财报净利润同比增长率 > 15
    latest_financial = df[df['fs_quarter'] == quarters[0]]
    if np.isnan(latest_financial.iloc[0]['fs_net_profit_yoy']) or latest_financial.iloc[0]['fs_net_profit_yoy'] <= 15:
        return
    
    # 找到年度财报日或者之前最近一天pe数据
    pe_history = pe_data[(pe_data['instrument'] == instrument) & (pe_data['date'] <= latest_financial.iloc[0]['date']) & (~np.isnan(pe_data['pe_ttm']))];
    pe_history.sort_values(by='date', ascending=False, inplace=True)
    
    if pe_history.shape[0] == 0:
        print ('no pe data of ' + instrument)
    # 市盈率 < 35
    if np.isnan(pe_history.iloc[0]['pe_ttm']) or pe_history.iloc[0]['pe_ttm'] >= 35:
        return
    
    # 找到一只股票，添加到result中
    stocks = stocks.append(pe_history.iloc[0][['instrument', 'name']], ignore_index=True)

financial_data.groupby('instrument').apply(good_stock)

print("共找到" + str(stocks.shape[0]) + "支好股票：")
print(stocks)

# 初始化虚拟账户状态，只在第一个交易日运行
def initialize(context):
    # 设置手续费，买入时万3，卖出是千分之1.3,不足5元以5元计
    context.set_commission(PerOrder(buy_cost=0.0003, sell_cost=0.0013, min_cost=5))

# 策略交易逻辑，每个交易日运行一次
def handle_data(context, data):
    for index, row in stocks.iterrows():  
        # 字符型股票代码转化成BigQuant回测引擎所需的股票代码
        instrument = context.symbol(row['instrument'])
        
        # 目前持仓数量
        curr_position =  context.portfolio.positions[instrument].amount
        # 第一个能买入的时间买入股票
        if curr_position == 0 and data.can_trade(instrument):
            # 买入一万元股票
            order_value(instrument, 10000)

# 策略比较参考标准，以沪深300为例
benchmark = '000300.INDX'
# 策略回测接口: https://bigquant.com/docs/strategy_backtest.html
# 回测最近一年数据，2016-06-01起的第一个工作日买入这些股票
m = M.backtest.v5(
    instruments=instruments,
    start_date="2017-05-01",
    end_date='2017-06-15',
    initialize=initialize,
    handle_data=handle_data,
    # 买入订单以开盘价成交
    order_price_field_buy='open',
    # 卖出订单以开盘价成交
    order_price_field_sell='open',
    # 初始资金
    capital_base=stocks.shape[0] * 10000,
    benchmark=benchmark,
)