使用python和tushare，1.选取前三年内的任意一天，如果当日上涨家数小于1700家，则这一天作为trade_date,2.从trade_date往前推90天，选出这90天内最高价是最低价的2倍以上的股票作为股票池，3.在trade_date当日如果，股票池股票收十字线，则买入，持有3天则卖出，4.起始资金1000万，每次最多买5只，打印出结果和收益曲线

好的，我会根据您的要求进行回答。首先，需要安装tushare模块，可以使用以下命令：

!pip install tushare

然后，需要获取股票数据。以下是获取tushare数据的代码：

import tushare as ts 

# 设置tushare token
ts.set_token('your_token_here')

# 初始化pro接口
pro = ts.pro_api()

# 获取交易日历
trade_cal = pro.trade_cal(exchange='', start_date='20180101', end_date='20201231', fields='cal_date, is_open')

# 获取前三年内的任意一天作为trade_date
import random

trade_date = random.choice(trade_cal[trade_cal['is_open'] == 1].cal_date.tolist())
print('选定的交易日为：', trade_date)

# 获取trade_date前90天的数据
df = pro.daily(trade_date=trade_date, fields='ts_code, trade_date, open, high, low, close, pre_close, pct_chg')

# 筛选出trade_date前90天最高价是最低价的2倍以上的股票作为股票池
df['spread'] = df['high'] / df['low']
stock_pool = df[df['trade_date'] < trade_date].groupby('ts_code').filter(lambda x: x['spread'].max() >= 2)

print('股票池选取的股票数为：', len(stock_pool))

接下来，需要编写判断十字线的函数：

# 判断是否是十字星线
def is_doji(df):
    # 获取开盘价和收盘价
    open_price = df.iloc[-2]['open']
    close_price = df.iloc[-2]['close']
    
    # 判断是否满足十字星线的条件
    if abs(open_price - close_price) / (df.iloc[-3]['high'] - df.iloc[-3]['low']) <= 0.1 and \
    min(open_price, close_price) > df.iloc[-3]['close']:
        return True
    else:
        return False

然后，需要编写买卖股票的函数：

# 买卖股票的函数
def trade(stock_pool, money):
    # 获取当前时间的股票数据
    df = pro.daily(trade_date=trade_date, ts_code=','.join(stock_pool['ts_code'].tolist()), fields='ts_code, trade_date, open, high, low, close, pre_close, pct_chg')
    
    # 筛选出当天收十字线的股票
    doji_stocks = []
    for ts_code in stock_pool['ts_code'].tolist():
        sub_df = df[df['ts_code'] == ts_code]
        if is_doji(sub_df):
            doji_stocks.append(ts_code)
    
    # 如果没有股票收十字线，则不进行交易
    if len(doji_stocks) == 0:
        return money
    
    # 计算每只股票的购买数量
    num_per_stock = min(5, len(doji_stocks))
    per_stock_money = money / num_per_stock
    num_per_stock = [int(per_stock_money / sub_df.iloc[-1]['close']) for ts_code, sub_df in df[df['ts_code'].isin(doji_stocks)].groupby('ts_code')]
    
    # 购买股票
    for i, ts_code in enumerate(doji_stocks[:num_per_stock]):
        sub_df = df[df['ts_code'] == ts_code]
        buy_price = sub_df.iloc[-1]['close']
        money -= num_per_stock[i] * buy_price
    
    # 持有3天后卖出
    sell_date = pro.trade_cal(exchange='', start_date=trade_date, end_date='20201231', fields='cal_date, is_open')
    sell_date = sell_date[sell_date['is_open'] == 1].iloc[3]['cal_date']
    df = pro.daily(trade_date=sell_date, ts_code=','.join(stock_pool['ts_code'].tolist()), fields='ts_code, trade_date, open, high, low, close, pre_close, pct_chg')
    sell_money = 0
    for ts_code, num in zip(doji_stocks[:num_per_stock], num_per_stock):
        sub_df = df[df['ts_code'] == ts_code]
        sell_price = sub_df.iloc[-1]['close']
        sell_money += num * sell_price
    
    return money + sell_money

最后，可以运行主程序，计算收益曲线：

# 主程序
money = 100000000
profits = []
for i, row in trade_cal[trade_cal['cal_date'] >= trade_date].iterrows():
    if row['is_open'] == 1:
        # 计算收益
        money = trade(stock_pool, money)
        profit = (money - 100000000) / 100000000
        profits.append(profit)
        print('交易日：{}，当前资产：{}，收益率：{}%'.format(row['cal_date'], money, round(profit * 100, 2)))

# 绘制收益曲线
import matplotlib.pyplot as plt

plt.plot(profits)
plt.xlabel('交易日')
plt.ylabel('收益率')
plt.title('收益曲线')
plt.show()

希望对您有帮助！