1.公式

```python

TYP:=(HIGH+LOW+CLOSE)/3;

攻击:(TYP-MA(TYP,10))/(0.015*AVEDEV(TYP,10)),LINETHICK1;

力量:(TYP-MA(TYP,30))/(0.015*AVEDEV(TYP,30)),COLORYELLOW,LINETHICK2;

通达信翻译：

TYP赋值:(最高价+最低价+收盘价)/3

输出攻击:(TYP-TYP的10日简单移动平均)/(0.015*TYP的10日平均绝对偏差),线宽为1

输出力量:(TYP-TYP的30日简单移动平均)/(0.015*TYP的30日平均绝对偏差),画黄色,线宽为2

```

2.代码化处理

```python

#1.该策略数据来源于baostock

#2.相关的计算与通达信的不同，由于对于计算机等常识的不懂，所以无法调用通达信数据进行分析

#但是实际计算的结果和通达信展示的不同，故而先进行计算，看看效果。

#3.重要的一点是，该数据貌似不是用前复权数据，因为从通达信的公式导出来的数据完全没有。

#4.先用回测看一下，所以将数据导入接入进来了。

import baostock as bs

import pandas as pd

import matplotlib.pyplot as plt

pd.set_option('expand_frame_repr',False)

#py文本名字：AttackPower.py

def dfload(code="sh.600789",startday='1990-01-01',endday='2020-01-01'):

#导入数据

lg=bs.login()

rs = bs.query_history_k_data_plus(code,

"date,code,open,high,low,close,preclose,volume,amount,adjustflag,turn,tradestatus,pctChg,isST",

start_date=startday, end_date=endday,

frequency="d", adjustflag="2")

data_list = []

while (rs.error_code == '0') & rs.next():

# 获取一条记录，将记录合并在一起

data_list.append(rs.get_row_data())

result = pd.DataFrame(data_list, columns=rs.fields)

return result

#5.由于baostock的数据导出均为object，所以需要转化

def astypedata(df):

df['high'] = df['high'].astype(float)

df['low'] = df['low'].astype(float)

df['open'] = df['open'].astype(float)

df['close'] = df['close'].astype(float)

df['date'] = pd.to_datetime(df['date'])

return df

#6.计算绝对误差值

def avedev(df):

'''

TYP:=(HIGH+LOW+CLOSE)/3;

攻击:(TYP-MA(TYP,10))/(0.015*AVEDEV(TYP,10)),LINETHICK1;

力量:(TYP-MA(TYP,30))/(0.015*AVEDEV(TYP,30)),COLORYELLOW,LINETHICK2;

金钻突破:力量-REF(力量,1)>90 OR 攻击-REF(攻击,1)>90;

由于金钻突破只有买点，没有买点，所以选择忽视，只以攻击和力量的交叉点进行判断买点卖点。

:param df:

:return:

'''

df['typ'] = (df['high'] + df['low'] + df['close']) / 3

df['avedev'] = abs(df['typ'] - sum(df['typ']) / len(df))

df['avedev10'] = df['avedev'].rolling(10).mean()

df['avedev30'] = df['avedev'].rolling(30).mean()

df['attack'] = (df['typ'] - df['typ'].rolling(10).mean()) / (0.015 * df['avedev10'])

df['power'] = (df['typ'] - df['typ'].rolling(30).mean()) / (0.015 * df['avedev30'])

df=df.fillna(0)

df=round(df,2)

return df

def sign(df):

for i in range(df.shape[0]):

if df['attack'][i]>df['power'][i] and df['attack'][i-1]<=df['power'][i-1]:

df.ix[i,'signal']=1

if df['attack'][i]<df['power'][i] and df['attack'][i-1]>df['power'][i-1]:

df.ix[i,'signal']=-1

return df

def position(df):

df['position'] = df['signal'].shift()

df['position'].fillna(method='ffill', inplace=True)

# 不能买的信号

cond_cannot_buy = df['open'] > df['close'].shift(1) * 1.097

# 将开盘涨停日、并且当天position为1时的'pos'设置为空值

df.loc[cond_cannot_buy & (df['position'] == 1), 'position'] = None

# 不能卖的信号

# 找出开盘跌停的日期

cond_cannot_sell = df['open'] < df['close'].shift(1) * 0.903 # 今天的开盘价相对于昨天的收盘价下得了9.7%

# 将开盘跌停日、并且当天position为0时的'pos'设置为空值

df.loc[cond_cannot_sell & (df['position'] == -1), 'position'] = None

# position为空的日期，不能买卖。position只能和前一个交易日保持一致。

df['position'].fillna(method='ffill', inplace=True)

# 在position为空值的日期，将position补全为0

df['position'].fillna(value=-1, inplace=True)

return df

#后面就是回测阶段，写到这里就可以了。

```