期货：高频日内交易

高频交易基于低手续费，且交易判断成功的概率远大于失败的基础上的。

朴素的思路是判断拐点，在拐点处产生快速交易。

首先导入某一期货品种(分钟K线). 

df = pd.read_csv("JqData/RB2205.csv", index_col='date',parse_dates=['date'])[['open','close','low','high']]

分K线走势是这样子的：

df = df[:120]
df[['close']].plot()
plt.show()

找到拐点，并标记出来。（注意：拐点判断有延时性，交易具有延时性）

算法;

求导(dx=1),根据本点的前两点,和上一拐点的性质和距离判断前一点是否是拐点（knee point),本点是否买入或卖出

# Create new columns to store the knee point flag, type, and open direction.
df['kpFlag'] = np.NaN
df['kpType'] = np.NaN
df['oDirection'] = np.NaN

prevKpClose = 0
minDistance = 3
prevKpType = KneePointType.Unknown
for i in range(2, len(df)):
    if (df['dy1'][i-1] >=0) and (df['dy1'][i] < 0):   # flat/up -> down
        df.loc[df.index[i-1], 'kpType'] = KneePointType.Down.name
        if (prevKpType != KneePointType.Down):
            df.loc[df.index[i-1], 'kpFlag'] = True
            prevKpType = KneePointType.Down       
            if (df['close'][i-1] - prevKpClose >= minDistance * 1):
                df.loc[df.index[i], 'oDirection'] = OpenDirection.Sell.name
                prevKpClose = df.loc[df.index[i-1], 'close'] 
    elif (df['dy1'][i-1] <=0) and (df['dy1'][i] > 0):   # flat/down -> up
        df.loc[df.index[i-1], 'kpType'] = KneePointType.Up.name
        if (prevKpType != KneePointType.Up):
            df.loc[df.index[i-1], 'kpFlag'] = True
            prevKpType = KneePointType.Up       
            if (df['close'][i-1] - prevKpClose <= minDistance * -1):
                df.loc[df.index[i], 'oDirection'] = OpenDirection.Buy.name
                prevKpClose = df.loc[df.index[i-1], 'close'] 

把拐点（K），买点（B），和卖点（S）图形化显示一下：

from matplotlib import pylab
z = df[['close','kpFlag']]
z.plot(marker='o') # Plot the data, with a marker set.
#pylab.xlim(0,3) # Change the axes limits so that we can see the annotations.
#pylab.ylim(0,4)
plt.rcParams["figure.figsize"] = (36,20)
ax = pylab.gca()

for i in z.index: # iterate through each index in the dataframe
    v = df.loc[i, 'close'] 
    f = df.loc[i, 'kpFlag'] 
    d = df.loc[i, 'oDirection']
    if f == True:
        ax.annotate('K',xy=(i,v),  bbox=dict(boxstyle='round,pad=0.2', fc='pink', alpha=0.5),)
    if d == OpenDirection.Buy.name:
        ax.annotate('B',xy=(i,v),bbox=dict(boxstyle='round,pad=0.2', fc='red', alpha=0.5),fontsize=20)
    if d == OpenDirection.Sell.name:
        ax.annotate('S',xy=(i,v),  bbox=dict(boxstyle='round,pad=0.2', fc='green', alpha=0.5), fontsize=20)

 看起来这个震荡行情下的表现还不错，单边行情中还需要做一些微调