Ta-Lib源码解析（三）：蜡烛图指标 (Candlestick Indicator) #（附Python重构代码和示意图）（补充中）

3.（蜡烛图指标）Candlestick Indicator

蜡烛图指标为识别K线形态的指标，笔者认为重点在于定义方式所提供的思路，背后的投资逻辑可以借鉴但不建议直接参考

文档的翻译内容和源码逻辑存在不一致的现象，笔者分别列出中文百科中的定义和源码实际逻辑，供大家参考

一. 前言

技术分析理论中的一些模糊词，ta-lib中给出了精确的定义

源码
见 ta_global.c

TA_RetCode TA_RestoreCandleDefaultSettings( TA_CandleSettingType settingType )
{
   
    const TA_CandleSetting TA_CandleDefaultSettings[] = {
   
        /* real body is long when it's longer than the average of the 10 previous candles' real body */
        {
    TA_BodyLong, TA_RangeType_RealBody, 10, 1.0 },
        /* real body is very long when it's longer than 3 times the average of the 10 previous candles' real body */
        {
    TA_BodyVeryLong, TA_RangeType_RealBody, 10, 3.0 },
        /* real body is short when it's shorter than the average of the 10 previous candles' real bodies */
        {
    TA_BodyShort, TA_RangeType_RealBody, 10, 1.0 },
        /* real body is like doji's body when it's shorter than 10% the average of the 10 previous candles' high-low range */
        {
    TA_BodyDoji, TA_RangeType_HighLow, 10, 0.1 },
        /* shadow is long when it's longer than the real body */
        {
    TA_ShadowLong, TA_RangeType_RealBody, 0, 1.0 },
        /* shadow is very long when it's longer than 2 times the real body */
        {
    TA_ShadowVeryLong, TA_RangeType_RealBody, 0, 2.0 },
        /* shadow is short when it's shorter than half the average of the 10 previous candles' sum of shadows */
        {
    TA_ShadowShort, TA_RangeType_Shadows, 10, 1.0 },
        /* shadow is very short when it's shorter than 10% the average of the 10 previous candles' high-low range */
        {
    TA_ShadowVeryShort, TA_RangeType_HighLow, 10, 0.1 },
        /* when measuring distance between parts of candles or width of gaps */
        /* "near" means "<= 20% of the average of the 5 previous candles' high-low range" */
        {
    TA_Near, TA_RangeType_HighLow, 5, 0.2 },
        /* when measuring distance between parts of candles or width of gaps */
        /* "far" means ">= 60% of the average of the 5 previous candles' high-low range" */
        {
    TA_Far, TA_RangeType_HighLow, 5, 0.6 },
        /* when measuring distance between parts of candles or width of gaps */
        /* "equal" means "<= 5% of the average of the 5 previous candles' high-low range" */
        {
    TA_Equal, TA_RangeType_HighLow, 5, 0.05 }
    };

    int i;
    if( settingType > TA_AllCandleSettings )
        return TA_BAD_PARAM;
    if( settingType == TA_AllCandleSettings )
        for( i = 0; i < TA_AllCandleSettings; ++i )
            TA_Globals->candleSettings[i] = TA_CandleDefaultSettings[i];
    else
        TA_Globals->candleSettings[settingType] = TA_CandleDefaultSettings[settingType];
    return TA_SUCCESS;
}

翻译

TA_BodyLong： 当期蜡烛体长度超过过去十日蜡烛体长度的平均值
TA_ShadowVeryLong： 当期蜡烛体长度超过过去十日蜡烛体长度的平均值的三倍
TA_BodyShort： 当期蜡烛体长度低于过去十日蜡烛体长度的平均值
TA_BodyDoji： 蜡烛体类似于十字线，短于过去十日蜡烛体长度的平均值的10%
TA_ShadowLong： 当期影线长度长于蜡烛体长度
TA_ShadowVeryLong： 当期影线长度长于蜡烛体长度两倍
TA_ShadowShort： 当期影线短于前十期影线长度平均值的一半
TA_ShadowVeryShort： 当期影线短于前十期high-low长度平均值的10%
TA_Near： 小于前五期20%的high-low长度平均值
TA_Far： 大于前五期60%的high-low长度平均值
TA_Equal： 小于前五期5%的high-low长度平均值

二. 单K线模型

CDLBELTHOLD（捉腰带线）

python函数原型：

cdlbelthold= CDL2BELTHOLD(open, high, low, close)

解释：

看涨为例，其开市价位于当日的最低点，然后市场一路上扬。

重构代码（来自ta_CDLBELTHOLD.c， 有修改， 只保留逻辑）：

def CDLBELTHOLD_(open, high, low, close):
    _len = len(open)
    open, high, low, close = pd.Series(open), pd.Series(high), pd.Series(low), pd.Series(close)
    if _len >= 11:
        high_low_range = high - low
        shadow_very_short = high_low_range.rolling(10).mean().shift(1) * 0.1
        real_body = abs(open - close)
        body_long = real_body.rolling(10).mean().shift(1)
        shadow_upper = high - np.maximum(close, open)
        shadow_lower = np.minimum(close, open) - low
        condition_1 = real_body > body_long
        condition_2_1 = (close > open) & (shadow_lower < shadow_very_short)
        condition_2_2 = (close < open) & (shadow_upper < shadow_very_short)
        cdlbelthold = np.where(condition_1 & condition_2_1, 100, 0)
        cdlbelthold = np.where(condition_1 & condition_2_2, -100, cdlbelthold)
    else:
        cdlbelthold = np.zeros(_len).astype(int)
    return cdlbelthold

CDLCLOSINGMARUBOZU （收盘缺影线）

python函数原型：

cdlclosingmarubozu= CDLCLOSINGMARUBOZU(open, high, low, close)

解释：

中文百科：一日K线模式，以阳线为例，最低价低于开盘价，收盘价等于最高价， 预示着趋势持续。
源码与抓腰带线几乎相同

重构代码（来自ta_CDLBELTHOLD.c， 与BELTHOLD极其相似应该直接复制粘贴该条件）：

def CDLCLOSINGMARUBOZU_(open, high, low, close):
    _len = len(open)
    open, high, low, close = pd.Series(open), pd.Series(high), pd.Series(low), pd.Series(close)
    if _len >= 11:
        high_low_range = high - low
        shadow_very_short = high_low_range.rolling(10).mean().shift(1) * 0.1
        real_body = abs(open - close)
        body_long = real_body.rolling(10).mean().shift(1)
        shadow_upper = high - np.maximum(close, open)
        shadow_lower = np.minimum(close, open) - low
        condition_1 = real_body > body_long
        condition_2_1 = (close > open) & (shadow_upper < shadow_very_short)
        condition_2_2 = (close < open) & (shadow_lower < shadow_very_short)
        cdlclosingmarubozu = np.where(condition_1 & condition_2_1, 100, 0)
        cdlclosingmarubozu = np.where(condition_1 & condition_2_2, -100, cdlclosingmarubozu)
    else:
        cdlclosingmarubozu = np.zeros(_len).astype(int)
    return cdlclosingmarubozu

CDLDOJI（doji十字）

python函数原型：

cdldoji = CDLDOJI(open, high, low, close)

解释：

单k线模型，十字线

重构代码（来自ta_CDLDOJI.c， 有修改， 只保留逻辑）：

def CDLDOJI_(open, high, low, close):
    _len = len(open)
    open, high, low, close = pd.Series(open), pd.Series(high), pd.Series(low), pd.Series(close)
    if _len >= 11:
        real_body = abs(open - close)
        high_low_range = abs(high - low)
        doji = high_low_range.rolling(10).mean().shift(1) * 0.1
        condition = real_body < doji
        cdldoji = np.where(condition, 100, 0)
    else:
        cdldoji = np.zeros(_len).astype(int)
    return cdldoji

CDLDRAGONFLYDOJI （蜻蜓十字）

python函数原型：

cdldragonflydoji= CDLDRAGONFLYDOJI (open, high, low, close)

解释：

单k线模型，T字线

重构代码（来自ta_CDLDRAGONFLYDOJI.c， 有修改， 只保留逻辑）：

def CDLDRAGONFLYDOJI_(open, high, low, close):
    _len = len(open)
    open, high, low, close = pd.Series(open), pd.Series(high), pd.Series(low), pd.Series(close)
    if _len >= 11:
        real_body = abs(open - close)
        high_low_range = abs(high - low)
        shadow_upper = high - np.maximum(close, open)
        shadow_lower = np.minimum(close, open) - low
        shadow_very_short = high_low_range.rolling(10).mean().shift(1) * 0.1
        doji = high_low_range.rolling(10).mean().shift(1) * 0.1
        condition = (real_body < doji) & (shadow_upper < shadow_very_short) & (shadow_lower > shadow_very_short)
        cdldragonflydoji = np.where(condition, 100, 0)
    else:
        cdldragonflydoji = np.zeros(_len).astype(int)
    return cdldragonflydoji

CDLGRAVESTONEDOJI （墓碑十字）

python函数原型：

cdlgravestonedoji = CDLGRAVESTONEDOJI(open, high, low, close)

解释：

单k线模型，倒T字线

重构代码（来自ta_CDLGRAVESTONEDOJI.c， 有修改， 只保留逻辑）：

def CDLGRAVESTONEDOJI_(open, high, low, close):
    _len = len(open)
    open, high, low, close = pd.Series(open), pd.Series(high), pd.Series(low), pd.Series(close)
    if _len >= 11:
        real_body = abs(open - close)
        high_low_range = abs(high - low)
        shadow_upper = high - np.maximum(close, open)
        shadow_lower = np.minimum(close, open) - low
        shadow_very_short = high_low_range.rolling(10).mean().shift(1) * 0.1
        doji = high_low_range.rolling(10).mean().shift(1) * 0.1
        condition = (real_body < doji) & (shadow_upper > shadow_very_short) & (shadow_lower < shadow_very_short)
        cdlgravestonedoji = np.where(condition, 100, 0)
    else:
        cdlgravestonedoji = np.zeros(_len).astype(int)
    return cdlgravestonedoji

三. 双K线模型

CDLCOUNTERATTACK （反击线）

python函数原型：

cdlcounterattack = CDLCOUNTERATTACK(open, high, low, close)

解释：

双k线模型，一大阴一大阳，且收盘价接近，
延续第二根K线地行情

重构代码（来自ta_CDLCOUNTERATTACK.c， 有修改， 只保留逻辑）：

def CDLCOUNTERATTACK_(open, high, low, close):
    _len = len(open)
    open, high, low, close = pd.Series(open), pd.Series(high), pd.Series(low), pd.Series(close)
    if _len >= 12:
        real_body = abs(open - close)
        high_low_range = abs(high - low)
        equal = high_low_range.rolling(5).mean().shift(1) * 0.05
        body_long = real_body.rolling(10).mean().shift(1)
        condition_1_1 = (close.shift(1) > open.shift(1)) & \
                        (close < open)
        condition_1_2 = (close.shift(1) < open.shift(1)) & \
                        (close > open)
        condition_2 = (real_body.shift(1) > body_long.shift(1)) & \
                      (real_body > body_long) & \
                      (close <= close.shift(1) + equal.shift(1)) & \
                      (close >= close.shift(1) - equal.shift(1))

        cdlcounterattack = np.where(condition_1_1 & condition_2, -100, 0)
        cdlcounterattack = np.where(condition_1_2 & condition_2, 100, cdlcounterattack)
    else:
        cdlcounterattack = np.zeros(_len).astype(int)
    return cdlcounterattack

CDLDARKCLOUDCOVER（乌云压顶）

python函数原型：

cdldarkcloudcover = CDLDARKCLOUDCOVER(open, high, low, close, penetration=0.5)

解释：

双k线模型，第一根大阳
第二根阴线开盘高于前一日最高
收盘再前一日实体下半部分
见顶信号

重构代码（来自ta_CDLDARKCLOUDCOVER.c， 有修改， 只保留逻辑）：

def CDLDARKCLOUDCOVER_(open, high, low, close, penetration=0.5):
    _len = len(open)
    open, high, low, close = pd.Series(open), pd.Series(high), pd.Series(low), pd.Series(close)
    if _len >= 12:
        real_body = abs(open - close)
        high_low_range = abs(high - low)
        equal = high_low_range.rolling(5).mean().shift(1) * 0.05
        body_long = real_body.rolling(10).mean().shift(1)
        condition_1 = (close.shift(1) > open.shift(1)) & \
                      (real_body.shift(1) > body_long.shift(1))
        condition_2 = (close < open) & \
                      (open > high.shift(1)) & \
                      (close > open.shift(1)) & \
                      (close < close.shift(1) - real_body.shift(1) * penetration)
        cdldarkcloudcover = np.where(condition_1 & condition_2, -100, 0)
    else:
        cdldarkcloudcover = np.zeros(_len).astype(int)
    return cdldarkcloudcover

CDLDOJISTAR（十字星）

python函数原型：

cdldojistar = CDLDOJISTAR(open, high, low, close)

解释：

双k线模型，第一根大阴或者大阳
第二根高开or低开，十字星
反转信号

重构代码（来自ta_CDLDOJISTAR.c， 有修改， 只保留逻辑）：

def CDLDOJISTAR_(open, high, low, close):
    _len = len(open)
    open, high, low, close = pd.Series(open), pd.Series(high), pd.Series(low), pd.Series(close)
    if _len >= 12:
        real_body = abs(open - close)
        high_low_range = abs(high - low)
        doji = high_low_range.rolling(10).mean().shift(1) * 0.1
        body_long = real_body.rolling(10).mean().shift(1)
        condition_1_1 = (close.shift(1) > open.shift(1)) & \
                        (np.maximum(close, open) > close.shift(1))
        condition_1_2 = (close.shift(1) < open.shift(1)) & \
                        (np.minimum(close, open) < close.shift(1))
        condition_2 = (real_body.shift(1) > body_long.shift(1)) & \
                      (real_body <= doji)
        cdldojistar = np.where(condition_1_1 & condition_2, -100, 0)
        cdldojistar = np.where(condition_1_2 & condition_2, 100, cdldojistar)
    else:
        cdldojistar = np.zeros(_len).astype(int)
    return cdldojistar

CDLENGULFING （吞噬模式）

python函数原型：

cdlengulfing= CDLENGULFING(open, high, low, close)

解释：

两日K线模式，分多头吞噬和空头吞噬，以多头吞噬为例，第一日为阴线， 第二日阳线，第一日的开盘价和收盘价在第二日开盘价收盘价之内，但不能完全相同。

重构代码（来自ta_CDLENGULFING.c， 有修改， 只保留逻辑）：

def CDLENGULFING_(open, high, low, close):
    _len = len(open)
    open, high, low, close = pd.Series(open), pd.Series(high), pd.Series(low), pd.Series(close)
    if _len >= 12:
        condition_1 = (close.shift(1) < open.shift(1)) & (close > open) & \
                      (close > open.shift(1)) & (open < close.shift(1))
        condition_2 = (close.shift(1) > open.shift(1)) & (close < open) & \
                      (close < open.shift(1)) & (open > close.shift(1))
        cdlengulfing = np.where(condition_1, 100, 0)
        cdlengulfing = np.where(condition_2, -100, cdlengulfing)
    else:
        cdlengulfing = np.zeros(_len).astype(int)
    return cdlengulfing

CDLHAMMER（锤头）

python函数原型：

cdlhammer = CDLHAMMER(open, high, low, close)

解释：

两日K线模式，第二日为长下影短上影的短实体，且实体下沿低于前一日最低价（Near）

重构代码（来自ta_CDLHAMMER.c， 有修改， 只保留逻辑）：

def CDLHAMMER_(open, high, low, close):
    _len = len(open)
    open, high, low, close = pd.Series(open), pd.Series(high), pd.Series(low), pd.Series(close)
    if _len >= 12:
        condition_1 = (close.shift(1) < open.shift(1)) & (close > open) & \
                      (close > open.shift(1)) & (open < close.shift(1))
        condition_2 = (close.shift(1) > open.shift(1)) & (close < open) & \
                      (close < open.shift(1)) & (open > close.shift(1))
        cdlengulfing = np.where(condition_1, 100, 0)
        cdlengulfing = np.where(condition_2, -100, cdlengulfing)
    else:
        cdlengulfing = np.zeros(_len).astype(int)
    return cdlengulfing

四. 三K线模型

CDL2CROW（双乌鸦）

python函数原型：

cdl2crows= CDL2CROWS(open, high, low, close)

解释：

对于双乌鸦形状的定义，中文百科和ta-lib中定义的方式不一样，

中文百科中第一天长阳线后，第二天高开收阴, 形成缺口，第三天的大阴线包裹住第二天的阴线

ta-lib中第一根和第二跟定义相同，第三跟阴线开盘价再第二根阴线开收之间，收盘价再第一根阳线开收之间

投资逻辑（仅供参考，不宜借鉴）：按照主散博弈模型，主力推高出大阳线吸引散户关注，高开吸引散户进场，主力出货一部分，再次高开吸引散户进场，主力在次出货

第一根长阳线：
$$OPE{N}_{-2}-CLOS{E}_{-2}>AVG(|OPEN-CLOSE|)$$
第二根阴线：
$$OPE{N}_{-1}>CLOS{E}_{-1}>CLOS{E}_{-2}$$
第三根阴线：
$$OPE{N}_{-1}<OPE{N}_0<CLOS{E}_{-2}$$

重构代码（来自ta_CDL2CROW.c， 有修改， 只保留逻辑）：

def CDL2CROWS_(open, high, low, close):
    _len = len(open)
    open, high, low, close = pd.Series(open), pd.Series(high), pd.Series(low), pd.Series(close)
    if _len >= 13:
        real_body = abs(open - close)
        body_long = real_body.rolling(10).mean().shift(1)
        # 第一根大阳线
        condition1 = (close.shift(2) - open.shift(2)) > body_long.shift(2)
        # 第二根阴线
        condition2 = (open.shift(1) > close.shift(1)) & (close.shift(1) > close.shift(2))
        # 第三根阴线
        condition3 = (open.shift(1) > open) & (open > close.shift(1)) & \
                     (close.shift(2) > close) & (close > open.shift(2))
        cdl2crow = np.where(condition1 & condition2 & condition3, -100, 0)
        cdl2crow[:12] = 0
    else:
        cdl2crow = np.zeros(_len).astype(int)
    return cdl2crow

CDL3BLACKCROW（三乌鸦）

python函数原型：

cdl3blackcrows= CDL3BLACKCROWS(open, high, low, close)

解释：

（1）连续三天长阴线。（ta-lib无此要求）
（2）每天的收盘出现新低。
（3）每天的开盘在前一天的实体内。
（4）每天的收盘等于或接近当天的最低。
（5）第一根阴线跌破前阳线最高点（ta-lib特有要求）

第一根阳线：
$$OPE{N}_{-3}-CLOS{E}_{-3}>0$$
第一根阴线：
O P E N − 2 > C L O S E − 2 C L O S E − 2 < H I G H − 3 C L O S E − 2 − L O W