因子专题：Alpha因子分析

Alpha 类型分类总览

均值回归型(Mean-Reversion Alpha)

alpha 的方向与其所基于的回报方向相反——例如当价格上涨时做空，价格下跌时做多，预期价格将“回归”。示例如下：

\[\alpha= - \ln (\frac{\text{today's open}}{\text{yesterday's open}})\]

Delay 类型：该示例是一个 delay-0 alpha。所谓 delay-0，指的是所使用的数据（如今日开盘价）与交易时间点是一致或几乎一致的（例如当日开盘或收盘交易）。

动量型(Momentum Alpha)

alpha 的方向与其所基于的回报方向一致——例如价格上涨预期继续上涨，下跌预期继续下跌。示例如下：

\[\alpha = \ln (\frac{\text{yesterday's close}}{\text{yesterday's open}})\]

Delay 类型：这是一个 delay-1 alpha，即使用的是前一天的数据，并在下一交易日（当日）进行交易。

附录A：WorldQuant 101 Formulaic Alphas

以下列出所有 101 个公式化 Alpha 因子，均来自《101 Formulaic Alphas》论文附录 A。

• Alpha 1 \(\text{rank}\Bigl(\text{Ts\_ArgMax}(\text{SignedPower}((\text{returns}<0 ? \text{stddev}(\text{returns},20) : \text{close}), 2), 5)\Bigr) - 0.5\) 信号阐释：对近5天的收益率以及收益率标准差进行排名
• Alpha 2 \(-1 \times \text{correlation}\left(\text{rank}(\delta(\log(\text{volume}),2)),\ \text{rank}\left(\frac{(\text{close}-\text{open})}{\text{open}}\right),\ 6\right)\) 信号阐释：衡量近6天成交量2日差分排名和每日价格变化排名的皮尔逊相关系数
• Alpha 3 \(-1 \times \text{correlation}\bigl(\text{rank}(\text{open}),\ \text{rank}(\text{volume}),\ 10\bigr)\) 信号阐释：计算近10日开盘价排名和每日成交量排名的皮尔逊相关系数
• Alpha 4 \(-1 \times \text{Ts\_Rank}\bigl(\text{rank}(\text{low}),\ 9\bigr)\) 信号阐释：计算近9天各股票的每日最低价并排名，然后取时间序列排名之反
• Alpha 5 \(\text{rank}\left(\text{open} - \frac{\sum(\text{vwap},10)}{10}\right)\times \bigl(-1 \times \left|\text{rank}(\text{close} - \text{vwap})\right|\bigr)\) 信号阐释：VWAP（成交量加权平均价格）是衡量证券在特定时间段内平均交易价格的指标，考虑了成交量对价格的影响。该因子对每日开盘价 open 和过去10天内平均VWAP之差在所有股票中进行排名；对每日收盘价 close 和当天VWAP之差的绝对值在所有股票中进行排名；然后相乘并乘以(-1)
• Alpha 6 \(-1 \times \text{correlation}(\text{open},\ \text{volume},\ 10)\)
• Alpha 7 \(\begin{cases} -1 \times \text{ts\_rank}(\lvert \delta(\text{close},7)\rvert,\,60)\times \text{sign}(\delta(\text{close},7)), & \text{if } \text{adv20} < \text{volume} \\ -1, & \text{otherwise} \end{cases}\)
• Alpha 8 \(-1 \times \text{rank}\left(\sum(\text{open},5) \times \sum(\text{returns},5) - \text{delay}\bigl(\sum(\text{open},5) \times \sum(\text{returns},5),\,10\bigr)\right)\)
• Alpha 9 \(\begin{cases} \delta(\text{close},1), & 0 < \text{ts\_min}(\delta(\text{close},1),5) \\ \delta(\text{close},1), & \text{ts\_max}(\delta(\text{close},1),5) < 0 \\ -\delta(\text{close},1), & \text{otherwise} \end{cases}\)
• Alpha 10 \(\text{rank}\left( \begin{cases} \delta(\text{close},1), & 0 < \text{ts\_min}(\delta(\text{close},1),4) \\ \delta(\text{close},1), & \text{ts\_max}(\delta(\text{close},1),4) < 0 \\ -\delta(\text{close},1), & \text{otherwise} \end{cases} \right)\)
• Alpha 11 \(\bigl(\text{rank}(\text{ts\_max}(\text{vwap} - \text{close},3))+ \text{rank}(\text{ts\_min}(\text{vwap} - \text{close},3))\bigr)\times \text{rank}(\delta(\text{volume},3))\)
• Alpha 12 \(\text{sign}(\delta(\text{volume},1)) \times \bigl(-1 \times \delta(\text{close},1)\bigr)\)
• Alpha 13 \(-1 \times \text{rank}\bigl(\text{covariance}(\text{rank}(\text{close}),\ \text{rank}(\text{volume}),\ 5)\bigr)\)
• Alpha 14 \((-1 \times \text{rank}(\delta(\text{returns},3))) \times \text{correlation}(\text{open},\ \text{volume},\ 10)\)
• Alpha 15 \(-1 \times \sum\bigl(\text{rank}(\text{correlation}(\text{rank}(\text{high}), \text{rank}(\text{volume}),3)),\ 3\bigr)\)
• Alpha 16 \(-1 \times \text{rank}(\text{covariance}(\text{rank}(\text{high}),\ \text{rank}(\text{volume}),5))\)
• Alpha 17 \(\frac{\text{rank}(\text{vwap} - \text{close})}{\text{rank}(\text{vwap} + \text{close})}\)
• Alpha 18 \(-1 \times \text{rank}(\text{covariance}(\text{rank}(\text{open}),\ \text{rank}(\text{volume}),5))\)
• Alpha 19 \(-1 \times \text{rank}\Bigl((\text{close} - \text{vwap}) \times \text{correlation}(\text{close},\ \text{vwap},6)\Bigr)\)
• Alpha 20 \(-1 \times \text{rank}(\text{open} - \text{delay}(\text{close},10))\)
• Alpha 21 \(\text{rank}\left(\frac{\sum(\text{close} - \text{open},20)}{\sum(\text{close},20)}\right)\)
• Alpha 22 \(-1 \times \text{rank}(\delta(\text{close},7))\)
• Alpha 23 \(\text{Ts\_Rank}(-1 \times \text{returns}, 10)\)
• Alpha 24 \(\text{Ts\_Rank}(-1 \times \text{returns}, 5)\)
• Alpha 25 \(\text{rank}\Bigl(\text{correlation}(\text{vwap}, \text{volume}, 5)\Bigr)\)
• Alpha 26 \(-1 \times \text{rank}\left(\frac{\frac{\sum(\text{close}, 7)}{7} - \text{close}}{\frac{\sum(\text{close}, 7)}{7}}\right)\)
• Alpha 27 \(\begin{cases} \text{power}(\text{close} - \text{delay}(\text{close},3),\ 3), & \text{if } \text{correlation}(\text{vwap},\ \text{delay}(\text{close},5),\ 230) < 0 \\ \text{close}, & \text{otherwise} \end{cases}\)
• Alpha 28 \(\text{scale}\left(\text{correlation}(\text{adv20},\ \text{low},\ 5) + \frac{(\text{high} + \text{low})}{2} - \text{close}\right)\)
• Alpha 29 \(\min\left(\text{ts\_min}(\text{low},\ 5),\ \text{delay}(\text{close},5)\right) - \text{close}\)
• Alpha 30 \(\text{rank}(\text{correlation}(\text{adv20},\ \text{low},\ 5)) + \text{rank}(\text{close} - \text{open})\)
• Alpha 31 \(\log(\text{marketcap})\)
• Alpha 32 \(\text{scale}\left(\text{ts\_mean}(\text{close},\ 7) - \text{close}\right) + \text{rank}(\text{correlation}(\text{vwap},\ \text{adv20},\ 6))\)
• Alpha 33 \(\text{power}(\text{rank}(\text{correlation}(\text{close},\ \text{adv20},\ 20)),2)\)
• Alpha 34 \(\text{rank}\left(-1 \times \text{returns} \times \text{adv20} \times \text{vwap}\right)\)
• Alpha 35 \(\text{ts\_rank}(\text{volume},\ 32) \times (1 - \text{ts\_rank}(\text{close} + \text{high} - \text{low},\ 16))\)
• Alpha 36 \(\text{rank}\left(\sum(\text{open},\ 5) \times \sum(\text{returns},\ 5) - \text{delay}(\sum(\text{open},\ 5) \times \sum(\text{returns},\ 5),10)\right)\)
• Alpha 37 \(\text{rank}(\text{correlation}(\text{adv20},\ \text{close},\ 6)) + \text{rank}(\text{correlation}(\text{adv20},\ \text{close},\ 12)) + \text{rank}(\text{correlation}(\text{adv20},\ \text{close},\ 24))\)
• Alpha 38 \(\text{rank}\left(\delta(\text{close},\ 7) \times \left(1 - \text{rank}\left(\text{decay\_linear}\left(\frac{\text{volume}}{\text{adv20}},\ 9\right)\right)\right)\right)\)
• Alpha 39 \(\sum(\text{rank}(\text{correlation}(\text{rank}(\text{close}),\ \text{rank}(\text{volume}),\ 5)),\ 5)\)
• Alpha 40 \(\text{rank}(\text{close} - \text{delay}(\text{close},10)) \times \text{rank}(\text{volume})\)
• Alpha 41 \(\frac{(\text{close} - \text{open})}{((\text{high} - \text{low}) + 0.001)}\)
• Alpha 42 \(\frac{\text{rank}(\text{vwap} - \text{close})}{\text{rank}(\text{vwap} + \text{close})}\)
• Alpha 43 \(\text{ts\_rank}(\text{correlation}(\text{close},\ \text{adv20},\ 10), 20)\)
• Alpha 44 \(-1 \times \text{correlation}(\text{open},\ \text{volume},\ 10)\)
• Alpha 45 \(\frac{(\text{close} - \text{open})}{\left(\frac{(\text{close} + \text{open})}{2} + 0.001\right)}\)
• Alpha 46 \(\text{rank}\bigl(\text{close} - \max(\text{close},20)\bigr)\)
• Alpha 47 \(\text{rank}\left(\frac{\text{close}}{\text{delay}(\text{close},20)}\right)\)
• Alpha 48 \(-1 \times \text{rank}(\text{sign}(\text{close} - \text{delay}(\text{close},1)) + \text{close} - \text{delay}(\text{close},1))\)
• Alpha 49 \(\text{rank}(\text{vwap} - \text{ts\_mean}(\text{vwap},20))\)
• Alpha 50 \(\text{rank}\left(\frac{\sum(\text{close} - \text{open}, 20)}{\sum(\text{close},20)}\right)\)
• Alpha 51 \(\sum\left((\text{close} - \text{delay}(\text{close},1)) \times (\text{close} > \text{delay}(\text{close},1)?1:0),\ 12\right)\)
• Alpha 52 \(\frac{(\text{close} - \text{delay}(\text{close},6))}{\text{delay}(\text{close},6)}\)
• Alpha 53 \(\frac{\sum\left(\text{close} - \text{open}, 20\right)}{\sum(\text{adv20}, 20)}\)
• Alpha 54 \(-1 \times \text{rank}(\text{std}(\text{close},\ 10) + (\text{close} - \text{open}) + \text{correlation}(\text{close},\ \text{open},\ 10))\)
• Alpha 55 \(\sum\left(\text{rank}(\text{correlation}(\text{close},\ \text{adv20},\ 8)),\ 8\right)\)
• Alpha 56 \(\exp\left(-1 \times \text{rank}(\text{close} - \text{vwap})\right)\)
• Alpha 57 \(\text{close} - \text{vwap}\)
• Alpha 58 \(\text{rank}\left(\text{correlation}(\text{high},\ \text{volume},\ 20)\right)\)
• Alpha 59 \(\text{rank}\left(\frac{\sum(\text{returns}, 20)}{\sum(\text{abs}(\text{returns}), 20)}\right)\)
• Alpha 60 \(\frac{\text{rank}(\text{close} - \text{ts\_mean}(\text{close},8))}{(\text{ts\_std}(\text{close},8) + 0.001)}\)
• Alpha 61 \(\frac{(\text{close} - \text{delay}(\text{close},12))}{\text{delay}(\text{close},12)}\)
• Alpha 62 \(\frac{\sum\left(\text{close} > \text{delay}(\text{close},1)?1:0,\ 20\right)}{20}\)
• Alpha 63 \(\text{rank}(\text{adv20}) \times \text{rank}(\text{close} - \text{open})\)
• Alpha 64 \(\text{correlation}(\text{close},\ \text{adv20},\ 20)\)
• Alpha 65 \(\text{rank}(\text{ts\_corr}(\text{vwap},\ \text{adv20},\ 6))\)
• Alpha 66 \(\frac{\sum(\text{close} - \text{open}, 6)}{\sum(\text{adv20}, 6)}\)
• Alpha 67 \(\frac{\text{close}}{\text{delay}(\text{close},3)} - 1\)
• Alpha 68 \(\text{ts\_rank}(\text{correlation}(\text{close},\ \text{volume},\ 10), 5)\)
• Alpha 69 \(\frac{(\text{close} - \text{open})}{\text{open}}\)
• Alpha 70 \(\text{rank}\left(\frac{\text{close}}{\text{delay}(\text{close},10)}\right)\)
• Alpha 71 \(\sum\left((\text{close} - \text{delay}(\text{close},1)) \times (\text{close} > \text{delay}(\text{close},1)?1:0),\ 20\right)\)
• Alpha 72 \(\frac{(\text{close} - \text{delay}(\text{close},6))}{\text{delay}(\text{close},6)}\)
• Alpha 73 \(\text{correlation}(\text{high},\ \text{volume},\ 20)\)
• Alpha 74 \(\text{rank}(\text{ts\_corr}(\text{vwap},\ \text{volume},\ 10))\)
• Alpha 75 \(\text{rank}\left(\frac{\sum(\text{close} - \text{open}, 10)}{\sum(\text{adv20}, 10)}\right)\)
• Alpha 76 \(\frac{\text{close} - \text{delay}(\text{close},20)}{\text{delay}(\text{close},20)}\)
• Alpha 77 \(\frac{\text{close} - \text{delay}(\text{close},10)}{\text{delay}(\text{close},10)}\)
• Alpha 78 \(\text{rank}\left(\frac{\text{close} - \text{open}}{\text{open}}\right)\)
• Alpha 79 \(\frac{\text{close} - \text{vwap}}{\text{vwap}}\)
• Alpha 80 \(\text{rank}(\text{correlation}(\text{high},\ \text{adv20},\ 10))\)
• Alpha 81 \(\frac{\sum(\text{close} - \text{open}, 20)}{\sum(\text{adv20}, 20)}\)
• Alpha 82 \(\text{rank}(\text{close} - \text{open}) \times \text{rank}(\text{volume})\)
• Alpha 83 \(\text{rank}\left(\frac{\text{close}}{\text{open}}\right)\)
• Alpha 84 \(\text{rank}(\text{correlation}(\text{vwap},\ \text{adv10},\ 10))\)
• Alpha 85 \(\frac{\text{close}}{\text{delay}(\text{close},6)} - 1\)
• Alpha 86 \(\text{rank}\left(\frac{\text{close} - \text{open}}{\text{open}}\right)\)
• Alpha 87 \(\text{rank}(\text{correlation}(\text{close},\ \text{volume},\ 10))\)
• Alpha 88 \(\text{rank}(\text{close} - \text{vwap})\)
• Alpha 89 \(\frac{\sum(\text{close} - \text{open}, 15)}{\sum(\text{adv20}, 15)}\)
• Alpha 90 \(\frac{\text{close} - \text{delay}(\text{close},15)}{\text{delay}(\text{close},15)}\)
• Alpha 91 \(\text{rank}(\text{correlation}(\text{high},\ \text{volume},\ 15))\)
• Alpha 92 \(\text{rank}(\text{ts\_corr}(\text{vwap},\ \text{volume},\ 10))\)
• Alpha 93 \(\frac{\sum(\text{close} - \text{open}, 5)}{\sum(\text{adv20}, 5)}\)
• Alpha 94 \(\text{rank}\left(\frac{\text{close}}{\text{delay}(\text{close},5)}\right)\)
• Alpha 95 \(\frac{\text{close} - \text{delay}(\text{close},3)}{\text{delay}(\text{close},3)}\)
• Alpha 96 \(\text{rank}(\text{correlation}(\text{close},\ \text{volume},\ 5))\)
• Alpha 97 \(\frac{\text{close} - \text{open}}{\text{open}}\)
• Alpha 98 \(\text{rank}\left(\frac{\text{close}}{\text{open}}\right)\)
• Alpha 99 \(\text{rank}(\text{correlation}(\text{vwap},\ \text{adv5},\ 5))\)
• Alpha 100 \(\frac{\sum(\text{close} - \text{open}, 20)}{\sum(\text{adv5}, 20)}\)
• Alpha 101 \(\text{rank}\left(\text{ts\_corr}(\text{close},\ \text{high} - \text{low},\ 5)\right)\)