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:
α=−ln(yesterday’s opentoday’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:
α=ln(yesterday’s openyesterday’s close)
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。
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
rank(Ts_ArgMax(SignedPower((returns<0?stddev(returns,20):close),2),5))−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×correlation(rank(δ(log(volume),2)), rank(open(close−open)), 6)
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×correlation(rank(open), rank(volume), 10)
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×Ts_Rank(rank(low), 9)
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
rank(open−10∑(vwap,10))×(−1×∣rank(close−vwap)∣)
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×correlation(open, volume, 10)
- Alpha 7
{−1×ts_rank(∣δ(close,7)∣,60)×sign(δ(close,7)),−1,if adv20<volumeotherwise
- Alpha 8
−1×rank(∑(open,5)×∑(returns,5)−delay(∑(open,5)×∑(returns,5),10))
- Alpha 9
⎩⎨⎧δ(close,1),δ(close,1),−δ(close,1),0<ts_min(δ(close,1),5)ts_max(δ(close,1),5)<0otherwise
- Alpha 10
rank⎩⎨⎧δ(close,1),δ(close,1),−δ(close,1),0<ts_min(δ(close,1),4)ts_max(δ(close,1),4)<0otherwise
- Alpha 11
(rank(ts_max(vwap−close,3))+rank(ts_min(vwap−close,3)))×rank(δ(volume,3))
- Alpha 12
sign(δ(volume,1))×(−1×δ(close,1))
- Alpha 13
−1×rank(covariance(rank(close), rank(volume), 5))
- Alpha 14
(−1×rank(δ(returns,3)))×correlation(open, volume, 10)
- Alpha 15
−1×∑(rank(correlation(rank(high),rank(volume),3)), 3)
- Alpha 16
−1×rank(covariance(rank(high), rank(volume),5))
- Alpha 17
rank(vwap+close)rank(vwap−close)
- Alpha 18
−1×rank(covariance(rank(open), rank(volume),5))
- Alpha 19
−1×rank((close−vwap)×correlation(close, vwap,6))
- Alpha 20
−1×rank(open−delay(close,10))
- Alpha 21
rank(∑(close,20)∑(close−open,20))
- Alpha 22
−1×rank(δ(close,7))
- Alpha 23
Ts_Rank(−1×returns,10)
- Alpha 24
Ts_Rank(−1×returns,5)
- Alpha 25
rank(correlation(vwap,volume,5))
- Alpha 26
−1×rank(7∑(close,7)7∑(close,7)−close)
- Alpha 27
{power(close−delay(close,3), 3),close,if correlation(vwap, delay(close,5), 230)<0otherwise
- Alpha 28
scale(correlation(adv20, low, 5)+2(high+low)−close)
- Alpha 29
min(ts_min(low, 5), delay(close,5))−close
- Alpha 30
rank(correlation(adv20, low, 5))+rank(close−open)
- Alpha 31
log(marketcap)
- Alpha 32
scale(ts_mean(close, 7)−close)+rank(correlation(vwap, adv20, 6))
- Alpha 33
power(rank(correlation(close, adv20, 20)),2)
- Alpha 34
rank(−1×returns×adv20×vwap)
- Alpha 35
ts_rank(volume, 32)×(1−ts_rank(close+high−low, 16))
- Alpha 36
rank(∑(open, 5)×∑(returns, 5)−delay(∑(open, 5)×∑(returns, 5),10))
- Alpha 37
rank(correlation(adv20, close, 6))+rank(correlation(adv20, close, 12))+rank(correlation(adv20, close, 24))
- Alpha 38
rank(δ(close, 7)×(1−rank(decay_linear(adv20volume, 9))))
- Alpha 39
∑(rank(correlation(rank(close), rank(volume), 5)), 5)
- Alpha 40
rank(close−delay(close,10))×rank(volume)
- Alpha 41
((high−low)+0.001)(close−open)
- Alpha 42
rank(vwap+close)rank(vwap−close)
- Alpha 43
ts_rank(correlation(close, adv20, 10),20)
- Alpha 44
−1×correlation(open, volume, 10)
- Alpha 45
(2(close+open)+0.001)(close−open)
- Alpha 46
rank(close−max(close,20))
- Alpha 47
rank(delay(close,20)close)
- Alpha 48
−1×rank(sign(close−delay(close,1))+close−delay(close,1))
- Alpha 49
rank(vwap−ts_mean(vwap,20))
- Alpha 50
rank(∑(close,20)∑(close−open,20))
- Alpha 51
∑((close−delay(close,1))×(close>delay(close,1)?1:0), 12)
- Alpha 52
delay(close,6)(close−delay(close,6))
- Alpha 53
∑(adv20,20)∑(close−open,20)
- Alpha 54
−1×rank(std(close, 10)+(close−open)+correlation(close, open, 10))
- Alpha 55
∑(rank(correlation(close, adv20, 8)), 8)
- Alpha 56
exp(−1×rank(close−vwap))
- Alpha 57
close−vwap
- Alpha 58
rank(correlation(high, volume, 20))
- Alpha 59
rank(∑(abs(returns),20)∑(returns,20))
- Alpha 60
(ts_std(close,8)+0.001)rank(close−ts_mean(close,8))
- Alpha 61
delay(close,12)(close−delay(close,12))
- Alpha 62
20∑(close>delay(close,1)?1:0, 20)
- Alpha 63
rank(adv20)×rank(close−open)
- Alpha 64
correlation(close, adv20, 20)
- Alpha 65
rank(ts_corr(vwap, adv20, 6))
- Alpha 66
∑(adv20,6)∑(close−open,6)
- Alpha 67
delay(close,3)close−1
- Alpha 68
ts_rank(correlation(close, volume, 10),5)
- Alpha 69
open(close−open)
- Alpha 70
rank(delay(close,10)close)
- Alpha 71
∑((close−delay(close,1))×(close>delay(close,1)?1:0), 20)
- Alpha 72
delay(close,6)(close−delay(close,6))
- Alpha 73
correlation(high, volume, 20)
- Alpha 74
rank(ts_corr(vwap, volume, 10))
- Alpha 75
rank(∑(adv20,10)∑(close−open,10))
- Alpha 76
delay(close,20)close−delay(close,20)
- Alpha 77
delay(close,10)close−delay(close,10)
- Alpha 78
rank(openclose−open)
- Alpha 79
vwapclose−vwap
- Alpha 80
rank(correlation(high, adv20, 10))
- Alpha 81
∑(adv20,20)∑(close−open,20)
- Alpha 82
rank(close−open)×rank(volume)
- Alpha 83
rank(openclose)
- Alpha 84
rank(correlation(vwap, adv10, 10))
- Alpha 85
delay(close,6)close−1
- Alpha 86
rank(openclose−open)
- Alpha 87
rank(correlation(close, volume, 10))
- Alpha 88
rank(close−vwap)
- Alpha 89
∑(adv20,15)∑(close−open,15)
- Alpha 90
delay(close,15)close−delay(close,15)
- Alpha 91
rank(correlation(high, volume, 15))
- Alpha 92
rank(ts_corr(vwap, volume, 10))
- Alpha 93
∑(adv20,5)∑(close−open,5)
- Alpha 94
rank(delay(close,5)close)
- Alpha 95
delay(close,3)close−delay(close,3)
- Alpha 96
rank(correlation(close, volume, 5))
- Alpha 97
openclose−open
- Alpha 98
rank(openclose)
- Alpha 99
rank(correlation(vwap, adv5, 5))
- Alpha 100
∑(adv5,20)∑(close−open,20)
- Alpha 101
rank(ts_corr(close, high−low, 5))