|
1
|
|
|
""" |
|
2
|
|
|
Technical Analysis Factors |
|
3
|
|
|
-------------------------- |
|
4
|
|
|
""" |
|
5
|
|
|
from bottleneck import ( |
|
6
|
|
|
nanargmax, |
|
7
|
|
|
nanmax, |
|
8
|
|
|
nanmean, |
|
9
|
|
|
nansum, |
|
10
|
|
|
) |
|
11
|
|
|
from numpy import ( |
|
12
|
|
|
abs, |
|
13
|
|
|
clip, |
|
14
|
|
|
diff, |
|
15
|
|
|
fmax, |
|
16
|
|
|
inf, |
|
17
|
|
|
isnan, |
|
18
|
|
|
NINF, |
|
19
|
|
|
) |
|
20
|
|
|
from numexpr import evaluate |
|
21
|
|
|
|
|
22
|
|
|
from zipline.pipeline.data import USEquityPricing |
|
23
|
|
|
from zipline.pipeline.term import SingleInputMixin |
|
24
|
|
|
from zipline.utils.control_flow import ignore_nanwarnings |
|
25
|
|
|
from .factor import CustomFactor |
|
26
|
|
|
|
|
27
|
|
|
|
|
28
|
|
|
class Returns(CustomFactor): |
|
29
|
|
|
""" |
|
30
|
|
|
Calculates the percent change in close price over the given window_length. |
|
31
|
|
|
|
|
32
|
|
|
**Default Inputs**: [USEquityPricing.close] |
|
33
|
|
|
""" |
|
34
|
|
|
inputs = [USEquityPricing.close] |
|
35
|
|
|
|
|
36
|
|
|
def compute(self, today, assets, out, close): |
|
37
|
|
|
out[:] = (close[-1] - close[0]) / close[0] |
|
38
|
|
|
|
|
39
|
|
|
|
|
40
|
|
|
class RSI(CustomFactor, SingleInputMixin): |
|
41
|
|
|
""" |
|
42
|
|
|
Relative Strength Index |
|
43
|
|
|
|
|
44
|
|
|
**Default Inputs**: [USEquityPricing.close] |
|
45
|
|
|
|
|
46
|
|
|
**Default Window Length**: 15 |
|
47
|
|
|
""" |
|
48
|
|
|
window_length = 15 |
|
49
|
|
|
inputs = (USEquityPricing.close,) |
|
50
|
|
|
|
|
51
|
|
|
def compute(self, today, assets, out, closes): |
|
52
|
|
|
diffs = diff(closes, axis=0) |
|
53
|
|
|
ups = nanmean(clip(diffs, 0, inf), axis=0) |
|
54
|
|
|
downs = abs(nanmean(clip(diffs, -inf, 0), axis=0)) |
|
55
|
|
|
return evaluate( |
|
56
|
|
|
"100 - (100 / (1 + (ups / downs)))", |
|
57
|
|
|
local_dict={'ups': ups, 'downs': downs}, |
|
58
|
|
|
global_dict={}, |
|
59
|
|
|
out=out, |
|
60
|
|
|
) |
|
61
|
|
|
|
|
62
|
|
|
|
|
63
|
|
|
class SimpleMovingAverage(CustomFactor, SingleInputMixin): |
|
64
|
|
|
""" |
|
65
|
|
|
Average Value of an arbitrary column |
|
66
|
|
|
|
|
67
|
|
|
**Default Inputs**: None |
|
68
|
|
|
|
|
69
|
|
|
**Default Window Length**: None |
|
70
|
|
|
""" |
|
71
|
|
|
# numpy's nan functions throw warnings when passed an array containing only |
|
72
|
|
|
# nans, but they still returns the desired value (nan), so we ignore the |
|
73
|
|
|
# warning. |
|
74
|
|
|
ctx = ignore_nanwarnings() |
|
75
|
|
|
|
|
76
|
|
|
def compute(self, today, assets, out, data): |
|
77
|
|
|
out[:] = nanmean(data, axis=0) |
|
78
|
|
|
|
|
79
|
|
|
|
|
80
|
|
|
class WeightedAverageValue(CustomFactor): |
|
81
|
|
|
""" |
|
82
|
|
|
Helper for VWAP-like computations. |
|
83
|
|
|
|
|
84
|
|
|
**Default Inputs:** None |
|
85
|
|
|
|
|
86
|
|
|
**Default Window Length:** None |
|
87
|
|
|
""" |
|
88
|
|
|
def compute(self, today, assets, out, base, weight): |
|
89
|
|
|
out[:] = nansum(base * weight, axis=0) / nansum(weight, axis=0) |
|
90
|
|
|
|
|
91
|
|
|
|
|
92
|
|
|
class VWAP(WeightedAverageValue): |
|
93
|
|
|
""" |
|
94
|
|
|
Volume Weighted Average Price |
|
95
|
|
|
|
|
96
|
|
|
**Default Inputs:** [USEquityPricing.close, USEquityPricing.volume] |
|
97
|
|
|
|
|
98
|
|
|
**Default Window Length:** None |
|
99
|
|
|
""" |
|
100
|
|
|
inputs = (USEquityPricing.close, USEquityPricing.volume) |
|
101
|
|
|
|
|
102
|
|
|
|
|
103
|
|
|
class MaxDrawdown(CustomFactor, SingleInputMixin): |
|
104
|
|
|
""" |
|
105
|
|
|
Max Drawdown |
|
106
|
|
|
|
|
107
|
|
|
**Default Inputs:** None |
|
108
|
|
|
|
|
109
|
|
|
**Default Window Length:** None |
|
110
|
|
|
""" |
|
111
|
|
|
ctx = ignore_nanwarnings() |
|
112
|
|
|
|
|
113
|
|
|
def compute(self, today, assets, out, data): |
|
114
|
|
|
drawdowns = fmax.accumulate(data, axis=0) - data |
|
115
|
|
|
drawdowns[isnan(drawdowns)] = NINF |
|
116
|
|
|
drawdown_ends = nanargmax(drawdowns, axis=0) |
|
117
|
|
|
|
|
118
|
|
|
# TODO: Accelerate this loop in Cython or Numba. |
|
119
|
|
|
for i, end in enumerate(drawdown_ends): |
|
120
|
|
|
peak = nanmax(data[:end + 1, i]) |
|
121
|
|
|
out[i] = (peak - data[end, i]) / data[end, i] |
|
122
|
|
|
|
quantopian /
zipline
