Skip to content

"Vibe-Trading: Your Personal Trading Agent"

fix(factors): restore Qlib ROC and CORD semantics

Yufeng Heea4b275b311b
diff --git a/agent/src/factors/zoo/qlib158/cord10.py b/agent/src/factors/zoo/qlib158/cord10.py
index 3e2be53d1..19d7f723c 100644
--- a/agent/src/factors/zoo/qlib158/cord10.py
+++ b/agent/src/factors/zoo/qlib158/cord10.py
@@ -2,10 +2,10 @@
# (Apache-2.0). Copyright (c) Microsoft Corporation.
# ============================================================
# 中文名称: 收益率-量变化相关性 10日
-# 简要说明: ts_corr(close/close_prev, log(v/v_prev), 10),10日收益率与成交量变化率的相关系数。
+# 简要说明: ts_corr(close/close_prev, log(v/v_prev + 1), 10),10日价格变化与对数量变比的相关系数。
# 典型用途: 衡量10日价格变动与成交量变动的相关性,反映量价配合程度。
# ============================================================
-"""qlib158 CORD10: formula = \\mathrm{ts\\_corr}(\\mathrm{close}/\\mathrm{close}_{{-1}}, \\log((\\mathrm{volume}+1)/(\\mathrm{volume}_{{-1}}+1)), 10)."""
+"""qlib158 CORD10: formula = \\mathrm{ts\\_corr}(\\mathrm{close}/\\mathrm{close}_{{-1}}, \\log(\\mathrm{volume}/\\mathrm{volume}_{{-1}} + 1), 10)."""
from __future__ import annotations
import numpy as np
@@ -15,7 +15,7 @@ from src.factors.base import safe_div, ts_corr
__alpha_meta__ = {
'id': 'qlib158_cord10',
'theme': ['volume', 'microstructure'],
- 'formula_latex': '\\\\mathrm{ts\\\\_corr}(\\\\mathrm{close}/\\\\mathrm{close}_{{-1}}, \\\\log((\\\\mathrm{volume}+1)/(\\\\mathrm{volume}_{{-1}}+1)), 10)',
+ 'formula_latex': '\\\\mathrm{ts\\\\_corr}(\\\\mathrm{close}/\\\\mathrm{close}_{{-1}}, \\\\log(\\\\mathrm{volume}/\\\\mathrm{volume}_{{-1}} + 1), 10)',
'columns_required': ['close', 'volume'],
'universe': ['equity_us', 'equity_cn', 'equity_hk', 'equity_in', 'equity_kr'],
'frequency': ['1d'],
@@ -29,6 +29,6 @@ def compute(panel: dict[str, pd.DataFrame]) -> pd.DataFrame:
c = panel['close']
v = panel['volume']
c_ret = safe_div(c, c.shift(1))
- v_ret = safe_div(v + 1.0, v.shift(1) + 1.0)
+ v_ret = safe_div(v, v.shift(1)) + 1.0
logvr = np.log(v_ret)
return ts_corr(c_ret, logvr, 10)
diff --git a/agent/src/factors/zoo/qlib158/cord20.py b/agent/src/factors/zoo/qlib158/cord20.py
index 0a4133cd1..90e6a7ba1 100644
--- a/agent/src/factors/zoo/qlib158/cord20.py
+++ b/agent/src/factors/zoo/qlib158/cord20.py
@@ -2,10 +2,10 @@
# (Apache-2.0). Copyright (c) Microsoft Corporation.
# ============================================================
# 中文名称: 收益率-量变化相关性 20日
-# 简要说明: ts_corr(close/close_prev, log(v/v_prev), 20),20日收益率与成交量变化率的相关系数。
+# 简要说明: ts_corr(close/close_prev, log(v/v_prev + 1), 20),20日价格变化与对数量变比的相关系数。
# 典型用途: 衡量20日价格变动与成交量变动的相关性,反映量价配合程度。
# ============================================================
-"""qlib158 CORD20: formula = \\mathrm{ts\\_corr}(\\mathrm{close}/\\mathrm{close}_{{-1}}, \\log((\\mathrm{volume}+1)/(\\mathrm{volume}_{{-1}}+1)), 20)."""
+"""qlib158 CORD20: formula = \\mathrm{ts\\_corr}(\\mathrm{close}/\\mathrm{close}_{{-1}}, \\log(\\mathrm{volume}/\\mathrm{volume}_{{-1}} + 1), 20)."""
from __future__ import annotations
import numpy as np
@@ -15,7 +15,7 @@ from src.factors.base import safe_div, ts_corr
__alpha_meta__ = {
'id': 'qlib158_cord20',
'theme': ['volume', 'microstructure'],
- 'formula_latex': '\\\\mathrm{ts\\\\_corr}(\\\\mathrm{close}/\\\\mathrm{close}_{{-1}}, \\\\log((\\\\mathrm{volume}+1)/(\\\\mathrm{volume}_{{-1}}+1)), 20)',
+ 'formula_latex': '\\\\mathrm{ts\\\\_corr}(\\\\mathrm{close}/\\\\mathrm{close}_{{-1}}, \\\\log(\\\\mathrm{volume}/\\\\mathrm{volume}_{{-1}} + 1), 20)',
'columns_required': ['close', 'volume'],
'universe': ['equity_us', 'equity_cn', 'equity_hk', 'equity_in', 'equity_kr'],
'frequency': ['1d'],
@@ -29,6 +29,6 @@ def compute(panel: dict[str, pd.DataFrame]) -> pd.DataFrame:
c = panel['close']
v = panel['volume']
c_ret = safe_div(c, c.shift(1))
- v_ret = safe_div(v + 1.0, v.shift(1) + 1.0)
+ v_ret = safe_div(v, v.shift(1)) + 1.0
logvr = np.log(v_ret)
return ts_corr(c_ret, logvr, 20)
diff --git a/agent/src/factors/zoo/qlib158/cord30.py b/agent/src/factors/zoo/qlib158/cord30.py
index 8696bf125..2e9841851 100644
--- a/agent/src/factors/zoo/qlib158/cord30.py
+++ b/agent/src/factors/zoo/qlib158/cord30.py
@@ -2,10 +2,10 @@
# (Apache-2.0). Copyright (c) Microsoft Corporation.
# ============================================================
# 中文名称: 收益率-量变化相关性 30日
-# 简要说明: ts_corr(close/close_prev, log(v/v_prev), 30),30日收益率与成交量变化率的相关系数。
+# 简要说明: ts_corr(close/close_prev, log(v/v_prev + 1), 30),30日价格变化与对数量变比的相关系数。
# 典型用途: 衡量30日价格变动与成交量变动的相关性,反映量价配合程度。
# ============================================================
-"""qlib158 CORD30: formula = \\mathrm{ts\\_corr}(\\mathrm{close}/\\mathrm{close}_{{-1}}, \\log((\\mathrm{volume}+1)/(\\mathrm{volume}_{{-1}}+1)), 30)."""
+"""qlib158 CORD30: formula = \\mathrm{ts\\_corr}(\\mathrm{close}/\\mathrm{close}_{{-1}}, \\log(\\mathrm{volume}/\\mathrm{volume}_{{-1}} + 1), 30)."""
from __future__ import annotations
import numpy as np
@@ -15,7 +15,7 @@ from src.factors.base import safe_div, ts_corr
__alpha_meta__ = {
'id': 'qlib158_cord30',
'theme': ['volume', 'microstructure'],
- 'formula_latex': '\\\\mathrm{ts\\\\_corr}(\\\\mathrm{close}/\\\\mathrm{close}_{{-1}}, \\\\log((\\\\mathrm{volume}+1)/(\\\\mathrm{volume}_{{-1}}+1)), 30)',
+ 'formula_latex': '\\\\mathrm{ts\\\\_corr}(\\\\mathrm{close}/\\\\mathrm{close}_{{-1}}, \\\\log(\\\\mathrm{volume}/\\\\mathrm{volume}_{{-1}} + 1), 30)',
'columns_required': ['close', 'volume'],
'universe': ['equity_us', 'equity_cn', 'equity_hk', 'equity_in', 'equity_kr'],
'frequency': ['1d'],
@@ -29,6 +29,6 @@ def compute(panel: dict[str, pd.DataFrame]) -> pd.DataFrame:
c = panel['close']
v = panel['volume']
c_ret = safe_div(c, c.shift(1))
- v_ret = safe_div(v + 1.0, v.shift(1) + 1.0)
+ v_ret = safe_div(v, v.shift(1)) + 1.0
logvr = np.log(v_ret)
return ts_corr(c_ret, logvr, 30)
diff --git a/agent/src/factors/zoo/qlib158/cord5.py b/agent/src/factors/zoo/qlib158/cord5.py
index f7a47cfb1..a05a3fe80 100644
--- a/agent/src/factors/zoo/qlib158/cord5.py
+++ b/agent/src/factors/zoo/qlib158/cord5.py
@@ -2,10 +2,10 @@
# (Apache-2.0). Copyright (c) Microsoft Corporation.
# ============================================================
# 中文名称: 收益率-量变化相关性 5日
-# 简要说明: ts_corr(close/close_prev, log(v/v_prev), 5),5日收益率与成交量变化率的相关系数。
+# 简要说明: ts_corr(close/close_prev, log(v/v_prev + 1), 5),5日价格变化与对数量变比的相关系数。
# 典型用途: 衡量5日价格变动与成交量变动的相关性,反映量价配合程度。
# ============================================================
-"""qlib158 CORD5: formula = \\mathrm{ts\\_corr}(\\mathrm{close}/\\mathrm{close}_{{-1}}, \\log((\\mathrm{volume}+1)/(\\mathrm{volume}_{{-1}}+1)), 5)."""
+"""qlib158 CORD5: formula = \\mathrm{ts\\_corr}(\\mathrm{close}/\\mathrm{close}_{{-1}}, \\log(\\mathrm{volume}/\\mathrm{volume}_{{-1}} + 1), 5)."""
from __future__ import annotations
import numpy as np
@@ -15,7 +15,7 @@ from src.factors.base import safe_div, ts_corr
__alpha_meta__ = {
'id': 'qlib158_cord5',
'theme': ['volume', 'microstructure'],
- 'formula_latex': '\\\\mathrm{ts\\\\_corr}(\\\\mathrm{close}/\\\\mathrm{close}_{{-1}}, \\\\log((\\\\mathrm{volume}+1)/(\\\\mathrm{volume}_{{-1}}+1)), 5)',
+ 'formula_latex': '\\\\mathrm{ts\\\\_corr}(\\\\mathrm{close}/\\\\mathrm{close}_{{-1}}, \\\\log(\\\\mathrm{volume}/\\\\mathrm{volume}_{{-1}} + 1), 5)',
'columns_required': ['close', 'volume'],
'universe': ['equity_us', 'equity_cn', 'equity_hk', 'equity_in', 'equity_kr'],
'frequency': ['1d'],
@@ -29,6 +29,6 @@ def compute(panel: dict[str, pd.DataFrame]) -> pd.DataFrame:
c = panel['close']
v = panel['volume']
c_ret = safe_div(c, c.shift(1))
- v_ret = safe_div(v + 1.0, v.shift(1) + 1.0)
+ v_ret = safe_div(v, v.shift(1)) + 1.0
logvr = np.log(v_ret)
return ts_corr(c_ret, logvr, 5)
diff --git a/agent/src/factors/zoo/qlib158/cord60.py b/agent/src/factors/zoo/qlib158/cord60.py
index 70f0f048d..57aee0086 100644
--- a/agent/src/factors/zoo/qlib158/cord60.py
+++ b/agent/src/factors/zoo/qlib158/cord60.py
@@ -2,10 +2,10 @@
# (Apache-2.0). Copyright (c) Microsoft Corporation.
# ============================================================
# 中文名称: 收益率-量变化相关性 60日
-# 简要说明: ts_corr(close/close_prev, log(v/v_prev), 60),60日收益率与成交量变化率的相关系数。
+# 简要说明: ts_corr(close/close_prev, log(v/v_prev + 1), 60),60日价格变化与对数量变比的相关系数。
# 典型用途: 衡量60日价格变动与成交量变动的相关性,反映量价配合程度。
# ============================================================
-"""qlib158 CORD60: formula = \\mathrm{ts\\_corr}(\\mathrm{close}/\\mathrm{close}_{{-1}}, \\log((\\mathrm{volume}+1)/(\\mathrm{volume}_{{-1}}+1)), 60)."""
+"""qlib158 CORD60: formula = \\mathrm{ts\\_corr}(\\mathrm{close}/\\mathrm{close}_{{-1}}, \\log(\\mathrm{volume}/\\mathrm{volume}_{{-1}} + 1), 60)."""
from __future__ import annotations
import numpy as np
@@ -15,7 +15,7 @@ from src.factors.base import safe_div, ts_corr
__alpha_meta__ = {
'id': 'qlib158_cord60',
'theme': ['volume', 'microstructure'],
- 'formula_latex': '\\\\mathrm{ts\\\\_corr}(\\\\mathrm{close}/\\\\mathrm{close}_{{-1}}, \\\\log((\\\\mathrm{volume}+1)/(\\\\mathrm{volume}_{{-1}}+1)), 60)',
+ 'formula_latex': '\\\\mathrm{ts\\\\_corr}(\\\\mathrm{close}/\\\\mathrm{close}_{{-1}}, \\\\log(\\\\mathrm{volume}/\\\\mathrm{volume}_{{-1}} + 1), 60)',
'columns_required': ['close', 'volume'],
'universe': ['equity_us', 'equity_cn', 'equity_hk', 'equity_in', 'equity_kr'],
'frequency': ['1d'],
@@ -29,6 +29,6 @@ def compute(panel: dict[str, pd.DataFrame]) -> pd.DataFrame:
c = panel['close']
v = panel['volume']
c_ret = safe_div(c, c.shift(1))
- v_ret = safe_div(v + 1.0, v.shift(1) + 1.0)
+ v_ret = safe_div(v, v.shift(1)) + 1.0
logvr = np.log(v_ret)
return ts_corr(c_ret, logvr, 60)
diff --git a/agent/src/factors/zoo/qlib158/roc10.py b/agent/src/factors/zoo/qlib158/roc10.py
index 1747c6976..6821f24f9 100644
--- a/agent/src/factors/zoo/qlib158/roc10.py
+++ b/agent/src/factors/zoo/qlib158/roc10.py
@@ -2,10 +2,10 @@
# (Apache-2.0). Copyright (c) Microsoft Corporation.
# ============================================================
# 中文名称: 变动率 10日
-# 简要说明: close_t / close_{t-10} - 1,10日收益率。
-# 典型用途: 经典的10日动量因子,正值为上涨趋势,负值为下跌趋势。
+# 简要说明: close_{t-10} / close_t,10日前收盘价除以最新收盘价。
+# 典型用途: 用于与 Qlib Alpha158 对齐的10日价格变化分析,不是常规收益率。
# ============================================================
-"""qlib158 ROC10: formula = \\mathrm{close}_t / \\mathrm{close}_{{t-10}} - 1."""
+"""qlib158 ROC10: formula = \\mathrm{close}_{{t-10}} / \\mathrm{close}_t."""
from __future__ import annotations
import pandas as pd
@@ -14,7 +14,7 @@ from src.factors.base import safe_div
__alpha_meta__ = {
'id': 'qlib158_roc10',
'theme': ['momentum'],
- 'formula_latex': '\\\\mathrm{close}_t / \\\\mathrm{close}_{{t-10}} - 1',
+ 'formula_latex': '\\\\mathrm{close}_{{t-10}} / \\\\mathrm{close}_t',
'columns_required': ['close'],
'universe': ['equity_us', 'equity_cn', 'equity_hk', 'equity_in', 'equity_kr'],
'frequency': ['1d'],
@@ -26,4 +26,4 @@ __alpha_meta__ = {
def compute(panel: dict[str, pd.DataFrame]) -> pd.DataFrame:
"""Return qlib158 ROC10 on the supplied OHLCV panel."""
c = panel['close']
- return safe_div(c, c.shift(10)) - 1.0
+ return safe_div(c.shift(10), c)
diff --git a/agent/src/factors/zoo/qlib158/roc20.py b/agent/src/factors/zoo/qlib158/roc20.py
index 1fe7bab8f..8a83d664a 100644
--- a/agent/src/factors/zoo/qlib158/roc20.py
+++ b/agent/src/factors/zoo/qlib158/roc20.py
@@ -2,10 +2,10 @@
# (Apache-2.0). Copyright (c) Microsoft Corporation.
# ============================================================
# 中文名称: 变动率 20日
-# 简要说明: close_t / close_{t-20} - 1,20日收益率。
-# 典型用途: 经典的20日动量因子,正值为上涨趋势,负值为下跌趋势。
+# 简要说明: close_{t-20} / close_t,20日前收盘价除以最新收盘价。
+# 典型用途: 用于与 Qlib Alpha158 对齐的20日价格变化分析,不是常规收益率。
# ============================================================
-"""qlib158 ROC20: formula = \\mathrm{close}_t / \\mathrm{close}_{{t-20}} - 1."""
+"""qlib158 ROC20: formula = \\mathrm{close}_{{t-20}} / \\mathrm{close}_t."""
from __future__ import annotations
import pandas as pd
@@ -14,7 +14,7 @@ from src.factors.base import safe_div
__alpha_meta__ = {
'id': 'qlib158_roc20',
'theme': ['momentum'],
- 'formula_latex': '\\\\mathrm{close}_t / \\\\mathrm{close}_{{t-20}} - 1',
+ 'formula_latex': '\\\\mathrm{close}_{{t-20}} / \\\\mathrm{close}_t',
'columns_required': ['close'],
'universe': ['equity_us', 'equity_cn', 'equity_hk', 'equity_in', 'equity_kr'],
'frequency': ['1d'],
@@ -26,4 +26,4 @@ __alpha_meta__ = {
def compute(panel: dict[str, pd.DataFrame]) -> pd.DataFrame:
"""Return qlib158 ROC20 on the supplied OHLCV panel."""
c = panel['close']
- return safe_div(c, c.shift(20)) - 1.0
+ return safe_div(c.shift(20), c)
diff --git a/agent/src/factors/zoo/qlib158/roc30.py b/agent/src/factors/zoo/qlib158/roc30.py
index 9906ca8dd..67a5054ea 100644
--- a/agent/src/factors/zoo/qlib158/roc30.py
+++ b/agent/src/factors/zoo/qlib158/roc30.py
@@ -2,10 +2,10 @@
# (Apache-2.0). Copyright (c) Microsoft Corporation.
# ============================================================
# 中文名称: 变动率 30日
-# 简要说明: close_t / close_{t-30} - 1,30日收益率。
-# 典型用途: 经典的30日动量因子,正值为上涨趋势,负值为下跌趋势。
+# 简要说明: close_{t-30} / close_t,30日前收盘价除以最新收盘价。
+# 典型用途: 用于与 Qlib Alpha158 对齐的30日价格变化分析,不是常规收益率。
# ============================================================
-"""qlib158 ROC30: formula = \\mathrm{close}_t / \\mathrm{close}_{{t-30}} - 1."""
+"""qlib158 ROC30: formula = \\mathrm{close}_{{t-30}} / \\mathrm{close}_t."""
from __future__ import annotations
import pandas as pd
@@ -14,7 +14,7 @@ from src.factors.base import safe_div
__alpha_meta__ = {
'id': 'qlib158_roc30',
'theme': ['momentum'],
- 'formula_latex': '\\\\mathrm{close}_t / \\\\mathrm{close}_{{t-30}} - 1',
+ 'formula_latex': '\\\\mathrm{close}_{{t-30}} / \\\\mathrm{close}_t',
'columns_required': ['close'],
'universe': ['equity_us', 'equity_cn', 'equity_hk', 'equity_in', 'equity_kr'],
'frequency': ['1d'],
@@ -26,4 +26,4 @@ __alpha_meta__ = {
def compute(panel: dict[str, pd.DataFrame]) -> pd.DataFrame:
"""Return qlib158 ROC30 on the supplied OHLCV panel."""
c = panel['close']
- return safe_div(c, c.shift(30)) - 1.0
+ return safe_div(c.shift(30), c)
diff --git a/agent/src/factors/zoo/qlib158/roc5.py b/agent/src/factors/zoo/qlib158/roc5.py
index 4b429c653..ceee0cacf 100644
--- a/agent/src/factors/zoo/qlib158/roc5.py
+++ b/agent/src/factors/zoo/qlib158/roc5.py
@@ -2,10 +2,10 @@
# (Apache-2.0). Copyright (c) Microsoft Corporation.
# ============================================================
# 中文名称: 变动率 5日
-# 简要说明: close_t / close_{t-5} - 1,5日收益率。
-# 典型用途: 经典的5日动量因子,正值为上涨趋势,负值为下跌趋势。
+# 简要说明: close_{t-5} / close_t,5日前收盘价除以最新收盘价。
+# 典型用途: 用于与 Qlib Alpha158 对齐的5日价格变化分析,不是常规收益率。
# ============================================================
-"""qlib158 ROC5: formula = \\mathrm{close}_t / \\mathrm{close}_{{t-5}} - 1."""
+"""qlib158 ROC5: formula = \\mathrm{close}_{{t-5}} / \\mathrm{close}_t."""
from __future__ import annotations
import pandas as pd
@@ -14,7 +14,7 @@ from src.factors.base import safe_div
__alpha_meta__ = {
'id': 'qlib158_roc5',
'theme': ['momentum'],
- 'formula_latex': '\\\\mathrm{close}_t / \\\\mathrm{close}_{{t-5}} - 1',
+ 'formula_latex': '\\\\mathrm{close}_{{t-5}} / \\\\mathrm{close}_t',
'columns_required': ['close'],
'universe': ['equity_us', 'equity_cn', 'equity_hk', 'equity_in', 'equity_kr'],
'frequency': ['1d'],
@@ -26,4 +26,4 @@ __alpha_meta__ = {
def compute(panel: dict[str, pd.DataFrame]) -> pd.DataFrame:
"""Return qlib158 ROC5 on the supplied OHLCV panel."""
c = panel['close']
- return safe_div(c, c.shift(5)) - 1.0
+ return safe_div(c.shift(5), c)
diff --git a/agent/src/factors/zoo/qlib158/roc60.py b/agent/src/factors/zoo/qlib158/roc60.py
index 6df7f754f..1df1b9ed0 100644
--- a/agent/src/factors/zoo/qlib158/roc60.py
+++ b/agent/src/factors/zoo/qlib158/roc60.py
@@ -2,10 +2,10 @@
# (Apache-2.0). Copyright (c) Microsoft Corporation.
# ============================================================
# 中文名称: 变动率 60日
-# 简要说明: close_t / close_{t-60} - 1,60日收益率。
-# 典型用途: 经典的60日动量因子,正值为上涨趋势,负值为下跌趋势。
+# 简要说明: close_{t-60} / close_t,60日前收盘价除以最新收盘价。
+# 典型用途: 用于与 Qlib Alpha158 对齐的60日价格变化分析,不是常规收益率。
# ============================================================
-"""qlib158 ROC60: formula = \\mathrm{close}_t / \\mathrm{close}_{{t-60}} - 1."""
+"""qlib158 ROC60: formula = \\mathrm{close}_{{t-60}} / \\mathrm{close}_t."""
from __future__ import annotations
import pandas as pd
@@ -14,7 +14,7 @@ from src.factors.base import safe_div
__alpha_meta__ = {
'id': 'qlib158_roc60',
'theme': ['momentum'],
- 'formula_latex': '\\\\mathrm{close}_t / \\\\mathrm{close}_{{t-60}} - 1',
+ 'formula_latex': '\\\\mathrm{close}_{{t-60}} / \\\\mathrm{close}_t',
'columns_required': ['close'],
'universe': ['equity_us', 'equity_cn', 'equity_hk', 'equity_in', 'equity_kr'],
'frequency': ['1d'],
@@ -26,4 +26,4 @@ __alpha_meta__ = {
def compute(panel: dict[str, pd.DataFrame]) -> pd.DataFrame:
"""Return qlib158 ROC60 on the supplied OHLCV panel."""
c = panel['close']
- return safe_div(c, c.shift(60)) - 1.0
+ return safe_div(c.shift(60), c)
diff --git a/agent/tests/factors/test_qlib158_regression.py b/agent/tests/factors/test_qlib158_regression.py
index 8fdc4bdba..b409c1ebb 100644
--- a/agent/tests/factors/test_qlib158_regression.py
+++ b/agent/tests/factors/test_qlib158_regression.py
@@ -1,4 +1,4 @@
-"""Qlib's BETA and RESI features share one trailing OLS fit."""
+"""Qlib Alpha158 compatibility regressions use independent formula oracles."""
from __future__ import annotations
@@ -63,3 +63,64 @@ def test_regression_features_require_a_complete_gap_free_window(stem: str) -> No
assert actual.iloc[3:8].isna().all()
assert actual.iloc[8:].notna().all()
+
+
+def _volume_through(window: int) -> pd.DataFrame:
+ steps = np.arange(window + 2, dtype=np.float64)
+ values = 1.0 + ((steps * 17.0 + 3.0) % 31.0)
+ return pd.DataFrame(values, columns=["A"])
+
+
[email protected]("window", _WINDOWS)
+def test_roc_is_delayed_close_over_latest_close(window: int) -> None:
+ close = _close_through(window + 1)
+ module = importlib.import_module(f"src.factors.zoo.qlib158.roc{window}")
+
+ actual = module.compute({"close": close})["A"]
+
+ assert actual.iloc[:window].isna().all()
+ for end in (window, window + 1):
+ expected = close["A"].iloc[end - window] / close["A"].iloc[end]
+ assert actual.iloc[end] == pytest.approx(expected, abs=1e-12)
+
+
[email protected]("window", _WINDOWS)
+def test_cord_uses_log_of_one_plus_volume_ratio(window: int) -> None:
+ close = _close_through(window + 1)
+ volume = _volume_through(window)
+ module = importlib.import_module(f"src.factors.zoo.qlib158.cord{window}")
+
+ actual = module.compute({"close": close, "volume": volume})["A"]
+
+ price_ratio = close["A"] / close["A"].shift(1)
+ volume_change = np.log(volume["A"] / volume["A"].shift(1) + 1.0)
+ assert actual.iloc[:window].isna().all()
+ for end in (window, window + 1):
+ window_slice = slice(end - window + 1, end + 1)
+ expected = np.corrcoef(
+ price_ratio.iloc[window_slice].to_numpy(),
+ volume_change.iloc[window_slice].to_numpy(),
+ )[0, 1]
+ assert actual.iloc[end] == pytest.approx(expected, abs=1e-12)
+
+
+def test_cord_plus_one_placement_can_flip_correlation_sign() -> None:
+ close = pd.DataFrame({"A": [10.0, 11.0, 15.0, 14.0, 20.0, 22.0]})
+ volume = pd.DataFrame({"A": [15.0, 29.0, 1.0, 1.0, 12.0, 28.0]})
+ module = importlib.import_module("src.factors.zoo.qlib158.cord5")
+
+ actual = module.compute({"close": close, "volume": volume})["A"].iloc[5]
+
+ assert actual == pytest.approx(0.34260936518939894, abs=1e-12)
+
+
+def test_cord_requires_a_complete_gap_free_volume_window() -> None:
+ close = _close_through(11)
+ volume = _volume_through(10)
+ volume.iloc[3, 0] = np.nan
+ module = importlib.import_module("src.factors.zoo.qlib158.cord5")
+
+ actual = module.compute({"close": close, "volume": volume})["A"]
+
+ assert actual.iloc[3:9].isna().all()
+ assert actual.iloc[9:].notna().all()