원클릭으로 Manus에서 모든 스킬 실행

시작하기

macro-calendar-modulation

스타2

포크0

업데이트2026년 2월 24일 14:46

CPI/GDP/retail sales date awareness and VIX regime scaling for calendar features

설치

Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.

Manus에서 실행

출처

smith6jt-cop

smith6jt-cop/Skills_Registry

GitHub 저장소 열기 Creator 저장소 보기

다운로드

Manus에서 실행

Macro Calendar Modulation

Experiment Overview

Item	Details
Date	2026-02-24
Goal	Add CPI/GDP/retail sales awareness and VIX regime scaling to calendar features without changing obs_dim
Environment	Python 3.11+, OpenBB FRED provider
Status	Success

Context

Calendar features only knew FOMC + employment first Friday. No VIX level, no CPI/GDP dates. Earnings cause 5-10% average moves; CPI/GDP releases move markets significantly.

Design constraint: NUM_FEATURES must stay at 7 (no obs_dim change). Solution: modulate existing hour_weight value.

Implementation

Macro Event Dates

CPI_RELEASE_DATES = [datetime(2024,1,11), datetime(2024,2,13), ...]  # 2024-2026
GDP_RELEASE_DATES = [datetime(2024,1,25), datetime(2024,2,28), ...]  # 2024-2026
RETAIL_SALES_DATES = [datetime(2024,1,17), datetime(2024,2,15), ...]  # 2024-2026
ALL_MACRO_EVENT_DATES = sorted(set(CPI + GDP + RETAIL))

hour_weight Modulation

CalendarFeatureCalculator(enable_macro_modulation=True)

# In _hour_weight():
base_weight = ... # existing market hours logic (0.1 - 1.0)

if enable_macro_modulation:
    # +20% on CPI/GDP/retail sales days
    if ts_date in self._macro_event_dates:
        base_weight *= 1.20

    # VIX regime scale (cached 4h via OpenBB FRED)
    vix_scale = _get_vix_regime_scale()  # 0.9x to 1.3x
    base_weight *= vix_scale

VIX Regime Scaling

VIX Level	Scale	Rationale
< 15	1.0x	Normal volatility
15-25	0.95x	Slightly reduce weight (elevated but manageable)
25-35	1.1x	Higher volatility = more signal in price moves
> 35	1.3x	Panic = high weight (big moves, high importance)

Caching

Module-level _macro_data_cache dict (not instance-level)
VIX cached for 4 hours via _fetch_vix_level() using OpenBB FRED
_macro_event_dates pre-built as set in __init__ for O(1) lookup

Failed Attempts

Attempt	Why it Failed	Lesson Learned
Adding new feature (earnings_proximity)	Changes obs_dim from 5900 to 6000, breaks all existing models	Modulate existing features instead
Instance-level VIX cache	Multiple CalendarFeatureCalculator instances waste API calls	Module-level cache shared across instances

Final Parameters

MACRO_EVENT_BOOST = 1.20       # +20% on macro release days
VIX_REGIME_SCALES = {
    'normal': 1.0,     # VIX < 15
    'elevated': 0.95,  # VIX 15-25
    'high': 1.1,       # VIX 25-35
    'panic': 1.3,      # VIX > 35
}
VIX_CACHE_TTL = 14400  # 4 hours

Files Modified

File	Change
`alpaca_trading/features/calendar_features.py`	Added date lists, `_fetch_vix_level()`, `_is_macro_event_day()`, `_get_vix_regime_scale()`, modulated `_hour_weight()`
`tests/test_openbb_integration.py`	6 tests for macro modulation

이 저장소의 다른 Skills

같은 저장소

cycle-dir-normalization

smith6jt-cop/Skills_Registry

Normalize long-form CODEX cycle folders to short form before notebooks run. Trigger: cyc001_reg001_*, hard-coded cyc paths breaking, staged CODEX raw data failing in Notebooks 1/2.

2026-04-162

joint-multi-tf-v560

smith6jt-cop/Skills_Registry

v5.6.0 joint multi-TF model: single model per symbol with broadcast 1Hour context replaces dual 15Min/1Hour models. Trigger: (1) replacing weighted-voting model aggregation, (2) adding broadcast features to vectorized env, (3) limited training data + worried about overfitting from doubling obs_dim, (4) backtest builder mismatch with newer feature counts.

2026-04-112

multi-timeframe-training

smith6jt-cop/Skills_Registry

DEPRECATED in v5.6.0 — see joint-multi-tf-v560 skill. Documents the v5.2.0 dual-model approach (train separate 15Min/1Hour models, combine via weighted voting). Still relevant for: (1) loading legacy v5.5.0 dual models, (2) understanding the historical aggregation layer, (3) resampling pattern via origin='start'.

2026-04-112

dashboard-feature-discovery

smith6jt-cop/Skills_Registry

Surface a shipped-but-undocumented CLI feature in user-facing docs. Trigger: user reports a known feature missing from README/readthedocs even though the CLI command exists.

2026-04-082

live-aware-account-routing

smith6jt-cop/Skills_Registry

KINTSUGI Snakefile + CLI changes that route SLURM jobs around accounts saturated by OTHER users on the same QOS pool. Trigger: QOSGrpMemLimit, jobs stuck pending despite available GPU slots in config, noisy neighbor on shared QOS, multi-user investment pool exhaustion, _build_cycle_assignment static-vs-live.

2026-04-082

slurm-concurrent-processing

smith6jt-cop/Skills_Registry

KINTSUGI SLURM batch processing: Maximize throughput using multi-account resource calculation with GPU+CPU pools per account. Trigger: SLURM job submission, batch processing, resource maximization, GPU+CPU concurrent, headless processing, resource pool.

2026-04-082

name	macro-calendar-modulation
description	CPI/GDP/retail sales date awareness and VIX regime scaling for calendar features
author	Claude Code
date	"2026-02-24T00:00:00.000Z"
version	4.4.0

Macro Calendar Modulation

Experiment Overview

Item	Details
Date	2026-02-24
Goal	Add CPI/GDP/retail sales awareness and VIX regime scaling to calendar features without changing obs_dim
Environment	Python 3.11+, OpenBB FRED provider
Status	Success

Context

Calendar features only knew FOMC + employment first Friday. No VIX level, no CPI/GDP dates. Earnings cause 5-10% average moves; CPI/GDP releases move markets significantly.

Design constraint: NUM_FEATURES must stay at 7 (no obs_dim change). Solution: modulate existing hour_weight value.

Implementation

Macro Event Dates

CPI_RELEASE_DATES = [datetime(2024,1,11), datetime(2024,2,13), ...]  # 2024-2026
GDP_RELEASE_DATES = [datetime(2024,1,25), datetime(2024,2,28), ...]  # 2024-2026
RETAIL_SALES_DATES = [datetime(2024,1,17), datetime(2024,2,15), ...]  # 2024-2026
ALL_MACRO_EVENT_DATES = sorted(set(CPI + GDP + RETAIL))

hour_weight Modulation

CalendarFeatureCalculator(enable_macro_modulation=True)

# In _hour_weight():
base_weight = ... # existing market hours logic (0.1 - 1.0)

if enable_macro_modulation:
    # +20% on CPI/GDP/retail sales days
    if ts_date in self._macro_event_dates:
        base_weight *= 1.20

    # VIX regime scale (cached 4h via OpenBB FRED)
    vix_scale = _get_vix_regime_scale()  # 0.9x to 1.3x
    base_weight *= vix_scale

VIX Regime Scaling

VIX Level	Scale	Rationale
< 15	1.0x	Normal volatility
15-25	0.95x	Slightly reduce weight (elevated but manageable)
25-35	1.1x	Higher volatility = more signal in price moves
> 35	1.3x	Panic = high weight (big moves, high importance)

Caching

Module-level _macro_data_cache dict (not instance-level)
VIX cached for 4 hours via _fetch_vix_level() using OpenBB FRED
_macro_event_dates pre-built as set in __init__ for O(1) lookup

Failed Attempts

Attempt	Why it Failed	Lesson Learned
Adding new feature (earnings_proximity)	Changes obs_dim from 5900 to 6000, breaks all existing models	Modulate existing features instead
Instance-level VIX cache	Multiple CalendarFeatureCalculator instances waste API calls	Module-level cache shared across instances

Final Parameters

MACRO_EVENT_BOOST = 1.20       # +20% on macro release days
VIX_REGIME_SCALES = {
    'normal': 1.0,     # VIX < 15
    'elevated': 0.95,  # VIX 15-25
    'high': 1.1,       # VIX 25-35
    'panic': 1.3,      # VIX > 35
}
VIX_CACHE_TTL = 14400  # 4 hours

Files Modified

File	Change
`alpaca_trading/features/calendar_features.py`	Added date lists, `_fetch_vix_level()`, `_is_macro_event_day()`, `_get_vix_regime_scale()`, modulated `_hour_weight()`
`tests/test_openbb_integration.py`	6 tests for macro modulation