zeberg-salomon-rotator

<essential_principles>

<principle name="two_state_model"> **Core of Two-State Switching Model**

The Zeberg–Salomon model simplifies the market into two states:

RISK_ON: Hold equities (SPY), business cycle expansion phase
RISK_OFF: Hold long-term bonds (TLT), business cycle contraction phase

The switching logic is based on the business cycle rule that "leading indicators weaken first, followed by confirmation from coincident indicators." </principle>

<principle name="leading_coincident"> **Leading vs Coincident Indicators**

Type	Function	Typical Components	Lead Time
Leading	Early Warning	Yield curve, new orders, housing permits	6-12 months
Coincident	Confirmation	Employment, industrial production, real income	Synchronous

Synthesis Method:

Transform each series (yoy/mom/diff)
Unify direction (+1/-1)
Rolling z-score standardization
EMA smoothing
Weighted synthesis </principle>

<principle name="iceberg_sinking"> **Iceberg Event vs Sinking Event**

Iceberg Event: LeadingIndex < iceberg_threshold
  → 預警：景氣開始轉弱
  → 搭配「領先指標下降」+ 可選「市場亢奮」濾鏡

Sinking Event: CoincidentIndex < sinking_threshold
  → 確認：實體經濟收縮
  → 通常在 Iceberg 之後數月發生

State Machine Logic:

RISK_ON → RISK_OFF: Iceberg Event confirmed consecutively + negative slope
RISK_OFF → RISK_ON: Leading indicator rebounds above (threshold + hysteresis) </principle>

<principle name="data_access"> **Data Acquisition Methods**

This skill uses data sources that do not require an API key:

FRED CSV:

https://fred.stlouisfed.org/graph/fredgraph.csv?id={SERIES_ID}

Yahoo Finance: Fetch SPY, TLT, VIX using the
```
yfinance
```
package

Scripts are located in the

scripts/

directory and can be executed directly. </principle>

</essential_principles>

<objective> Implement the Zeberg–Salomon two-state rotation strategy:

Construct Indicators: Synthesize LeadingIndex and CoincidentIndex from FRED data
Detect Events: Identify "Iceberg" (leading indicator weakening) and "Sinking" (coincident indicator confirmation) events
Generate Switch Signals: Produce RISK_ON ↔ RISK_OFF switch events
Backtest Performance: Calculate cumulative returns, MaxDD, CAGR, and compare with benchmarks

Outputs: List of switch events, indicator time series, backtest summary, diagnostic information. </objective>

<quick_start>

Fastest Way: Run Default Backtest

bash

cd skills/zeberg-salomon-rotator
pip install pandas numpy yfinance pandas-datareader  # First-time use
python scripts/rotator.py --quick

Sample Output:

json

{
  "state": "RISK_ON",
  "latest_indices": {"LeadingIndex": 0.41, "CoincidentIndex": 0.22},
  "iceberg_event": false,
  "sinking_event": false,
  "last_switch": {"date": "2023-06-30", "action": "EXIT_LONG_BOND_ENTER_EQUITY"}
}

Full Backtest:

bash

python scripts/rotator.py --start 2000-01-01 --end 2026-01-01 --output result.json

</quick_start>

<intake> What operation do you need to perform?

Quick Check - View current business cycle state and latest indicators
Full Backtest - Perform complete historical backtest and performance analysis
Visualization Charts - Generate multi-panel backtest result charts
Monitoring Mode - Set up continuous monitoring and switch alerts
Methodology Learning - Understand the logic of the Zeberg-Salomon model

Please select or provide analysis parameters directly. </intake>

<routing> | Response | Action | |-------------------------------|---------------------------------------------| | 1, "quick", "check" | Execute `python scripts/rotator.py --quick` | | 2, "backtest", "full" | Read `workflows/backtest.md` and execute | | 3, "chart", "plot" | Read `workflows/visualize.md` and execute | | 4, "monitor", "alert" | Read `workflows/monitor.md` and execute | | 5, "learn", "methodology", "why" | Read `references/methodology.md` | | Provide parameters (e.g., date range) | Read `workflows/backtest.md` and execute with parameters |

After routing, read the corresponding document and execute. </routing>

<directory_structure>

zeberg-salomon-rotator/
├── SKILL.md                           # 本文件（路由器）
├── skill.yaml                         # 前端展示元數據
├── manifest.json                      # 技能元數據
├── workflows/
│   ├── backtest.md                    # 完整回測工作流
│   ├── visualize.md                   # 視覺化工作流
│   ├── monitor.md                     # 持續監控工作流
│   └── analyze.md                     # 深度分析工作流
├── references/
│   ├── data-sources.md                # FRED 系列代碼與資料來源
│   ├── methodology.md                 # Zeberg-Salomon 方法論解析
│   └── input-schema.md                # 完整輸入參數定義
├── templates/
│   ├── output-json.md                 # JSON 輸出模板
│   └── output-markdown.md             # Markdown 報告模板
└── scripts/
    ├── rotator.py                     # 主輪動腳本
    ├── visualize.py                   # 視覺化繪圖工具
    └── fetch_data.py                  # 數據抓取工具

</directory_structure>

<reference_index>

Methodology: references/methodology.md

Zeberg-Salomon model concepts
Definitions of Iceberg/Sinking events
Two-state switching logic

Data Sources: references/data-sources.md

FRED series codes (leading/coincident)
Yahoo Finance asset codes
Data frequency and alignment

Input Parameters: references/input-schema.md

Complete parameter definitions
Default values and recommended ranges

</reference_index>

<workflows_index>

Workflow	Purpose	Usage Scenario
backtest.md	Complete historical backtest	When performance analysis is needed
visualize.md	Generate visualization charts	When chart presentation is needed
monitor.md	Continuous state monitoring	For daily monitoring or alerts
analyze.md	In-depth indicator analysis	To understand current market conditions
</workflows_index>

<templates_index>

Template	Purpose
output-json.md	JSON output structure definition
output-markdown.md	Markdown report template
</templates_index>

<scripts_index>

Script	Command	Purpose
rotator.py	`--quick`	Quick check of current state
rotator.py	`--start DATE --end DATE`	Complete backtest
visualize.py	`-i result.json -o chart.png`	Generate visualization charts
fetch_data.py	`--series T10Y3M,PAYEMS`	Fetch FRED data
</scripts_index>

<input_schema_summary>

Core Parameters

Parameter	Type	Default Value	Description
start_date	string	2000-01-01	Backtest start date
end_date	string	today	Backtest end date
freq	string	M	Frequency (M=Monthly)
equity_proxy	string	SPY	Risk asset proxy
bond_proxy	string	TLT	Long-term bond proxy

Threshold Parameters

Parameter	Type	Default Value	Description
iceberg_threshold	number	-0.3	Leading indicator threshold
sinking_threshold	number	-0.5	Coincident indicator threshold
confirm_periods	int	2	Number of consecutive confirmation periods
hysteresis	number	0.15	Entry/exit gap

Complete parameter definitions can be found in

references/input-schema.md

</input_schema_summary>

<output_schema_summary>

json

{
  "skill": "zeberg-salomon-rotator",
  "as_of": "2026-01-14",
  "state": "RISK_ON",
  "latest_indices": {
    "LeadingIndex": 0.41,
    "CoincidentIndex": 0.22,
    "iceberg_event": false,
    "sinking_event": false
  },
  "switch_events": [...],
  "backtest_summary": {
    "cagr": 0.123,
    "max_drawdown": -0.27,
    "turnovers": 10
  }
}

Complete output structure is available in

templates/output-json.md

. </output_schema_summary>

<success_criteria> When execution is successful, the following should be produced:

Current state (RISK_ON or RISK_OFF)
Values of LeadingIndex and CoincidentIndex
Iceberg/Sinking event determination
List of switch events (including date, reason)
Backtest performance summary (CAGR, MaxDD, number of turnovers)
Comparison with benchmarks (buy-and-hold, 60/40)
Diagnostic information (contribution of each indicator)
Visualization charts (optional, output to
```
output/
```
directory) </success_criteria>

zeberg-salomon-rotator

NPX Install

Tags

SKILL.md Content (Chinese)