detect-palladium-lead-silver-turns

<essential_principles>

The hypothesis that "palladium leads silver" requires quantifiable verification:

Use cross-correlation to estimate the optimal lead-lag
When silver shows a turning point, check whether palladium first or simultaneously shows a same-direction turning point within the confirmation window
Unconfirmed turning points are treated as candidates for "failed pushes"

Lead-Lag = argmax(cross_correlation(pd_ret[t-k:t], ag_ret[t:t+k]))
Confirmed = pd_turn exists within [ag_turn.ts - window, ag_turn.ts + window]

Method	Principle	Applicable scenarios
`pivot`	Local extrema within N bars to the left and right	Trends with clear structure
`peaks`	scipy find_peaks + prominence	Automated density control
`slope_change`	Trend slope changes from positive to negative or vice versa	Smoothed trend tracking

It is recommended to start from pivot with 3-5 bars on each side, then adjust as needed.

There are multiple ways to measure whether palladium "participates" in silver's moves:

Indicator	Definition	Suggested threshold
`returns_corr`	Rolling correlation coefficient of returns	> 0.5
`direction_agree`	Proportion of same-direction rises and falls	> 60%
`vol_expansion`	Both volatilities expand in sync	σ_pd / σ_ag > 0.8
`breakout_confirm`	When silver breaks out, palladium also breaks out	Same-direction breakout

When the threshold is not met, silver's move may be "liquidity noise" rather than trend confirmation.

Turn "silver moves without palladium participation" into backtestable rules:

Rule	Definition	Consequence
`no_confirm_then_revert`	No confirmation + silver retraces past the starting point within N bars	Mark as failed_move
`no_confirm_then_break_fail`	No confirmation + after silver breaks out, it falls back below the breakout point	False breakout

Historical statistics: failure rate of unconfirmed events vs. success rate of confirmed events.

</essential_principles>

Data acquisition: OHLCV for silver and palladium (yfinance: SI=F, PA=F)
Turning point detection: identify local highs and lows of both (pivot / peaks / slope_change)
Lead-lag estimation: use cross-correlation to find the optimal lag
Cross-metal confirmation: whether silver's turning point is confirmed by a same-direction palladium turning point within the window
Failed move determination: whether unconfirmed silver turning points meet the failure rules

Output: confirmation rate, failure rate, detailed determination for each event, and risk control recommendations.

<quick_start>

The fastest way: detect whether silver's recent turning points are confirmed by palladium

cd skills/detect-palladium-lead-silver-turns
pip install pandas numpy yfinance scipy statsmodels  # 首次使用
python scripts/palladium_lead_silver.py --silver SI=F --palladium PA=F --quick

Example output:

{
  "symbol_pair": {"silver": "SI=F", "palladium": "PA=F"},
  "as_of": "2026-01-14",
  "timeframe": "1h",
  "estimated_pd_leads_by_bars": 6,
  "lead_lag_corr": 0.42,
  "confirmation_rate": 0.71,
  "unconfirmed_failure_rate": 0.64,
  "latest_event": {
    "ts": "2026-01-15T14:00:00Z",
    "turn": "top",
    "confirmed": false,
    "participation_ok": false,
    "failed_move": true
  }
}

Full analysis：

python scripts/palladium_lead_silver.py --silver SI=F --palladium PA=F --timeframe 1h --lookback 1000 --output result.json

Generate Bloomberg-style visualization charts(recommended):

pip install matplotlib yfinance  # 首次使用
python scripts/plot_bloomberg_style.py --input result.json --output output/palladium_silver_2026-01-26.png

Chart features:

Bloomberg professional color scheme: dark background, orange-red silver line, orange-yellow palladium line
Background color band markers: green background = confirmed turning point area, red background = unconfirmed turning point area (does not block the trend lines)
Latest event annotation: prominently mark the confirmation status and price of the latest turning point
Pd/Ag price ratio chart: shows the relative price change of palladium versus silver, including a 20-period moving average
Rolling confirmation rate: dynamically shows the effectiveness trend of the confirmation logic
Statistics Dashboard: Key metrics such as confirmation rate, failure rate, total number of pivots
Market Interpretation: Current state assessment and actionable recommendations

Traditional Three-in-One Chart(Technical Analysis-oriented):

python scripts/plot_palladium_silver.py --silver SI=F --palladium PA=F --output output/

Includes:

Silver/Palladium price overlay with pivot markers
Confirmed/Unconfirmed event distribution
Rolling correlation coefficient time series
Failed trend statistics

</quick_start>

Quick Detection - Check whether the most recent silver pivot is confirmed by palladium
Historical Backtest - Retrospectively analyze the effectiveness of cross-metal confirmation
Continuous Monitoring - Set alerts to notify when a new pivot appears
Parameter Tuning - Find the optimal confirmation window and participation threshold
Methodology Learning - Understand the theoretical basis of cross-metal lead-lag

Please select or directly provide analysis parameters to begin.

After routing, read the corresponding document and execute.

<directory_structure>

detect-palladium-lead-silver-turns/
├── 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/
│   ├── detect.md                      # 單次偵測工作流
│   ├── backtest.md                    # 歷史回測工作流
│   └── monitor.md                     # 持續監控工作流
├── references/
│   ├── methodology.md                 # 跨金屬領先滯後方法論
│   ├── input-schema.md                # 完整輸入參數定義
│   └── data-sources.md                # 資料來源說明
├── templates/
│   ├── output-json.md                 # JSON 輸出模板
│   └── output-markdown.md             # Markdown 報告模板
├── scripts/
│   ├── palladium_lead_silver.py       # 主偵測腳本
│   ├── plot_bloomberg_style.py        # Bloomberg 風格視覺化（推薦）
│   └── plot_palladium_silver.py       # 傳統三合一圖表
└── examples/
    └── silver-palladium-2024.json     # 範例輸出

</directory_structure>

<reference_index>

Methodology: references/methodology.md

Cross-Metal Lead-Lag Principles
Detailed Explanation of the Three Pivot Detection Methods
Participation and Confirmation Logic
Market Implications of Failed Trends

Data Sources: references/data-sources.md

Yahoo Finance Futures Symbols
Macro Filter Data Sources
Data Frequency and Alignment

Input Parameters: references/input-schema.md

Complete Parameter Definitions
Default Values and Recommended Ranges

</reference_index>

<workflows_index>

Workflow	Purpose	When to Use
detect.md	Single Detection	Check a specific time range
backtest.md	Historical Backtest	Validate the effectiveness of the confirmation logic
monitor.md	Continuous Monitoring	Daily tracking or alerts
</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
palladium_lead_silver.py	`--silver SI=F --palladium PA=F --quick`	Quickly check the current status
palladium_lead_silver.py	`--silver SI=F --palladium PA=F --lookback 1000`	Complete historical analysis
plot_bloomberg_style.py	`--input result.json --output output/chart.png`	Bloomberg-style chart (recommended)
plot_palladium_silver.py	`--silver SI=F --palladium PA=F --output dir/`	Traditional Three-in-One Chart
</scripts_index>

<input_schema_summary>

Core Parameters

Parameter	Type	Default Value	Description
silver_symbol	string	(required)	Silver instrument symbol
palladium_symbol	string	(required)	Palladium instrument symbol
timeframe	string	1h	Analysis timeframe
lookback_bars	int	1000	Number of bars to look back

Pivot Detection Parameters

Parameter	Type	Default Value	Description
turn_method	string	pivot	Pivot detection method
pivot_left	int	3	Number of confirmation bars on the left side of the pivot
pivot_right	int	3	Number of confirmation bars on the right side of the pivot
confirm_window_bars	int	6	Cross-metal confirmation window
lead_lag_max_bars	int	24	Maximum lag in bars for lead-lag

Participation Parameters

Parameter	Type	Default Value	Description
participation_metric	string	direction_agree	Participation measurement method
participation_threshold	float	0.6	Participation threshold
failure_rule	string	no_confirm_then_revert	Failed trend rules

See for complete parameter definitions references/input-schema.md。

</input_schema_summary>

<output_schema_summary>

{
  "skill": "detect-palladium-lead-silver-turns",
  "symbol_pair": {"silver": "SI=F", "palladium": "PA=F"},
  "as_of": "2026-01-14",
  "timeframe": "1h",
  "lookback_bars": 1200,
  "summary": {
    "estimated_pd_leads_by_bars": 6,
    "lead_lag_corr": 0.42,
    "confirmation_rate": 0.71,
    "unconfirmed_failure_rate": 0.64,
    "total_ag_turns": 24,
    "confirmed_turns": 17,
    "failed_moves": 5
  },
  "events": [
    {
      "ts": "2026-01-08T10:00:00Z",
      "turn": "bottom",
      "confirmed": true,
      "confirmation_lag_bars": -3,
      "participation_ok": true,
      "failed_move": false
    }
  ],
  "interpretation": {
    "regime_assessment": "...",
    "tactics": ["...", "..."]
  }
}

See for the complete output structure templates/output-json.md。 </output_schema_summary>

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

Best lead-lag estimate of palladium versus silver (bars)
Correlation coefficient of the lead-lag
Confirmation rate of silver turning points confirmed by palladium
Proportion of failed trends among unconfirmed events
Detailed assessment of each silver turning point (confirmed, participation, failed)
Market interpretation and tactical recommendations
Time series data (optional, for visualization)
Visualization chart PNG (optional, using plot_palladium_silver.py) </success_criteria>

About this skill