detect-palladium-lead-silver-turns

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

<essential_principles>

<principle name="cross_metal_confirmation"> **跨金屬確認核心**

「鈀金領先白銀」的假說需要可量化驗證：

以 cross-correlation 估計最佳領先滯後（lead-lag）
當銀出現拐點時，檢查鈀金是否在確認窗口內先行或同步出現同向拐點
未被確認的拐點視為「失敗推動」的候選

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]

</principle> <principle name="turning_point_detection"> **拐點偵測三法**

方法	原理	適用場景
`pivot`	左右 N 根K棒內的局部極值	結構明確的趨勢
`peaks`	scipy find_peaks + prominence	自動化密度控制
`slope_change`	趨勢斜率由正轉負或反之	平滑趨勢追蹤

建議從

pivot

開始，左右各 3-5 根K棒，再依需求調整。 </principle>

<principle name="participation_judgment"> **參與度判定**

鈀金是否「參與」銀的走勢，有多種衡量方式：

指標	定義	門檻建議
`returns_corr`	報酬率滾動相關係數	> 0.5
`direction_agree`	同向漲跌的比例	> 60%
`vol_expansion`	兩者波動同步擴張	σ_pd / σ_ag > 0.8
`breakout_confirm`	銀突破時鈀金也突破	同向突破

未達門檻時，銀的動作可能是「流動性噪音」而非趨勢確認。 </principle>

<principle name="failure_move_detection"> **失敗走勢判定**

將「無鈀金參與的銀動作」落地為可回測的規則：

規則	定義	後果
`no_confirm_then_revert`	無確認 + 銀在 N 根K內回撤過起點	標記為 failed_move
`no_confirm_then_break_fail`	無確認 + 銀突破後回落跌破突破點	假突破

歷史統計：未確認事件的失敗率 vs 已確認事件的成功率。 </principle>

</essential_principles>

<objective> 偵測「鈀金領先、白銀跟隨」的跨金屬拐點：

數據取得：白銀與鈀金的 OHLCV（yfinance: SI=F, PA=F）
拐點偵測：識別兩者的局部高低點（pivot / peaks / slope_change）
領先滯後估計：cross-correlation 找最佳 lag
跨金屬確認：銀的拐點是否在窗口內被鈀金同向拐點確認
失敗走勢判定：未確認的銀拐點是否符合失敗規則

輸出：確認率、失敗率、每個事件的詳細判定、風控建議。 </objective>

<quick_start>

最快的方式：偵測白銀近期拐點是否被鈀金確認

bash

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

輸出範例：

json

{
  "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
  }
}

完整分析：

bash

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

生成 Bloomberg 風格視覺化圖表（推薦）：

bash

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

圖表特色：

Bloomberg 專業配色：深色背景、橙紅色白銀線、橙黃色鈀金線
背景色帶標記：綠色背景 = 已確認拐點區域，紅色背景 = 未確認拐點區域（不擋住走勢線）
最新事件標註：醒目標示最新拐點的確認狀態與價格
Pd/Ag 價格比率圖：顯示鈀金對白銀的相對價格變化，含 20 期均線
滾動確認率：動態顯示確認邏輯的有效性趨勢
統計面板：確認率、失敗率、總拐點數等關鍵指標
行情解讀：當前狀態評估與可操作建議

傳統三合一圖表（技術分析向）：

bash

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

包含：

銀/鈀價格疊加與拐點標記
確認/未確認事件分布
滾動相關係數時間序列
失敗走勢統計

</quick_start>

<intake> 需要進行什麼操作？

快速偵測 - 檢查最近白銀拐點是否被鈀金確認
歷史回測 - 回溯分析跨金屬確認的有效性
持續監控 - 設定警報當出現新拐點時通知
參數調校 - 找出最佳的確認窗口與參與度門檻
方法論學習 - 了解跨金屬領先滯後的理論基礎

請選擇或直接提供分析參數開始。 </intake>

<routing> | Response | Action | |------------------------------------|--------------------------------------------------------| | 1, "快速", "quick", "check" | 執行 `python scripts/palladium_lead_silver.py --quick` | | 2, "回測", "backtest", "history" | 閱讀 `workflows/backtest.md` 並執行 | | 3, "監控", "monitor", "alert" | 閱讀 `workflows/monitor.md` 並執行 | | 4, "調校", "optimize", "tune" | 閱讀 `workflows/detect.md` 的參數調校部分 | | 5, "學習", "方法論", "why" | 閱讀 `references/methodology.md` | | 提供參數（如 timeframe, lookback） | 閱讀 `workflows/detect.md` 並使用參數執行 |

路由後，閱讀對應文件並執行。 </routing>

<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>

方法論: references/methodology.md

跨金屬領先滯後原理
拐點偵測三法詳解
參與度與確認邏輯
失敗走勢的市場含義

資料來源: references/data-sources.md

Yahoo Finance 期貨代碼
宏觀濾鏡數據來源
數據頻率與對齊

輸入參數: references/input-schema.md

完整參數定義
預設值與建議範圍

</reference_index>

<workflows_index>

Workflow	Purpose	使用時機
detect.md	單次偵測	檢查特定時間範圍
backtest.md	歷史回測	驗證確認邏輯有效性
monitor.md	持續監控	日常追蹤或警報
</workflows_index>

<templates_index>

Template	Purpose
output-json.md	JSON 輸出結構定義
output-markdown.md	Markdown 報告模板
</templates_index>

<scripts_index>

Script	Command	Purpose
palladium_lead_silver.py	`--silver SI=F --palladium PA=F --quick`	快速檢查當前狀態
palladium_lead_silver.py	`--silver SI=F --palladium PA=F --lookback 1000`	完整歷史分析
plot_bloomberg_style.py	`--input result.json --output output/chart.png`	Bloomberg 風格圖表（推薦）
plot_palladium_silver.py	`--silver SI=F --palladium PA=F --output dir/`	傳統三合一圖表
</scripts_index>

<input_schema_summary>

核心參數

參數	類型	預設值	說明
silver_symbol	string	(required)	白銀標的代碼
palladium_symbol	string	(required)	鈀金標的代碼
timeframe	string	1h	分析時間尺度
lookback_bars	int	1000	回溯K棒數

拐點偵測參數

參數	類型	預設值	說明
turn_method	string	pivot	拐點偵測方法
pivot_left	int	3	pivot 左側確認K數
pivot_right	int	3	pivot 右側確認K數
confirm_window_bars	int	6	跨金屬確認窗口
lead_lag_max_bars	int	24	領先滯後最大滯後K數

參與度參數

參數	類型	預設值	說明
participation_metric	string	direction_agree	參與度衡量方式
participation_threshold	float	0.6	參與度門檻
failure_rule	string	no_confirm_then_revert	失敗走勢規則

完整參數定義見

references/input-schema.md

。

</input_schema_summary>

<output_schema_summary>

json

{
  "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": ["...", "..."]
  }
}

完整輸出結構見

templates/output-json.md

。 </output_schema_summary>

<success_criteria> 執行成功時應產出：

鈀金對白銀的最佳領先滯後估計（bars）
領先滯後的相關係數
白銀拐點被鈀金確認的確認率
未確認事件的失敗走勢比例
每個白銀拐點的詳細判定（confirmed, participation, failed）
行情解讀與戰術建議
時間序列資料（可選，用於視覺化）
視覺化圖表 PNG（可選，使用 plot_palladium_silver.py） </success_criteria>

<essential_principles>

<principle name="cross_metal_confirmation"> **Core of Cross-Metal Confirmation**

The hypothesis of "Palladium leads Silver" requires quantifiable verification:

Estimate the optimal lead-lag using cross-correlation
When a turning point appears in silver, check if palladium has a同向 turning point leading or synchronously within the confirmation window
Unconfirmed turning points are candidates for "failed moves"

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]

</principle> <principle name="turning_point_detection"> **Three Methods for Turning Point Detection**

Method	Principle	Applicable Scenario
`pivot`	Local extremum within N candlesticks on both sides	Trends with clear structure
`peaks`	scipy find_peaks + prominence	Automated density control
`slope_change`	Trend slope reverses from positive to negative or vice versa	Smooth trend tracking

It is recommended to start with

pivot

, using 3-5 candlesticks on each side, then adjust as needed. </principle>

<principle name="participation_judgment"> **Participation Judgment**

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

Indicator	Definition	Recommended Threshold
`returns_corr`	Rolling correlation coefficient of returns	> 0.5
`direction_agree`	Proportion of same-direction price movements	> 60%
`vol_expansion`	Synchronized volatility expansion of both	σ_pd / σ_ag > 0.8
`breakout_confirm`	Palladium also breaks out when silver breaks out	Same-direction breakout

When thresholds are not met, silver's movements may be "liquidity noise" rather than trend confirmation. </principle>

<principle name="failure_move_detection"> **Failed Trend Judgment**

Translate "silver movements without palladium participation" into backtestable rules:

Rule	Definition	Consequence
`no_confirm_then_revert`	No confirmation + silver retraces beyond the starting point within N candlesticks	Marked as failed_move
`no_confirm_then_break_fail`	No confirmation + silver breaks out then falls back below the breakout level	False breakout

Historical statistics: Failure rate of unconfirmed events vs success rate of confirmed events. </principle>

</essential_principles>

<objective> Detect cross-metal turning points where "Palladium leads, Silver follows":

Data Acquisition: OHLCV data of 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: Find the optimal lag using cross-correlation
Cross-Metal Confirmation: Check if silver's turning point is confirmed by a same-direction turning point of palladium within the window
Failed Trend Judgment: Check if unconfirmed silver turning points meet the failure rules

Output: Confirmation rate, failure rate, detailed judgment of each event, risk control suggestions. </objective>

<quick_start>

Quickest Way: Detect if recent silver turning points are confirmed by palladium

bash

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

Output Example:

json

{
  "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:

bash

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

Generate Bloomberg-Style Visualization Chart (Recommended):

bash

pip install matplotlib yfinance  # First-time use
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 Band Marking: Green background = Confirmed turning point area, Red background = Unconfirmed turning point area (does not block price lines)
Latest Event Annotation: Prominently marks the confirmation status and price of the latest turning point
Pd/Ag Price Ratio Chart: Shows relative price changes of palladium to silver, including 20-period moving average
Rolling Confirmation Rate: Dynamically displays the validity trend of the confirmation logic
Statistics Panel: Key indicators such as confirmation rate, failure rate, total number of turning points
Market Interpretation: Current status assessment and actionable suggestions

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

bash

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

Includes:

Silver/Palladium price overlay with turning point markings
Distribution of confirmed/unconfirmed events
Time series of rolling correlation coefficient
Failed trend statistics

</quick_start>

<intake> What operation do you need to perform?

Quick Detection - Check if the latest silver turning point is confirmed by palladium
Historical Backtest - Backtest the effectiveness of cross-metal confirmation
Continuous Monitoring - Set up alerts to notify when new turning points appear
Parameter Tuning - Find the optimal confirmation window and participation thresholds
Methodology Learning - Understand the theoretical basis of cross-metal lead-lag

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

<routing> | Response | Action | |------------------------------------|--------------------------------------------------------| | 1, "quick", "fast", "check" | Execute `python scripts/palladium_lead_silver.py --quick` | | 2, "backtest", "history", "historical" | Read `workflows/backtest.md` and execute | | 3, "monitor", "alert", "watch" | Read `workflows/monitor.md` and execute | | 4, "optimize", "tune", "adjust" | Read the parameter tuning section in `workflows/detect.md` | | 5, "learn", "methodology", "theory" | Read `references/methodology.md` | | Provide parameters (e.g., timeframe, lookback) | Read `workflows/detect.md` and execute with parameters |

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

<directory_structure>

detect-palladium-lead-silver-turns/
├── SKILL.md                           # This file (router)
├── skill.yaml                         # Frontend display metadata
├── manifest.json                      # Skill metadata
├── workflows/
│   ├── detect.md                      # One-time detection workflow
│   ├── backtest.md                    # Historical backtest workflow
│   └── monitor.md                     # Continuous monitoring workflow
├── references/
│   ├── methodology.md                 # Cross-metal lead-lag methodology
│   ├── input-schema.md                # Complete input parameter definition
│   └── data-sources.md                # Data source description
├── templates/
│   ├── output-json.md                 # JSON output structure definition
│   └── output-markdown.md             # Markdown report template
├── scripts/
│   ├── palladium_lead_silver.py       # Main detection script
│   ├── plot_bloomberg_style.py        # Bloomberg-style visualization (Recommended)
│   └── plot_palladium_silver.py       # Traditional three-in-one chart
└── examples/
    └── silver-palladium-2024.json     # Example output

</directory_structure>

<reference_index>

Methodology: references/methodology.md

Cross-metal lead-lag principles
Detailed explanation of three turning point detection methods
Participation and confirmation logic
Market implications of failed trends

Data Sources: references/data-sources.md

Yahoo Finance futures codes
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	Applicable Scenario
detect.md	One-time detection	Check specific time ranges
backtest.md	Historical backtest	Verify the validity of 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`	Quick check of 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 target code
palladium_symbol	string	(required)	Palladium target code
timeframe	string	1h	Analysis time scale
lookback_bars	int	1000	Number of historical candlesticks to look back

Turning Point Detection Parameters

Parameter	Type	Default Value	Description
turn_method	string	pivot	Turning point detection method
pivot_left	int	3	Number of confirmation candlesticks on the left for pivot method
pivot_right	int	3	Number of confirmation candlesticks on the right for pivot method
confirm_window_bars	int	6	Cross-metal confirmation window
lead_lag_max_bars	int	24	Maximum number of lag candlesticks for lead-lag estimation

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 rule

Complete parameter definition can be found in

references/input-schema.md

</input_schema_summary>

<output_schema_summary>

json

{
  "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": ["...", "..."]
  }
}

Complete output structure can be found in

templates/output-json.md

. </output_schema_summary>

<success_criteria> Successful execution should produce:

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