Using LLM For Stock Analysis

The advent of AI, or rather LLM (Large Language Models) to be more specific, has made life for many easier. From getting work done to doing searches, LLM like Gemini, ChatGPT and Claude are heavyweights and for many, have become part of our daily and work lives. Today, we will talk about using LLMs to run research on financial decisions, ie stock picking and trading.

Many people run their LLMs using simple instructions like: run an analysis on the financial reports i have attached and tell me what is a reasonable price of the stock. This kind of instruction is vague and the results may not be optimal. The more specific your instruction is, the better your report will be. Today, we will see how we can churn out a detailed prompt for the LLM to work on.

Example used: China Hainan Rubber Industry Group (Stock Code: 601118)
Task summary: I bought 5900 Hainan Rubber shares at 4.63 RMB per share. I want to know when to sell.

Here are the steps to consider:

1. I used Deepseek (China’s LLM) and Gemini (US’ LLM). Key in what you want and ask them to generate a prompt. Take note their job is not to analyse the stock but to generate a prompt to put into different LLMs.
2. Your initial instructions should also include basic knowledge about the stock. For example, in the case of rubber, the prompt should also include an instruction to analyse the financial reports (which you will download and attach), news and economic cycle, production cycle of rubber etc. It should also include an instruction to include any critical information which may be useful but which you forgot to mention.
3. It should also include your main objective: At what price should you enter or to sell, your budget and your plan. If you have a plan, to critically assess the plan and propose a new plan.
4. Run the same instructions on both Deepseek and Gemini. Both are free with generous limits so you can run many iterations.
5. Copy each LLM’s output to the other LLM, asked it to analyse the other LLM’s output, to critically analyse it, combine the best of both practices. Do this 2 or 3 times, then use only one but alternate between the LLMs for each iteration, asking them to improve further. For every iteration, read carefully what the changes are. You may need to manually edit an instruction (do inform the LLM if you do and why). This will run for another few iterations.
6. Then use the prompt to run on Claude and ChatGPT, which has more usage limits but will give you a better report. Do note that every LLM has their own strengths and weaknesses. Gemini is suitable for search, Deepseek is suitable for logical thinking and Mathematics, but neither of them match up to Claude or ChatGPT.
7. Read both ChatGPT and Claude’s report.

This method will help you crunch through financial reports quickly, get new ideas and perspective which you may not have thought of it yourself and also new suggestions which you may consider.

But most importantly, do note that LLM makes mistakes and it is still a tool. Do not trust the LLM 100%. The LLM may give you a proposal, but it is the user who must look through it, verify if a certain fact or claim in the report is valid and finally decide what to do with it.

A story about AI: A man saw some mushrooms on the ground and asked the AI if he could eat the mushrooms. The AI said yes, and the man ate it. He got poisoned in the end and questioned the AI. The AI replied: Yes you are right. I am sorry I made a mistake. The mushroom is poisoned and cannot be eaten.

The LLM is a great tool, but it is ultimately just a tool.

A sample prompt which I used have been attached for reference.

# PROMPT FOR LLM ANALYSIS: HAINAN RUBBER (601118.SH)

## ROLE & PERSPECTIVE
Act as a seasoned agricultural commodities equity analyst and cyclical value investor. Combine deep fundamental rigor with specialized knowledge of global natural rubber (NR) and synthetic rubber (SR) markets, plantation economics, processing margins, biological asset accounting (IAS 41 / CAS 5), and Chinese state-owned enterprise (SOE) dynamics. You are analytical, objective, skepticism-driven, and focused on risk-adjusted medium-to-long-term capital deployment. Prioritize cash flow reality, plantation lifecycle dynamics (immature vs mature acreage), commodity price cycles, inventory dynamics, and balance sheet resilience over accounting top-line noise.

—

## USER TRADING CONTEXT & STRATEGY ASSUMPTIONS
– Target Asset: China Hainan Rubber Industry Group Co., Ltd. (A-share: 601118.SH).
– Current Position: 5,900 shares acquired at an average cost of 4.63 RMB (total baseline investment ~27,317 RMB).
– Order Execution Unit: Fixed tranche size of ~25,000 RMB per buy order. Convert to actual share counts using Shanghai Stock Exchange round-lot rules (exact multiples of 100 shares closest to ~25,000 RMB based on the current market price cited); do NOT assume a stale price or fractional/odd lots.
– Accumulation Strategy Mandate:
* The user previously used a rigid “buy one tranche for every 5% drop from 4.63 RMB” rule.
* You are NOT bound to this rule. You MUST critically assess it (highlighting shortcomings such as ignoring technical support levels, seasonal rubber cycles, and volatility, which risk premature capital exhaustion).
* You MUST propose and justify a superior, fundamentally and cyclically anchored accumulation framework (e.g., P/B valuation floors, SHFE/SICOM cost-curve supports, or seasonal Q1 wintering lulls). State clearly what replaces the 5% rule and why in a dedicated summary paragraph.
– Holding Period Tolerance: Weeks to 12 months (up to 1 year or more). The user does not trade daily and waits patiently for fundamental price misalignment.
– Investment Objective: Determine optimal SELL zones for existing shares (blended profit-taking) and optimal RE-BUY / ACCUMULATION zones for future tranches. Immediate execution is NOT required.
– Output Language: English.

—

## INPUTS PROVIDED
Use the attached official filings and disclosures, including:
– 3 attached financial reports on Hainan Rubber (601118.SH), covering annual, interim, or quarterly disclosures.

—

## [ATTACHMENT & AS-OF CHECK] At the very top of your output, state:
1. Which documents are actually present in context, and which are missing.
2. The reference price, basic and diluted share count, and as-of date used for all valuation multiples.
If a required document or key table is missing, list it and proceed with available data, noting limitations. Do NOT fabricate numbers, citations, or disclosures. Mark missing items as “Not Disclosed” or “Not Available” with a one-line explanation.

—

## TASK
Produce a comprehensive, institutional-grade investment analysis for Hainan Rubber (601118.SH) using the attached official reports and verified external sector data.

1. Deconstruct core business fundamentals using attached filings.
2. Cross-reference market context: natural rubber prices (Shanghai Futures Exchange RU/NR contracts, Singapore SICOM TSR20, Tokyo TOCOM); synthetic rubber (butadiene, SBR) spreads and crude oil correlation; global supply-demand dynamics (Thailand, Indonesia, Vietnam, Ivory Coast production; ANRPC forecasts); weather patterns (typhoons, El Niño/La Niña impacts on Hainan and Southeast Asia); Chinese domestic agricultural policy (subsidies, rubber price insurance, SOE reform, rural revitalization); downstream tire demand (auto sales, heavy truck market, EV weight impact on tire wear, export tariffs, EUDR compliance); and Hainan Free Trade Port (FTP) policy impacts. Cite source and as-of date for every external data point. If live external data is unavailable, say so and label scenario assumptions [Speculation].
3. Construct a fundamental & cycle-driven multi-tranche accumulation ladder and profit-taking exit framework tailored to a ~25,000 RMB execution model, substituting the rigid 5% rule with your recommended cyclical/technical spacing. Provide specific price levels for selling and re-buying with exact rationales.
4. Uncover non-obvious structural nuances, hidden accounting factors (such as biological asset fair value adjustments under IAS 41 / CAS 5 and high-volume low-margin trading revenue dilution), or operational risks specific to plantation agriculture and SOE governance.

—

## CRITICAL ANALYTICAL RULES & DATA HANDLING

1. Source Hierarchy (apply in order of precedence):
– Tier 1: Attached company filings (Annual Report, Interim Report, exchange disclosures)
– Tier 2: Company IR materials, earnings calls, official announcements
– Tier 3: Regulator / exchange data (SSE, CSRC, Shanghai Futures Exchange, Hainan provincial government)
– Tier 4: Reputable third-party market data (ANRPC, IRSG, Bloomberg/Reuters, Wind, Mysteel)
– Tier 5: Analyst estimates / broker research (label as [Speculation] unless corroborated)
– Tier 6: Model inference (label [Interpretation] or [Speculation])
– Tie-break rule: When sources conflict, state the conflict, apply the higher tier, and note the lower-tier figure as a reconciliation item.

2. Strict Source Citations: Cite exact report sections, tables, or notes for every figure sourced from filings (e.g., [2025 Annual Report – Section 3, Main Financial Data]). If the exact section/table cannot be located in context, write “Section not identifiable in provided context” — do NOT invent a citation.

3. Restatements & Reclassifications: Flag any restated prior-period figures, segment reclassifications, biological asset revaluations, or accounting policy changes disclosed in the filings. If none are disclosed, state “No restatement disclosed.”

4. Financial & Accounting Rigor:
– Separate core segments: (a) Natural Rubber Plantation & Processing (upstream), (b) Rubber Trading & Distribution, (c) Rubber Wood & By-Products, (d) Synthetic Rubber / Other Chemical, (e) Non-core / Government Subsidies / Other.
– Adjust Net Profit for: Non-recurring items (government subsidies, asset disposals), biological asset fair value gains/losses (IAS 41 / CAS 5 impact), inventory write-downs, capitalized interest, and non-cash depletion/depreciation of mature plantations.
– State the annualization method used (H1 × 2, TTM, or run-rate) and the reason.
– Capital Intensity: Compute Capex/Revenue, Capex/D&A, and maintenance (replanting) vs growth capex split.
– Biological Asset Accounting: Detail valuation methodology (cost model vs fair value model), discount rate assumptions, and plantation age profile (immature %, mature %, over-mature %).
– Debt & Interest Coverage: Evaluate interest-bearing debt, short-term debt obligations, finance costs, and interest coverage ratios (EBITDA / Financial Costs).
– Government Subsidies & Policy Insurance: Identify the proportion of net profit dependent on Hainan government subsidies, rubber price insurance, or strategic stockpile allocations.

5. Data Gap Handling: Do NOT invent data. Label missing items “Not Disclosed” or “Not Available.”

6. Explicit Fact Tagging: Tag all non-historical claims as [Fact], [Interpretation], or [Speculation].

7. Scenario Modeling Required: Provide quantitative Bull, Base, and Bear scenarios with explicit:
– Natural rubber price assumptions (RMB/tonne for SHFE/domestic SCRWF or STR20; USD/tonne for SICOM TSR20)
– Production volume (tonnes) and yield per hectare assumptions
– Gross margin assumptions, processing volume assumptions, USD/RMB FX assumptions, and government subsidy assumptions
Assign probability weights (must sum to 100%) and compute probability-weighted expected value. Provide specific share price targets for each scenario.

8. Units & Currency: Default currency RMB. Rubber market prices in RMB/tonne (SHFE/Domestic) and USD/tonne (TSR20). Production and inventory in tonnes. State the USD/RMB and THB/RMB rates used.

9. Zero Trade Auditing: Do NOT critique or grade the user’s initial purchase of 5,900 shares at 4.63 RMB. Focus strictly on forward-looking strategy execution.

10. Bull Case Steelman: Before critiquing the bull thesis, state it in its strongest form (e.g., global rubber supply deficits from tree aging in Thailand/Indonesia, geopolitical strategic stockpile value, EUDR regulation compliance advantage for traceable Hainan supply, downstream tire demand recovery, EV weight-driven tire replacement acceleration).

11. Global vs Direct Exposure: Clearly separate direct operational exposure (Hainan plantation risks, labor, SOE governance, Hainan FTP policy) from indirect global exposure (Thai/Indonesian production, oil/synthetic rubber prices, Chinese auto/tire demand, weather patterns, FX fluctuations).

12. Peer Comparability: Compare Hainan Rubber with domestic/international peers (Sinochem International, Halcyon Agri, Sri Trang Agro-Industry, Thai Rubber Latex, Yunnan Rubber, Sailun) across P/E (reported & core), P/B, EV/EBITDA, gross margin, self-sufficiency ratio, and dividend yield. Flag comparability limits.

13. Reverse DCF / Implied Expectations: State what the current market price implies about forward natural rubber prices (RMB/tonne), volume growth, and operating margins using a simplified run-rate or 2-stage DCF framework. Compare implied expectations to your base case.

14. Cyclical vs Structural Framing: Classify the current situation as cyclical, structural, or mixed. State what evidence would shift the classification.

15. Internal Consistency Check: Verify that every number appearing in more than one section is identical. Flag and resolve any inconsistency rather than silently reconciling.

16. Disclaimer: Educational and strategic analysis only. Not financial advice. Include this disclaimer in Section 9.

17. Pacing & Length: Dense, rigorous, and exhaustive. Use markdown tables for all numeric comparisons. Only stop early if approaching output limits — complete through Section 5, write “[CONTINUE IN NEXT TURN]” and await user prompt. When continuing, do NOT repeat or re-summarize earlier sections; maintain all previously stated numbers and assumptions.

—

## OUTPUT FORMAT & REQUIRED SECTIONS

### 0. Attachment & Assumption Header
– Documents present / missing.
– Reference price, basic and diluted share count, and as-of date for all multiples.
– FX rates used (USD/RMB, THB/RMB).
– User’s current position: 5,900 shares @ 4.63 RMB.

### 0A. Key Metrics Dashboard
– One compact table of the 15–20 most important figures used throughout the report (Revenue, Core EBIT, Net Profit reported vs adjusted, OCF, FCF, Net Debt/EBITDA, annual NR production volume, self-owned yield vs external procurement, average selling price, plantation age profile, inventory, biological asset value, gross margin, P/E, P/B, EV/EBITDA, dividend yield).

### 1. Executive Summary
– 5–8 high-density bullets covering thesis, normalized valuation verdict, probability-weighted expected value, top 3 risks, cyclical vs structural classification, specific SELL zone for existing shares, specific RE-BUY zones (per your recommended ladder), actionable stance, and one-sentence bull case steelman.

### 2. Capital Deployment, Accumulation Ladder & Exit Strategy
– Position Summary: 5,900 shares @ 4.63 RMB (~27,317 RMB baseline cost).
– **Critical Assessment of the 5% Drop Rule & Replacement Rationale:** Evaluate the structural shortcomings of a rigid 5% drop rule for a cyclical commodity stock. In a dedicated summary paragraph, state explicitly what you recommend replacing it with and why (incorporating technical support, rubber seasonality, and volatility).
– **Recommended Multi-Tranche Accumulation Ladder:** Define 3–5 accumulation tiers with specific trigger conditions (technical, seasonal, P/B floors, cost-curve supports). Calculate exact SSE round-lot share counts for ~25,000 RMB per tranche, total capital deployment, and spacing rationale.
– **Structured Exit & Re-Buy Ladder:** Define explicit profit-taking targets for the existing 5,900 shares (specific sell targets, share amounts, fundamental rationales) and exact conditions to re-buy shares post-profit taking. Evaluate liquidity constraints (ADV, bid-ask spreads, A-share ±10% limits, T+1 rules).
– **Catalyst Calendar (3–18 months):** Seasonal tapping resumption/wintering months (April/May and November/December), wintering lulls (December–March), typhoon season (June–October), futures trends, earnings releases, ANRPC data, tire demand data, government subsidies, Hainan FTP policy updates, and dividend dates.

### 3. Segment-Level Operations, Tapping Cycle & Asset Lifecycle
– Revenue, Gross Profit, and EBIT by segment: (a) Plantation & Processing, (b) Trading & Distribution, (c) Rubber Wood & By-Products, (d) Synthetic Rubber, (e) Other / Eliminations.
– Natural Rubber Production & Tapping Cycle Analysis: Hainan plantation acreage (mature vs immature vs over-mature), yield per hectare, tapping labor costs, seasonal factors (wintering Q1 drop vs peak Q3-Q4 volume), tree lifecycle and replanting cycles, global ANRPC supply trends, processing capacity/margins, inventory dynamics, and operational headwinds (typhoons, leaf fall disease).

### 4. Balance Sheet Safety, Debt & Financial Quality Matrix
– Financial Quality Matrix Table: YoY revenue growth, segment gross margins, core adjusted EBIT (ex-hedging/subsidies), OCF, FCF, Net Debt/EBITDA, cash conversion ratio.
– Quality of Earnings & Subsidies: Government subsidy contribution %, biological asset fair value impact, related-party revenue %, working capital dynamics (DSO, DIO, DPO), and trading segment low-margin pass-through dilution.
– Debt & Liquidity Structure: Short-term debt refinancing risks, pledged land use rights/equipment, interest coverage, and SOE implicit guarantees.
– Capital Allocation & Shareholder Structure: Controlling shareholder (Hainan SASAC / State Farms) share pledges, dividend payout policy and coverage, and equity history.

### 5. Downstream Demand, Industry Macro & Peer Positioning
– Downstream Sector Linkages: Tire manufacturing operating rates (OEM vs replacement), commercial vehicle/heavy truck sales, EUDR environmental compliance advantages, and EV weight-driven tire wear acceleration.
– Cost Curve & Peer Comparison Table: Hainan Rubber vs Sinochem International, Halcyon Agri, Sri Trang Agro-Industry, Yunnan Rubber, and Sailun across P/E, P/B, EV/EBITDA, gross margin, self-sufficiency ratio, and dividend yield. Flag comparability limits.
– Moat Assessment: State-owned land rights, scale advantages, and pricing power.

### 6. Valuation & Scenario Analysis
– Valuation Multiples: P/E (reported vs core), P/B, P/S, EV/EBITDA, FCF yield.
– Reverse DCF / Implied Expectations: What rubber price and margin does the current market valuation imply compared to your base case?
– NR Price Sensitivity Matrix: Impact of SHFE/domestic rubber prices (12,000, 14,000, 16,000, 18,000 RMB/t) on annual net profit and EPS.
– Scenario Model (Bull / Base / Bear): Independent derivation, probability weightings, probability-weighted expected value, specific share price targets, and reconciliation to user entry price (4.63 RMB).
– Margin of Safety & Valuation Verdict: Discount/premium to intrinsic value and stance [Undervalued | Fairly Valued | Overvalued].

### 7. Critical Unasked Insights & Structural Nuances
– 3–5 hidden operational or accounting factors: Trading revenue vs core processing margin dilution, government insurance scheme mechanisms, Hainan land leasehold renewal terms, biological asset fair value accounting distortions, and USD-denominated sales vs RMB cost currency mismatches.
– Market Mispricing: What the market is under-appreciating or over-estimating regarding Hainan Rubber’s cyclical turn.

### 8. Top 5 Risk Matrix
– Risk Ranking & Mitigation: Weather/typhoons, NR price collapse, high debt/refinancing, biological asset write-downs, and downstream demand slumps.
– Format: Risk | Probability (1–5) | Severity (1–5) | Early Warning Indicator | Risk Trigger Action.

### 9. Actionable Stance & Quarterly Monitoring KPIs
– Medium-Term Stance: [Accumulate on Weakness | Hold | Partial Profit Taking | Exit].
– Specific SELL Strategy: First and second sell zones with exact share amounts and rationales.
– Specific RE-BUY Strategy: Summary of your recommended tranche triggers and exact SSE round-lot share counts for ~25,000 RMB per tranche.
– Falsification Conditions: 3 specific, measurable conditions that would invalidate the thesis.
– Top 3 KPIs for Next Financial Report: Specific quantitative thresholds compared to current baseline values.
– Analytical Confidence Score: [Low | Medium | High] with rationale.
– Disclaimer: Educational and strategic analysis only. Not financial advice.

—

## CRITICAL REMINDERS
– User’s Primary Questions:
1. When should I SELL my 5,900 shares @ 4.63 RMB?
2. When should I RE-BUY? (You are free to propose your own accumulation spacing, not necessarily 5%.)
3. What other critical information do I need?
– Do NOT:
– Critique the 4.63 RMB entry
– Provide generic advice
– Fabricate data
– Assume the 5% drop rule is mandatory; critically assess it and propose your own.
– Do:
– State explicitly what you recommend replacing the 5% rule with and why
– Provide specific price targets (SELL and RE-BUY) with exact share counts (round lots)
– Analyze biological asset accounting (IAS 41 / CAS 5) and trading revenue dilution explicitly
– Justify your accumulation spacing with technical, seasonal, and fundamental reasoning

—

**End of Prompt**

---

AI的崛起，或更準確的說，大型語言模型 (LLM), 讓很多人的生活輕鬆多了。從工作到搜索，LLM已是很多人生活不可或缺的一部分。今天，我們探討使用LLM來做選股票和交易的研究。

很多人使用LLM的時候都是給很簡單的指示：分析我附上的財務報告，並且告訴我什麼是合理的買賣價格。這種指示既模糊，成果也不良好。指示越具體，報告就越好。以下我們看看怎麼用LLM做出一個詳細的指令。

例：海南橡膠（股票601118）
任務簡介：在每股4.63人民幣買了5900股。需知什麼價格能賣。

以下步驟供讀者參考：

1. 我使用 Deepseek(中國 LLM) 和 Gemini(美國 LLM)。輸入我要的內容，讓他們生成一個提示詞。注意它們的任務不是分析股票，而是生成一個可以用在別的 LLM上的提示詞。

2. 初步指示應包含股票的基本認知。以橡膠為例，提示詞應包含一個指令分析你附上的財務報告，新聞，經濟週期和橡膠的生產週期等等。報告也得包含任何對你有用，但是你忽略掉的關鍵信息。

3. 提示詞也必須包含你主要的目的。股票在什麼價格交易，預算和計劃是什麼。若有計劃，讓LLM批判性的評估你的計劃且提出一個新計劃。

4. 同樣指令，在 Deepseek 和 Gemini 都過一遍。兩個 LLM 的免費版本額度挺高的，可進行多次迭代。

5. 把雙方生成出來的提示詞複製黏貼到另一個LLM，讓他們批判性分析，並結合兩者做法最佳的部分。重複2到3遍，然後每個迭代就只使用一個LLM，輪迴換，每一次迭代都讓他們在最新基礎上再進步。每一次的迭代你都必須細讀有什麼更改，有必要時還得手動更改一些指示（記得通知LLM並解釋緣由）。這個過程需進行多幾次迭代。

6. 最後將你滿意的成果用在 Claude 或 ChatGPT。這兩個的LLM使用額度比較低，但是生成的報告質量較高。每個 LLM 都有自己的優劣勢。 Deepseek善於邏輯思考和數學， Gemini擅長搜索。但在深度研究和報告上，比不上 Claude 和 ChatGPT。

這方法能快速的幫你啃下財務報告，挖掘新點子和你沒考慮過的思路，也提供了一些不一樣的意見。

但切記，LLM 還是個會犯錯的工具，不能全信。LLM 可給你一個計劃，但用戶需謹慎看過，自行驗證真假和最終自行做決定。

一個 AI 故事：有個人在地上看到了一些蘑菇，並問 AI 可否食用。AI 說可以，那人便吃了。最後中毒，質問 AI，AI便回：是的你說的對。抱歉我弄錯了。這蘑菇有毒且不能食用。

以下附上一個提示詞的樣本：

# 海南橡膠（601118.SH）LLM分析提示詞

## 角色與視角
扮演一位經驗豐富的農產品股票分析師與週期性價值投資者。將扎實的基本面分析功底與對全球天然橡膠（NR）和合成橡膠（SR）市場、種植園經濟學、加工利潤、生物資產會計（IAS 41 / CAS 5）以及中國國有企業（SOE）動態的專業知識相結合。你應具備分析性、客觀性、懷疑精神，並專注於風險調整後的中長期資本配置。優先關注現金流現實、種植園生命週期動態（未成熟與成熟面積）、商品價格週期、庫存動態以及資產負債表韌性，而非會計層面的營收噪音。

—

## 用戶交易背景與策略假設
– 目標資產：中國海南橡膠產業集團股份有限公司（A股：601118.SH）。
– 當前持倉：5,900股，平均成本4.63元人民幣（總基準投資約27,317元人民幣）。
– 訂單執行單位：每筆買單固定份額約25,000元人民幣。根據你引用的當前市場價格，按上海證券交易所整手規則（最接近25,000元人民幣的100股整數倍）換算為實際股數；不要假設過時價格或零碎/非整手股數。
– 建倉策略要求：
* 用戶此前使用 rigid “每下跌5%從4.63元人民幣買入一份” 的規則。
* 你不受此規則約束。你必須批判性地評估該規則（指出其缺陷，如忽視技術支撐位、橡膠季節性週期和波動性，可能導致資本過早耗盡）。
* 你必須提出並論證一個更優的、以基本面和週期為錨定的建倉框架（例如，P/B估值底部、SHFE/SICOM成本曲線支撐位或Q1冬季停割期低點）。在一段專門的總結段落中明確說明你建議用什麼替代5%規則以及為什麼。
– 持有期容忍度：數週至12個月（或更長）。用戶不每日交易，耐心等待基本面價格錯配。
– 投資目標：確定現有股票的賣出區域（分批止盈）以及未來份額的再買入/建倉區域。不要求立即執行。
– 輸出語言：英文。

—

## 輸入材料
使用所附官方文件與披露，包括：
– 3份關於海南橡膠（601118.SH）的所附財務報告，涵蓋年度、中期或季度披露。

—

## [附件與截止日期檢查] 在輸出最頂部，說明：
1. 上下文中實際存在哪些文件，缺少哪些文件。
2. 用於所有估值倍數的參考價格、基本和稀釋股數以及截止日期。
如果所需文件或關鍵表格缺失，請列出並基於可用數據繼續，註明局限性。不要捏造數字、引用或披露。將缺失項標記為「未披露」或「不可獲得」，並附一行解釋。

—

## 任務
使用所附官方報告和經核實的外部行業數據，為海南橡膠（601118.SH）製作一份全面的、機構級投資分析。

1. 使用所附文件解構核心業務基本面。
2. 交叉參考市場背景：天然橡膠價格（上海期貨交易所RU/NR合約、新加坡SICOM TSR20、東京TOCOM）；合成橡膠（丁二烯、SBR）價差與原油相關性；全球供需動態（泰國、印度尼西亞、越南、科特迪瓦產量；ANRPC預測）；天氣模式（颱風、厄爾尼諾/拉尼娜對海南和東南亞的影響）；中國國內農業政策（補貼、橡膠價格保險、國企改革、鄉村振興）；下游輪胎需求（汽車銷量、重卡市場、電動汽車重量對輪胎磨損的影響、出口關稅、EUDR合規）；以及海南自由貿易港（FTP）政策影響。為每個外部數據點引用來源和截止日期。如果實時外部數據不可用，請說明並標註情景假設為[推測]。
3. 構建一個以基本面和週期驅動的多份額建倉階梯和止盈退出框架，適用於約25,000元人民幣的執行模式，用你推薦的週期性/技術性間距替代僵化的5%規則。提供具體的賣出和再買入價格水平，並給出確切理由。
4. 揭示非顯而易見的深層結構、隱藏的會計因素（如IAS 41 / CAS 5下的生物資產公允價值調整以及高銷量低利潤率的貿易收入稀釋），或種植園農業和國企治理特有的運營風險。

—

## 關鍵分析規則與數據處理

1. 來源層級（按優先級順序應用）：
– 第一層：所附公司文件（年度報告、中期報告、交易所披露）
– 第二層：公司IR材料、財報電話會議、官方公告
– 第三層：監管機構/交易所數據（SSE、CSRC、上海期貨交易所、海南省政府）
– 第四層：信譽良好的第三方市場數據（ANRPC、IRSG、彭博/路透、Wind、Mysteel）
– 第五層：分析師估計/經紀商研究（除非得到證實，否則標註為[推測]）
– 第六層：模型推斷（標註為[解讀]或[推測]）
– 打破平局規則：當來源衝突時，說明衝突，採用較高層級，並將較低層級數據作為調節項註明。

2. 嚴格來源引用：為每個來自文件的數字引用確切的報告章節、表格或附註（例如，[2025年年度報告 – 第3節，主要財務數據]）。如果確切章節/表格無法在上下文中定位，請寫「在提供的上下文中無法識別章節」——不要捏造引用。

3. 重述與重分類：標記文件中披露的任何重述前期數字、分部重分類、生物資產重估或會計政策變更。如果未披露，請說明「未披露重述」。

4. 財務與會計嚴謹性：
– 分離核心分部：(a) 天然橡膠種植與加工（上游），(b) 橡膠貿易與分銷，(c) 橡膠木與副產品，(d) 合成橡膠/其他化工，(e) 非核心/政府補貼/其他。
– 調整淨利潤：非經常性項目（政府補貼、資產處置）、生物資產公允價值損益（IAS 41 / CAS 5影響）、存貨減值、資本化利息以及成熟種植園的非現金折耗/折舊。
– 說明所使用的年化方法（H1×2、TTM或運行率）及原因。
– 資本密集度：計算資本支出/收入、資本支出/折舊與攤銷，以及維護性（補種）與增長性資本支出的劃分。
– 生物資產會計：詳述估值方法（成本模型與公允價值模型）、折現率假設以及種植園樹齡結構（未成熟%、成熟%、過熟%）。
– 債務與利息覆蓋：評估有息債務、短期債務義務、財務費用以及利息覆蓋比率（EBITDA / 財務費用）。
– 政府補貼與政策保險：識別淨利潤中依賴海南省政府補貼、橡膠價格保險或戰略收儲分配的比例。

5. 數據缺口處理：不要捏造數據。將缺失項標記為「未披露」或「不可獲得」。

6. 明確事實標註：將所有非歷史性主張標註為[事實]、[解讀]或[推測]。

7. 情景建模要求：提供定量的牛市、基準和熊市情景，明確：
– 天然橡膠價格假設（SHFE/國內SCRWF或STR20為元人民幣/噸；SICOM TSR20為美元/噸）
– 產量（噸）和每公頃單產假設
– 毛利率假設、加工量假設、美元/人民幣匯率假設以及政府補貼假設
為每個情景分配概率權重（總和必須為100%），並計算概率加權期望值。為每個情景提供具體的股價目標。

8. 單位與貨幣：默認貨幣為人民幣。橡膠市場價格以元人民幣/噸（SHFE/國內）和美元/噸（TSR20）表示。產量和庫存以噸表示。說明所使用的美元/人民幣和泰銖/人民幣匯率。

9. 零交易審計：不要批評或評價用戶最初以4.63元人民幣購買5,900股的行為。嚴格聚焦於前瞻性策略執行。

10. 看漲情景最強論證：在批評看漲論點之前，以其最強形式陳述（例如，泰國/印度尼西亞樹齡老化導致全球橡膠供應短缺、地緣政治戰略收儲價值、EUDR法規合規優勢使海南可追溯供應受益、下游輪胎需求復甦、電動汽車重量驅動的輪胎更換加速）。

11. 全球與直接敞口：清晰區分直接運營敞口（海南種植園風險、勞動力、國企治理、海南FTP政策）與間接全球敞口（泰國/印度尼西亞產量、石油/合成橡膠價格、中國汽車/輪胎需求、天氣模式、匯率波動）。

12. 同業可比性：將海南橡膠與國內/國際同行（中化國際、Halcyon Agri、Sri Trang Agro-Industry、Thai Rubber Latex、雲南橡膠、賽輪）在P/E（報告與核心）、P/B、EV/EBITDA、毛利率、自給率和股息率方面進行比較。標註可比性局限。

13. 反向DCF / 隱含預期：說明當前市場價格隱含的遠期天然橡膠價格（元人民幣/噸）、銷量增長和營業利潤率，使用簡化運行率或兩階段DCF框架。將隱含預期與你的基準情景進行比較。

14. 週期性與結構性框架：將當前情況分類為週期性、結構性或混合型。說明什麼證據會改變分類。

15. 內部一致性檢查：驗證在多個部分出現的每個數字是否完全相同。標記並解決任何不一致，而不是默默調和。

16. 免責聲明：僅用於教育和策略分析。不構成財務建議。在第九節中納入此免責聲明。

17. 節奏與長度：密集、嚴謹、詳盡。所有數字比較使用markdown表格。只有在接近輸出限制時才提前停止——完成至第5節，寫「[在下一輪繼續]」並等待用戶提示。繼續時，不要重複或重新總結前面的部分；保持所有先前陳述的數字和假設。

—

## 輸出格式與必需章節

### 0. 附件與假設標題
– 存在/缺失的文件。
– 參考價格、基本和稀釋股數以及所有倍數的截止日期。
– 使用的匯率（美元/人民幣、泰銖/人民幣）。
– 用戶當前持倉：5,900股 @ 4.63元人民幣。

### 0A. 關鍵指標儀表盤
– 一個緊湊表格，包含報告中使用的15–20個最重要數字（收入、核心EBIT、報告與調整後淨利潤、OCF、FCF、淨債務/EBITDA、年度NR產量、自有產量與外購比例、平均售價、種植園樹齡結構、庫存、生物資產價值、毛利率、P/E、P/B、EV/EBITDA、股息率）。

### 1. 執行摘要
– 5–8個高密度要點，涵蓋論點、正常化估值結論、概率加權期望值、前三大風險、週期性與結構性分類、現有股票的具體賣出區域、未來份額的具體再買入區域（根據你推薦的階梯）、可操作立場，以及一句話看漲情景最強論證。

### 2. 資本部署、建倉階梯與退出策略
– 持倉摘要：5,900股 @ 4.63元人民幣（約27,317元人民幣基準成本）。
– **對5%下跌規則的批判性評估與替代理由：** 評估僵化的5%下跌規則對週期性商品股票的結構性缺陷。在一段專門的總結段落中，明確說明你建議用什麼替代它以及為什麼（結合技術支撐、橡膠季節性和波動性）。
– **推薦的多份額建倉階梯：** 定義3–5個建倉層級，具有具體的觸發條件（技術、季節性、P/B底部、成本曲線支撐）。計算每份約25,000元人民幣對應的確切SSE整手股數、總資本部署以及間距理由。
– **結構化退出與再買入階梯：** 為現有5,900股定義明確的止盈目標（具體賣出目標、股數、基本面理由）以及止盈後再買入的確切條件。評估流動性約束（ADV、買賣價差、A股±10%漲跌幅限制、T+1規則）。
– **催化劑日曆（3–18個月）：** 季節性開割/停割月份（4月/5月和11月/12月）、冬季停割期（12月–3月）、颱風季節（6月–10月）、期貨趨勢、財報發布、ANRPC數據、輪胎需求數據、政府補貼、海南FTP政策更新以及股息日期。

### 3. 分部運營、割膠週期與資產生命週期
– 按分部劃分的收入、毛利和EBIT：(a) 種植與加工，(b) 貿易與分銷，(c) 橡膠木與副產品，(d) 合成橡膠，(e) 其他/抵消。
– 天然橡膠生產與割膠週期分析：海南種植面積（成熟 vs 未成熟 vs 過熟）、每公頃單產、割膠勞動力成本、季節性因素（Q1冬季停割供應下降 vs Q3-Q4產量高峰）、樹木生命週期和補種週期、全球ANRPC供應趨勢、加工產能/利潤、庫存動態以及運營逆風（颱風、落葉病）。

### 4. 資產負債表安全、債務與財務質量矩陣
– 財務質量矩陣表：同比收入增長、分部毛利率、核心調整後EBIT（剔除套保/補貼）、OCF、FCF、淨債務/EBITDA、現金轉化率。
– 盈利質量與補貼：政府補貼貢獻百分比、生物資產公允價值影響、關聯方收入百分比、營運資金動態（DSO、DIO、DPO）以及貿易分部低利潤率過手稀釋。
– 債務與流動性結構：短期債務再融資風險、質押的土地使用權/設備、利息覆蓋以及國企隱性擔保。
– 資本配置與股東結構：控股股東（海南省國資委/海南農墾）股份質押、股息支付政策與覆蓋以及股權歷史。

### 5. 下游需求、行業宏觀與同業定位
– 下游行業關聯：輪胎製造開工率（OEM vs 替換）、商用車/重卡銷量、EUDR環境合規優勢以及電動汽車重量驅動的輪胎磨損加速。
– 成本曲線與同業比較表：海南橡膠 vs 中化國際、Halcyon Agri、Sri Trang Agro-Industry、雲南橡膠和賽輪，在P/E、P/B、EV/EBITDA、毛利率、自給率和股息率方面進行比較。標註可比性局限。
– 護城河評估：國有土地權、規模優勢和定價權。

### 6. 估值與情景分析
– 估值倍數：P/E（報告與核心）、P/B、P/S、EV/EBITDA、FCF收益率。
– 反向DCF / 隱含預期：當前市場估值隱含的橡膠價格和利潤率與你的基準情景相比如何？
– NR價格敏感性矩陣：SHFE/國內橡膠價格（12,000、14,000、16,000、18,000元人民幣/噸）對年度淨利潤和EPS的影響。
– 情景模型（牛市/基準/熊市）：獨立推導、概率權重、概率加權期望值、具體股價目標，以及與用戶入場價（4.63元人民幣）的對賬。
– 安全邊際與估值結論：相對於內在價值的折價/溢價以及立場[低估 | 公允 | 高估]。

### 7. 關鍵未問及見解與結構性細微差別
– 3–5個隱藏的運營或會計因素：貿易收入與核心加工利潤稀釋、政府保險機制、海南土地租賃續期條款、生物資產公允價值會計扭曲以及美元計價銷售與人民幣成本貨幣錯配。
– 市場錯誤定價：市場對海南橡膠週期轉折低估或高估了什麼。

### 8. 前五大風險矩陣
– 風險排序與緩解：天氣/颱風、NR價格暴跌、高債務/再融資、生物資產減值和下游需求下滑。
– 格式：風險 | 概率（1–5）| 嚴重性（1–5）| 預警指標 | 風險觸發行動。

### 9. 可操作立場與季度監控KPI
– 中期立場：[弱勢建倉 | 持有 | 部分止盈 | 退出]。
– 具體賣出策略：第一和第二賣出區域，附確切股數和理由。
– 具體再買入策略：總結你推薦的建倉觸發條件以及每份約25,000元人民幣對應的確切SSE整手股數。
– 證偽條件：3個具體的、可衡量的、會使論點失效的條件。
– 下一份財務報告的三大KPI：與當前基線值相比的具體定量閾值。
– 分析置信度評分：[低 | 中 | 高]及理由。
– 免責聲明：僅用於教育和策略分析。不構成財務建議。

—

## 關鍵提醒
– 用戶的主要問題：
1. 我應該在什麼時候賣出我的5,900股 @ 4.63元人民幣？
2. 我應該在什麼時候再買入？（你可以自由提出自己的建倉間距，不一定是5%。）
3. 我還需要哪些其他關鍵信息？
– 不要：
– 批評4.63元人民幣的入場
– 提供泛泛的建議
– 捏造數據
– 假設5%下跌規則是強制性的；批判性地評估它並提出你自己的方案。
– 要：
– 明確說明你建議用什麼替代5%規則以及為什麼
– 提供具體的價格目標（賣出和再買入）以及確切的股數（整手）
– 明確分析生物資產會計（IAS 41 / CAS 5）和貿易收入稀釋
– 用技術、季節性和基本面理由論證你的建倉間距

—

**提示詞結束**