docs: Update AI deep analysis description to a P1-P4 priority chain and revise trading tips.

README.md

@@ -46,13 +46,14 @@ Evaluates environmental stability from the last 4 METAR observations. Higher = m
 
 ### 3. 🤖 AI Deep Analysis (Groq LLaMA 3.3 70B)
 
-Feeds wind speed, wind direction, cloud cover, solar radiation, and METAR trend data into LLaMA 70B:
+Feeds all weather data into LLaMA 70B, analyzed via a **P1→P4 Priority Chain**:
 
-- **Logical Reasoning**: 2-3 sentences analyzing airport dynamics, explicitly referencing Open-Meteo forecast and DEB blended values as benchmarks.
-- **Time Awareness**: Analysis considers how much time remains until the predicted peak, judging remaining warming potential.
-- **Market Call**: Explicitly states the expected peak time window and specific temperature betting range. Calls "dead market" when cooling is confirmed.
-- **Confidence Score**: Quantitative 1-10 confidence rating.
-- **High Availability**: Built-in auto-retry + fallback model degradation (70B → 8B) to withstand Groq API 500/503 outages.
+- **P1 Real-Time Rhythm** (highest priority): 2 consecutive METAR highs → still warming; 2 non-highs past peak → dead market. Warming under low radiation → advection-driven, forecasts often underestimate.
+- **P2 Inhibitors**: Humidity >80% **and** BKN/OVC sustained 2 reports → effective suppression. "Partly cloudy" alone is insufficient.
+- **P3 Math Probability**: References settlement probability but cannot override P1 observations.
+- **P4 Forecast Background**: DEB/forecasts used for ceiling estimation; downweighted when actuals exceed them.
+- **Dead Market Trigger**: Past peak window + 2 consecutive non-highs + cloud buildup or precipitation → dead market declared.
+- **High Availability**: Auto-retry + fallback model degradation (70B → 8B) to withstand Groq API outages.
 
 ### 4. ⏱️ Real-time Airport Observations (Zero-Cache METAR)
 
@@ -151,12 +152,12 @@ graph TD
 
 ## 💡 Trading Tips
 
-1. **Watch Settlement Probability**: The probability engine is math-based and more objective than AI judgment. When one temperature has > 65% probability, the direction is relatively clear.
-2. **Observe Time Decay**: Probabilities auto-lock as time progresses. After peak hours, the engine narrows σ dramatically, concentrating results around the observed max.
-3. **Reference DEB Blended Value**: When models diverge, the DEB corrected value is usually more reliable than any single forecast.
-4. **Observe AI Confidence**: A score below 5 indicates high uncertainty—consider staying on the sidelines.
-5. **Watch Settlement Boundaries**: When the observed high is near X.5, be wary of rounding jumps during WU settlements.
-6. **Distribution Center μ**: The μ value shown in the probability display represents the algorithm's expected most likely actual high temperature—compare it directly with the Polymarket odds.
+1. **Real-time Rhythm First**: AI analysis follows P1→P4 priority. If live METAR trends (P1) conflict with math probabilities (P3)—e.g., probability favors 7°C but it’s still surging toward 8°C—always prioritize the live trend.
+2. **Watch Settlement Probabilities**: Based on Gaussian models, direction is most certain when a temperature has > 70% probability while P1 rhythm is flat.
+3. **Reference DEB Bias**: Use `/deb` to check for systematic bias. If a city is consistently "underestimated," habitually bid one WU notch higher.
+4. **Identify Dead Market Signals**: When AI declares a "Dead Market," it usually means warming power is exhausted (post-peak window + no new highs + cloud buildup). This is an opportunity to harvest remaining value.
+5. **Mind the Boundaries**: When the observed high is near X.5 (e.g., 7.50°C), be wary of Wunderground rounding up to 8 due to tiny fluctuations.
+6. **Center Point μ**: The μ value represents the expected actual high. When market prices deviate significantly from μ, an arbitrage opportunity may exist.
 
 ---
 

README_ZH.md

@@ -46,13 +46,14 @@ PolyWeather 是一款专为 **Polymarket** 等预测市场打造的天气分析
 
 ### 3. 🤖 AI 深度分析 (Groq LLaMA 3.3 70B)
 
-将风速、风向、云量、太阳辐射、METAR 趋势等全部投喂给 LLaMA 70B 大模型：
+将全部气象数据投喂给 LLaMA 70B，按 **P1→P4 优先级链** 分析：
 
-- **逻辑推演**：用 2-3 句话深度分析机场动力因子对升温的促进/阻碍，明确引用 Open-Meteo 预报值和 DEB 融合值作为对标。
-- **时间感知**：分析中会考虑当前时间距预计最热时段的距离，判断剩余升温空间。
-- **盘口判定**：明确给出预计最热时段和具体博弈温度区间。降温确认后直接给出死盘结论。
-- **置信度评分**：1-10 分量化置信度参考。
-- **高可用保障**：内置自动重试 + 备用模型降级机制（70B → 8B），抵御 Groq API 503/500 故障。
+- **P1 实况节奏**（最高优先级）：连续 2 报创新高→升温未止；连续 2 报未创新高且过峰→偏死盘。低辐射升温→暖平流驱动。
+- **P2 阻碍因子**：湿度>80% **且** BKN/OVC 持续 2 报→压温有效。单"多云"不足以判断受限。
+- **P3 数学概率**：参考结算概率分布，但不可压过 P1 实况。
+- **P4 预报背景**：DEB/预报用于判断上沿，实测超预报时降权。
+- **死盘判定**：峰值窗口已过 + 连续 2 报未创新高 + 云量回补或降水→判定死盘。
+- **高可用保障**：自动重试 + 备用模型降级（70B → 8B），抵御 Groq API 故障。
 
 ### 4. ⏱️ 实时机场观测 (Zero-Cache METAR)
 
@@ -151,12 +152,12 @@ graph TD
 
 ## 💡 交易提示
 
-1. **紧盯结算概率**：概率引擎基于数学模型计算，比 AI 的主观判断更客观。当某个温度概率 > 65%，说明方向较为明确。
-2. **关注时间衰减**：概率会随时间推进自动锁定。峰值过后，概率引擎自动缩窄 σ，结果高度集中在实测最高值附近。
-3. **参考 DEB 融合值**：当多模型分歧较大时，DEB 的修正值通常比单一预报更具参考意义。
-4. **观察 AI 置信度**：置信度低于 5 分时，说明当前气象环境处于高度不确定状态，建议观望。
-5. **注意结算边界**：实测最高温接近 X.5 时，需警惕 Wunderground 结算时的进位跳动。
-6. **分布中心 μ**：概率展示中的 μ 值代表算法预期的最可能实际最高温，可与 Polymarket 盘口价格直接对比。
+1. **实况节奏优先**：AI 分析遵循 P1→P4 优先级。如果实况趋势（P1）与数学概率（P3）冲突（例如概率看好 7°C 但实况仍猛涨冲向 8°C），请务必以实况走势为准。
+2. **紧盯结算概率**：概率引擎基于数学模型，当某个温度概率 > 70% 且 P1 节奏持平时，方向最为明确。
+3. **参考 DEB 偏差**：通过 `/deb` 查看城市的系统性偏差。如果某个城市经常“低估”，交易时应习惯性看高 1 个 WU 档位。
+4. **识别死盘信号**：当 AI 判定“死盘”时，通常意味着升温动力彻底枯竭（峰值窗后+不创新高+云量回补），此时是反向对收割残余价值的机会。
+5. **注意结算边界**：实测最高温接近 X.5（如 7.50°C）时，Wunderground 可能会因极微小波动从 7 进位到 8，需防范“偷鸡”。
+6. **分布中心 μ**：μ 值代表算法预期的实际最高温中心点，直接对标盘口价格。当价格严重偏离 μ，通常存在套利空间。
 
 ---