2025 年 08 月 03 日 止跌的三个观察视角 金融工程定期报告 证券研究报告 本期要点：止跌的三个观察视角 上期我们认为短期或有颠簸，但是整体趋势状态或仍然向上。事后来 看，市场确实在 3600 附近出现了自 6 月底以来相对较大的震荡调整。 从走势对称的角度以及从重要关口阻力位的角度看，市场对当前位置 震荡调整的关注度比较高，一类是担心本轮上涨是不是已经结束了， 一类是关心本轮调整何时将再次迎来机会。 除非是出现特别疯狂的走势或基本面（含基本面预期）出现确定性变 化，否则一波趋势的结束往往是需要反复验证的。在当前状态下，我 们新研发的择时系统的中期趋势指标仍然维持在偏强状态，这往往表 明需要反复测试 3670-3674 这一区域才有足够的证据来讨论趋势是 否结束的问题，因此本期我们主要讨论这波调整的技术面止跌信号。 从缠论视角看，一个有效的顶分型放量跌破 5 日或 10 日均线之后， 往往很可能会出现一波明显的调整，也就是至少会形成日线级别的向 下一笔。 从量能角度看，牛市中的下跌缩量到一定程度之后往往是企稳的重要 标志，快速上涨的牛市中一波像样的调整往往要求至少缩量到 55%之 下，而慢牛中的缩量 ...

Quantitative Models and Construction Methods 1. Model Name: Divergence-Based Turning Points - **Model Construction Idea**: The model identifies potential price turning points by observing divergences between price trends and volume or technical indicators (e.g., MACD). Divergences signal weakening momentum, suggesting a possible reversal in the price trend [2][9]. - **Model Construction Process**: 1. Identify price making new highs or lows. 2. Check if corresponding volume or MACD fails to make new highs or lows, indicating divergence. 3. Confirm divergence using additional signals: - For MACD, use green bar shortening or yellow/white line crossover for confirmation [10]. - Apply wave theory to filter valid signals by identifying five-wave structures in trends [14][16]. 4. Combine divergence signals with moving averages to determine market conditions (e.g., bull or bear market) [14][16]. - **Model Evaluation**: Divergence signals alone have a success rate of less than 55% for predicting turning points. However, combining them with wave theory and moving averages improves accuracy to over 65% [13][16][17]. 2. Model Name: V-Shaped Reversal Turning Points - **Model Construction Idea**: This model uses the "Temperature Indicator" to measure the degree of price deviation from moving averages, identifying extreme conditions that may signal V-shaped reversals [3][18][19]. - **Model Construction Process**: 1. Calculate a moving average (e.g., 60-day for short-term trends, annual for long-term trends). 2. Shift the moving average left by half its parameter length. 3. Linearly extrapolate the last two points of the shifted moving average. 4. Compute the deviation (bias) of each price point from the extrapolated moving average. 5. Calculate the percentile rank of the deviation over a rolling window to derive the "Temperature Indicator," which ranges from 0 to 100 [18][19]. 6. Define thresholds for identifying turning points: - In bear markets, both high-frequency and low-frequency temperature indicators must fall below 10. - In range-bound markets, only the high-frequency indicator below 10 is sufficient. - In bull markets, consider additional risks and adjust thresholds (e.g., high-frequency indicator below 15 or 10) [19][21][26]. - **Model Evaluation**: The model effectively identifies turning points in various market conditions but requires adjustments for bull markets to account for strong trends and potential false signals [27][28]. --- Model Backtesting Results 1. Divergence-Based Turning Points - **Accuracy**: Basic divergence signals have a success rate below 55% but improve to over 65% when combined with wave theory and moving averages [13][16][17]. 2. V-Shaped Reversal Turning Points - **Bear Market**: High-frequency and low-frequency temperature indicators below 10 successfully identified rebounds in January, February, and April 2022, each lasting over a month [21]. - **Range-Bound Market**: High-frequency temperature indicator below 10 identified rebounds in January and April 2025 during a three-quarter-long consolidation phase [22][25]. - **Bull Market**: High-frequency temperature indicator below 10 or 15 identified five strong buying opportunities in gold futures from 2023 to 2025 [26]. --- Quantitative Factors and Construction Methods 1. Factor Name: Temperature Indicator - **Factor Construction Idea**: Measures price deviation from moving averages to identify extreme overbought or oversold conditions [18][19]. - **Factor Construction Process**: 1. Use a moving average (e.g., 60-day or annual) to represent the trend. 2. Shift the moving average left by half its parameter length. 3. Linearly extrapolate the last two points of the shifted moving average. 4. Calculate the deviation (bias) of each price point from the extrapolated moving average. 5. Compute the percentile rank of the deviation over a rolling window to derive the factor value, ranging from 0 to 100 [18][19]. - **Factor Evaluation**: The factor effectively identifies extreme market conditions but requires different thresholds for bear, range-bound, and bull markets to optimize performance [19][21][26]. --- Factor Backtesting Results 1. Temperature Indicator - **Bear Market**: High-frequency and low-frequency indicators below 10 identified rebounds in January, February, and April 2022 [21]. - **Range-Bound Market**: High-frequency indicator below 10 identified rebounds in January and April 2025 [22][25]. - **Bull Market**: High-frequency indicator below 10 or 15 identified five strong buying opportunities in gold futures from 2023 to 2025 [26].