主动买入 vs 主动卖出: 资金方向揭秘
45s清晰解释股市资金流入流出的核心概念，适合新手投资者快速理解。
▶ Play ClipAI Summary
This video explores the research on individual stock capital flows and main force behavior, distinguishing between capital direction (active buying/selling) and main force behavior. The creator tests whether reconstructing split orders from quantitative algorithms can improve stock price prediction, finding no significant improvement.
Chapters
[00:00]
Two dimensions of stock capital research
The video distinguishes between capital direction (active buying vs selling) and main force behavior. Active buying is defined as trades at bid-1 or higher, considered bullish (inflow), while active selling is at bid-1 or lower, considered bearish (outflow).
[00:41]
Main force behavior estimation
Main force behavior data varies across platforms. The previous video discussed split order algorithms used by quantitative institutions to hide large orders by breaking them into smaller ones. The goal is to reconstruct these split orders to estimate true main force behavior.
[01:05]
Threshold differences across platforms
Platforms like East Money set thresholds (e.g., orders over 500k yuan) to classify as medium or large orders. However, thresholds differ across platforms (e.g., Tonghuashun, Tongdaxin, East Money), leading to different daily capital flow data, though the underlying logic is the same.
[01:54]
Research approach
The research aims to test if active buying/selling (as seen in software like Tuoer) predicts stock prices, and whether reconstructed main force behavior adds predictive information. The study used 16 stocks with Level 2 data from August 3 to September 4, 2026.
[02:33]
Initial findings
Results show that even after reconstructing large/small orders, prediction improvement is not significant. The research is in a bottleneck. The creator suggests longer timeframes might help, and notes that defining large/small orders varies by stock (e.g., 200k for inactive stocks, 500k for large caps like Zhongji Xuchuang).
The research on reconstructing split orders to predict stock prices did not yield significant improvements, highlighting the complexity of defining main force behavior and the need for further exploration.
Mentioned in this Video
💡 Key Takeaways
📊
Defining capital direction
Provides clear definitions of active buying/selling, foundational for understanding capital flow analysis.
💡
Platform threshold differences
Highlights that capital flow data varies across platforms due to different thresholds, a key practical insight.
01:05
💡
Reconstruction doesn't improve prediction
Challenges the assumption that reconstructing split orders adds predictive value, a significant negative result.
02:33
⚖️
Stock-specific order size definitions
Emphasizes the need to customize large/small order thresholds per stock, adding complexity to analysis.
03:00Full Transcript
[00:00] 今天简单给大家录一个关于 个股资金和主力行为的研究 我们这里头要区分两个维度 第一个是资金的方向 就是主动买入和主动卖出
[00:15] 主动性买入指的是 以买一及以上价格成交 视为看多 即如流入 主动性卖出呢 主动性卖出就叫内盘
[00:27] 以买一及以下价格成交 视为看空 就为流出 那除了这个资金的方向之外 还有一个主力行为 目前主力行为呢 市场上的数据是五花八门
[00:41] 什么都有 那我们上一个视频里头 讲到那个拆单算法 也是想还原一些量化机构 或者说主力通过把大单拆成小单的形式
[00:53] 隐藏行为的这种方式来做 那就是怎么去更准确的估算这个主力行为 那在主力行为上目前来讲
[01:05] 就是包括东方财富 他们都是设定一个预值 你比如说这一笔大于50万 就认为说这是中单或者大单 那所以这个标准的划分
[01:19] 每一家都不一样 比如同花顺同达信 东方财富都有些微的差异 你跟他我们从他各个软件里都看到的当天的资金流入 或者说主大单的行为其实是不一样的
[01:33] 但是他的逻辑是一样的 那现在我们做的工作就是什么呢 就是说从被量化软件 用拆单算法拆单的这些行为把它还原成真正的主力行为那么现在做的工作呢就是第一个我们想看一下这个资金方向
[01:54] 就是当天的主动买入或主动卖出 就是我们看到那个托儿的软件里头的 那个主力净流入主力净量
[02:06] 这个结果呢会对股价有没有预测的作用 另外一个就是还原后的主力行为 能否增加预测信息 目前我是选了16只个股
[02:20] 因为Level 2的数据特别大 我是用了2026年8月3号到9月4号 这么一个数据做了一个研究和分析
[02:33] 目前来看其实结果并没有我们想象那样 就是说他即使把他把这个大单小单进行还原之后 他的预测也没有显著的改善
[02:47] 所以这个呢 嗯 这个研究目前来讲也陷入一些瓶颈 或者说大家有没有什么更好的研究方向也可以提一下 当然我们这个做的这个研究是时间比较短
[03:00] 也许长一点会好一点 另外一个还发现的就是说 这个大单和小单的这个定义 其实对某些股票是不一样的 你比如不是特别活跃的股票 你可能定于20万以上就算大单了
[03:13] 但对于中继虚创这样的大盘股的话 你可能50万以上才能算作中单 所以这个方面的话呢 也需要根据每个个股去来定义
[03:25] 所以这个工作量就会比较大了 那么讲来讲 这个算下来呀 也没有什么特别的收获 就跟大家分享这么多 谢谢大家
