NED 原理 3.2

moreliver

mrbean

turntome

喜欢读！！！

大牛Steve

befriend

uddi

tcm328pan

chaoskid

wzpzpw

源济

谢谢分享！

meshfree

谢谢分享！

ricky

莫大看看我理解得对不对。x是NED， y 是由3020个NED数据点拟合出来的回报。假设实际回报是z那么R=(y*z/z^2).
我觉得z应该参与拟合而切y应该是NED此时和过去一段时间的函数。

如果莫大有兴趣可以share一下NED和实际回报数据大家一起坦讨一下。

littletiger

2012 paper (it has 71 pages...) is very helpful to explain NED. Some takeaways from reading the paper. (i may misunderstand some of the points)
1) it seems to suggests use Tick data (volume buckets instead of time buckets), right?
2) using that method, paper shows it has about 40% correlation between return and VPIN, (i.e.) NED here.
3) It draws several conclusions such as high VPIN cause high volatility of price movement, i.e., imbalance of demand and supply will cause big swing of
price movement.
4) on the other hand, Not all volatility is due to toxicity (high VPIN)

randomperson

tao8688

randomperson 发表于 2020-6-8 09:17 AM

mitwkwk

拟合的时候，点越多，效果就越差，因为很多noise都被带进去了。可以试试看做个平均以后，把曲线平滑了再拟合，可能会好点

chaoskid

如果把相同的程序用在每日的开盘，最高，最低会不会给出类同的R2值？若是，那说明该程序方法有效性值得商酌
或用在future上会不会给出不同的R2值？
瞎想的

georgeyn

高级算法

square