Serving the Quantitative Finance Community

 
User avatar
CommodityQuant
Topic Author
Posts: 35
Joined: July 5th, 2007, 6:16 am

Combining volatility with time

September 15th, 2022, 4:29 pm

If we want to see whether trades were "close" together in time, we might want to take volaility into account.
For example, if the time distance between the trades stays constant, we would regard trades in low-volatility conditions
as being closer together because we anticipate less market activity between the two times.

So there's a natural concept called "volatility-time" which takes into account both volatility and time in a single measurement of proximity.
Surely, this is done in finance?  However, I have been unable to google any reference to this concept whatsoever. I'd be grateful for any suggestions or references for such a metric.

Thank You,

CommodityQuant
 
User avatar
Alan
Posts: 2560
Joined: December 19th, 2001, 4:01 am
Location: California
Contact:

Re: Combining volatility with time

September 15th, 2022, 5:28 pm

google "business time sampling"
 
User avatar
DavidJN
Posts: 202
Joined: July 14th, 2002, 3:00 am

Re: Combining volatility with time

September 17th, 2022, 3:15 pm

Um… don’t we just call that (Sigma*Root Time) in finance? 
 
User avatar
Alan
Posts: 2560
Joined: December 19th, 2001, 4:01 am
Location: California
Contact:

Re: Combining volatility with time

September 19th, 2022, 2:25 am

not intraday
 
User avatar
ExSan
Posts: 415
Joined: April 12th, 2003, 10:40 am

Re: Combining volatility with time

September 19th, 2022, 12:46 pm

not intraday
Why not?
°°° #OpenToWork https://bit.ly/3klz5ys <--- my Tweets °| ° |°My Portfolio ---> https://bit.ly/3siPhoZ #OpenToWork °°°
 
User avatar
Alan
Posts: 2560
Joined: December 19th, 2001, 4:01 am
Location: California
Contact:

Re: Combining volatility with time

September 22nd, 2022, 2:53 pm

google "business time sampling"
 
User avatar
DavidJN
Posts: 202
Joined: July 14th, 2002, 3:00 am

Re: Combining volatility with time

September 23rd, 2022, 2:40 pm

There are more ticks at open and close, but it seems to me that tick time is business time and anything else is… something else.
 
From one of the related papers: “Our results show that the BTS (business time sampling) scheme performs better than the CTS (calendar time sampling) and the TTS (tick time sampling) schemes in yielding Gaussian returns.”
 
While transforming data to achieve more desirable statistical properties (e.g. first-differencing) is a time honoured tradition, I wasn’t aware that, despite the obvious convenience, achieving Gaussian returns was an explicit aim of financial engineering. That kind of thinking led to a world of hurt in the credit markets, no?
 
I’ve only looked at two or three papers, but the output of this research seems to be generating statistics instead of generating trading signals. You can sample such that you measure normal returns, okay but so what? What can you do with these supposedly improved estimates? What are the applications other than creatively characterizing data? How does this knowledge help the high frequency trader? What is the purpose of the paper other than another publication?
 
User avatar
bearish
Posts: 4877
Joined: February 3rd, 2011, 2:19 pm

Re: Combining volatility with time

September 23rd, 2022, 9:05 pm

I’m all in favor of convenience and, as an aside, I don’t think the Gaussian copula was a (let alone the) primary cause of the GFC, but in my experience time change can work great if you are in a univariate situation. Once you have more than one thing to worry about, though, they never seem to obey the same clock.
 
User avatar
Gamal
Posts: 956
Joined: February 26th, 2004, 8:41 am

Re: Combining volatility with time

Yesterday, 4:09 pm

I don’t think the Gaussian copula was a (let alone the) primary cause of the GFC.
Yes and no. On one hand it was obviously greed and short term thinking (up to the next bonus time) about long term investment but on the other hand the Gaussian copula does not control risk transfer within CDO^2 and higher powers. Michael Burry invented a simple non-mathematical trick and won.
 
User avatar
katastrofa
Posts: 6974
Joined: August 16th, 2007, 5:36 am
Location: Alpha Centauri

Re: Combining volatility with time

Yesterday, 8:36 pm

Humble noncognoscenti here: if you don’t trade, you don’t know the price, not to mention the volatility, so you have to assume some price distribution and then you can come up with the realised volatility, e.g. by estimating the variance = average of squared log(price1 / price2) / sqrt(t2-t1) over all increments?
 
User avatar
bearish
Posts: 4877
Joined: February 3rd, 2011, 2:19 pm

Re: Combining volatility with time

Yesterday, 9:14 pm

I don’t think the Gaussian copula was a (let alone the) primary cause of the GFC.
Yes and no. On one hand it was obviously greed and short term thinking (up to the next bonus time) about long term investment but on the other hand the Gaussian copula does not control risk transfer within CDO^2 and higher powers. Michael Burry invented a simple non-mathematical trick and won.
The main problem with the Gaussian copula and its cousins was that the presence of a model with a carefully crafted low-dimensional parameter space that could be calibrated to market data gave “senior management” the impression that the risk was manageable and indeed managed. A former manager of mine held strongly to the view that “it takes a model to beat a model”. The counter argument is that sometimes the lack of an *acceptable* model should be a reason not to do the trade. At least not to do it in size.
 
User avatar
Gamal
Posts: 956
Joined: February 26th, 2004, 8:41 am

Re: Combining volatility with time

Yesterday, 9:30 pm

The counter argument is that sometimes the lack of an *acceptable* model should be a reason not to do the trade. At least not to do it in size.
That's the statement that makes all traders furious.