May

25

Closing Time

May 25, 2021 |

Sushil Kedia writes:

Closing TIme of key contracts, around the world has the same character feeding the vig. Irrespective of whether this character speaks Japanese, Korean, Chinese, Malaysian, Hindi, Pashto, Hebrew, German, English or American English. 

The compulsion to not carry a losing trade overnight, to square off excess positions that cannot be funded overnight etc. etc. provide a good enough number of hands who are willing to be forced out and required to be forced out at close. 

If I can spot, from my back-benches in global finance a ready made bunch of pigs to be slaughtered everyday, I am wondering why wont the 200 Billion Dollar Liquidity pumps whether run by a rocket scientist or by anyone would not be already squeezing them hour by hour as the sun moves from the East to the West? 

Wondering what is a good way to structure a study that tries to isolate statistical evidence for reversing extremas N minutes before the close of related exchanges. 

Say the Closing TIme of top 5 liquidity producing exchangs of crude future world over are noted down and a statistical study of N minutes before closing time and after closing time of each of these 5 exchangs throws up a pattern? 

And then if equity index futures that produce the top 10 volumes even if each equity index contrat is a distinct entity, is there a closing time ebb and flow that is being created by the Scientists' algorithms?  

Victor Niederhoffer writes

This is a very interesting and  an suggestive post. let's have some   feedback on ow to approach this query 

Jared Albert  writes

I think closing time/price as the sole predictor is too broad and noise will swamp any effect. 

So, to me,  the first step is classify the various conditions that exist before the close. For example, days up vs down, up/down on day, distance from x day max/min etc.

There are so many predictor variables that I don't think this is a frequentist kind of problem lending itself to logistic regression and lots of crosstabs for example.

So step one is a machine learning classification model to separate the states using the closing time movement as the target for training. 

IF it turns out that there are classifiable 'set-ups', then one could run analysis within the most promising classifications.


Comments

WordPress database error: [Table './dailyspeculations_com_@002d_dailywordpress/wp_comments' is marked as crashed and last (automatic?) repair failed]
SELECT * FROM wp_comments WHERE comment_post_ID = '13349' AND comment_approved = '1' ORDER BY comment_date

Name

Email

Website

Speak your mind

Archives

Resources & Links

Search