Forums  > Risk Management  > Predicting Settlement SPAN on CME products before the SPAN file comes out?  
Page 1 of 1
Display using:  


Total Posts: 125
Joined: Nov 2005
Posted: 2017-04-18 22:26

CME produces files that describe the SPAN scenarios several times a day. We find that intraday SPAN (available at 1pm ET) can materially differ from Settlement SPAN (available ~6pm).

Given that we are primarily interested in SPAN on products which settle at 2:30 pm it seems possible to generate our own SPAN file at 2:30 by adjusting the intraday file or the previous day's settlement file with the newly known settlements.

Sounds like a fair bit of work. Anyone tried this or know of a vendor that provides this solution?

Founding Member

Total Posts: 1363
Joined: Mar 2004
Posted: 2017-04-18 23:32
I have not tried that. From working with those files a bit my hunch is the straight SPAN scenarios are pretty easy to get right - it's easy to imply out what the various price and vol bumps are and apply to new settles. Any of the more complicated rules not just relating to span scenarios I wouldn't know how to work with though, much less well documented (i.e. any offsets and such).

Also note that the final settles aren't officially out until a fair bit later than 2:30pm. Not that they tend to change much for products that settles primarily from screen trades. Sometimes some options may have small changes though.

Capital Structure Demolition LLC Radiation


Total Posts: 5
Joined: Jun 2016
Posted: 2017-04-29 07:03
Here we are trying to implement some probabilistic math does do not rely on classic assumptions nor in direct relation of the variables involved or the model from which they come but exclusively on the likelihood /unlikelihood of the scrambled serials involved.

To make things more difficult; like it is computationally unfeasible to do it by exhaustion; we need to rely on algorithms to reduce serials likelihood computation time.

So as our research depends only on the serials involved, the proper dataset; or a similar to desired datasets should be provided to us in order to experiment with the data and the outcome. As it is evident; nothing can go out if the inputs have not a relationship between them (even though; these relationships may be unknown)

The more data columns we have the better the chances we will find those they are the most suitable; because we try all the combinations in between.

Time interpolation (as you describe) is another matter again related to the nature of the variables involved. (Personally I see instruments related to time –like interest rates; or options-; with a better chance to perform)

As expected all this takes time; moreover considering we shall do it on a part time basis. Anyone is welcomed to submit us data for analysis; but also consider we are overloaded, so please have some patience.

All data format shall be numerical; (continuous or discrete) and formats should be sql; csv, or even txt

Do not hesitate to email us directly for more information on specific details.

(We are answering emails on the basis first come; first served ._))

Cheers, and all the best!
Previous Thread :: Next Thread 
Page 1 of 1