[This message was posted by Richard Labs of CL&B Capital Management, LLC 
<[email protected]> to the "Algorithmic Trading" discussion forum at 
http://fixprotocol.org/discuss/31. You can reply to it on-line at 
http://fixprotocol.org/discuss/read/905908df - PLEASE DO NOT REPLY BY MAIL.]

Jason,

Thanks very much for the post! 

The WG has other requests for just that functionality so we are likely going to 
elevate looking at this. I want to strongly encourage you to stay with us on 
this and provide further input!

We created the "indexing/categorization" set of attributes initially as a 
marketing concept. We wanted allow the buy side, when faced with a "shoebox" 
filled to the brim with various and sundry FIXatdl algos from a wide variety of 
sources, (including entities they had never done business with before) to have 
"hooks" attached to each strategy to allow browsing through that "pile" and 
presenting to traders sub sets that _might_ be relevant to their trading 
objective, right at the "purchasing moment". 

In the schema, we isolated the index entries for that marketing purpose 
thinking, "what if a BD gets aggressive and loose with their indexing" (i.e. 
the strategy works well in most of that region specified however there are 
several troublesome exceptions". We didn't want "looseness" in 
categorization/indexing/marketing area to leak into the much more tightly 
controlled  "operations" end. 

There was much debate regarding how to index. With nowhere near universal 
agreement. Significant dissension. In the end the majority ruled. So, it's not 
an exceptionally tight spec in that area. 

The idea, back then was to keep use of the indexing entries isolated. 

We also, from the start, came up with the idea "don't allow/encourage" a single 
strategy to "morph" itself into a different strategy. The idea was each 
strategy was to accomplish a single trading technique (what you see is what you 
get). 

Lastly, a few years back when we did the original design, we found that BD's 
often had completely different data centers / computer set ups in different 
regions. Therefore a VWAP in North America, ran on a data center in that 
region, while the VWAP in EMEA ran on a wholly different data center in that 
region. Often times this set up was left over from M&A activity and still 
fairly loosely integrated. Although the VWAP (and other algo) names were the 
same (and marketing worldwide was consistent), and the algos were indeed 
similar, they were NOT 100% identical. (Often the underlying FIX messages were 
radically different. And routing might also be different.) 

Each strategy also had the potential to behave differently (for example 
different risks in use: near options/future expiration, same named parameter 
such as "aggression level" working radically different in different regions, 
different "got-ya's" such as around market open/close, holidays, circuit 
breaker interruptions, resets after interruptions, etc.). We also felt the 
written disclosure document would likely be wholly different for each region 
(if the underlying execution infrastructure was different). Back then we had 
the concept of one strategy, one disclosure document. KISS. 

We also have the one strategy, one FIX order type concept. That is, if the 
strategy results in a new order single, there is no way to "morph" that into a 
basket order. 

The OMS/EMS faces a daunting job of integrating across BD offerings if 
strategies are not granular and "atomic". If they morph (and each will morph in 
different ways), with some BDs offering 5 underlying strategies under a single 
name with morphing, while the next BD presents them as individual strategies, 
that makes the "selection" process exceptionally difficult. Some OMS's want to 
take all VWAP strategies from all its BDs and present each strategy to the 
trader with all the fields in identical position. That requires granularity 

Way, way down the road (perhaps 5+ years out) we also envisions "algos of 
algos" - that is the ability to build a higher level algo that presented a nice 
clean UI but in fact "under the hood" deployed multiple individual algos (even 
across BDs!) to execute its objectives. For that kind of "tinker toy" building 
of complex strategies to work the underlying algos at the base level really 
need generic interfaces. 

Just wanted to give some background. WG needs more input here - particularly to 
gauge 1. how urgent is this need to use indexing data in "operations", 2. will 
the loose "indexing" we now permit cause real trouble if it leaks into to 
operations, 3. where is the best place to "draw the line" on morphing 
strategies - AT THIS TIME? What allows the broadest FIXatdl deployment without 
causing too much implementation pain "downstream" from the FIXatdl file itself. 
Often we have to take "baby steps". (Particularly if there is "looseness")

I'd strongly encourage others to weight in here! Please offer specific use 
cases if at all possible and clearly indicate how large the challenge is, how 
urgent the solution sought is.  

A simple change to FIXatdl (really just an ADDITION to the schema, no other 
structural changes or deletions) might be rolled in going from v1.1 -> v1.2. A 
complex change (altering anything now deployed in the marketplace would be v1.1 
-> v2.0 and would take much longer (i.e. years) to go through a much more 
formal and rigorous release process. 

Looking for all input. Thanks again Jason for posting!

[You can unsubscribe from this discussion group by sending a message to 
mailto:[email protected]]

-- 
You received this message because you are subscribed to the Google Groups 
"Financial Information eXchange" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to 
[email protected].
For more options, visit this group at 
http://groups.google.com/group/fix-protocol?hl=en.

Reply via email to