[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: A/C/E "one second netting"



PureBytes Links

Trading Reference Links

You may be interest to hear that both Reuters & Bloomberg appear to
be performing a version of this in-house (naturally without telling).

If you get the chance, DDE a Bloomberg & Reuters to the same Excel
screen, choose something very active like the BBO (best bid/offer) of a
highly liquid contract and watch for yourself... What it looked like to
me is Reuters were taking one second snapshots of the price, and
Bloomberg was averaging the activity over one second...

Both are bad, with the first you simply loose data, with the second you
don't see trades that actually took place and do see a trade that
*NEVER* (as a single transaction) took place.

I have tik data from the DTB (pre Eurex days) showing the Dax doing over
100 trades in under a minute.  If that's the case, the Bid/Offer should
just be a blur to the human eye... Is it?

I personally have never understood why the exchanges don't implement
the most basic of quality control over the data vendors.  I have the CME
spec for data vendors and they recommend 6 ISDN lines for a full
exchange feed.  Do the maths on that and then tell me how Joe Blogs data
vendor fits that down a 28.8k/56k modem with TCP/IP overheads...
Impossible, the only way its done is to throw away (or loose) data, and
that's exactly what the vendors do.  I tested some internet vendors and
found up to 50% (in some cases more) of data was lost during an average
hour.

You have to remember, if guys like Bloomberg & Reuters who force clients
to have a dedicated 2 Mbit (minimum) pipe to them, still have to cut
corners, what hope do the low cost IP vendors have?

You also have to remember when talking data feeds, go as close to the 
source as you can.  A lot of data vendors are re-sellers of re-sellers
etc, some of the exchanges even encourage this as its less directly 
connecting infrastructure they have to worry about.  I would say as a
rule of thumb, you probably loose ~ 10% of data with each hop, so if
your 4 hops away from the exchange (exchange - large vendor -
smaller vendor - internet vendor) you can quickly see how the 50% loss
figure above is reached.

At one stage I was in Australia and subscribed to an SFE data feed
(yes the whole exchange).  The company providing the feed was supplied a
1200 bps RS-232 from the exchange and they literally just did an RX/TX 
crossover and feed this into racks and racks of modems, you then dialed
into your assigned modem.  Zero features, but boy was it honest, best
feed I ever had (remember SFE is a low volume exchange).

To me, the data vendors should think of themselves more like Telco's and
less like value added re-sellers as they currently do.  I think most
people don't care for how many indicators the proprietary vendor package
has and simply wants the data moved from A -> B without massaging or 
mangling it.


Justin

---

Stefan Schulz wrote:

I suspect it has something to do with sheer tick volume - Eurex has the 
highest turnover of all futures exchanges worldwide.  Reporting every 
tick would impose unreasonable bandwidth requirements, and institutional 
players (heck, not even I) don't care about every tick.  For an example 
of what a really liquid market reporting every tick looks like, check 
the E-Mini S&P or the CAC40 at busy times.  Now imagine 5 times that 
volume (Bund, Bobl, Schatz, DAX, Eurostoxx) plus all the "smaller" 
contracts, it's just not doable across a normal feed.

The sensible thing would have been to modify the feed to report highs 
and lows of shapshot intervals which are slightly longer.  Instead they 
went for reporting interval averages - that's one number instead of two 
per interval.  Cheap and cheerful.

At 22:50 10/09/03 +1000, Adrian Pitt wrote:

Thanks for the URL Stefan...very fascinating...who on earth invented
such a ludicrous system in this day and age of computers? LOL.

Adrian

> -----Original Message-----
> From: Stefan Schulz [mailto:prog1@xxxxxxxxxxxxxx]
> Sent: Wednesday, 10 September 2003 10:28 PM
> To: omega-list@xxxxxxxxxx
> Subject: RE: A/C/E "one second netting"
>
>
>  From kaj's earlier post:
> http://www.optioninvestor.com/futurescorner/fc_052703_01.asp
>
> At 21:26 10/09/03 +1000, Adrian Pitt wrote:
> >Pardon my ignorance, but can you tell me what 'A/C/E "one second
> >netting" ' is please.
> >
> >Adrian
>
>