Re: S&P Tick Counts

[Carroll Slemaker <cslemaker1@xxxxxxxx>]

Mark Brown wrote:

> Otherwise you can expect data buffer overflow.  Which is exactly
> what many of you guys are experiencing when you compare ticks
> sent each day.

James Mazzulla wrote:

> With BMI and Signal there's no "data cleansing" like there
> is with CQG and, ostensibly, PCQuote. Consequently, our data
> contains all of the busted trades and none of the
> insertions. This may account for the discrepancies.

I don't believe that either of these hypotheses is the real reason for
the low tick counts we're seeing on the S&P.

First, I'm interested in trading only the S&P so I do not collect data
on thousands, or even hundreds, of issues as many who trade stocks do. 
Second, traders with a wide range of computing power have all reported
low counts.  In fact, I've yet to see a report from a SINGLE DBC
subscriber reporting a count close to what it should be.  Finally, I
have on several occasions compared a tick file of S&P data from other
DBC subscribers with my own collected data and I have found VERY few
discrepancies.  If either of us were losing a substantial portion of the
data within our own systems, the discrepancies would be much higher.

As for deletes and inserts in the official CME files, the program I have
written to obtain statistics from those files takes those kinds of
entries into consideration (it includes deletes, excludes inserts). 
When counting in the time-&-sales file it also excludes multiple
consecutive trades at the same price.  After processing in this manner,
both the time-&-sales file and the volume-by-tick file give counts which
are very close to one another - and my counts on collected DBC data USED
to be within 2% or 3% of these CME counts.  But not now.

Regards,

Carroll Slemaker