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

INDU Data Error / 60-Tick Phenomenon


  • To: IUhrik@xxxxxxx
  • Subject: INDU Data Error / 60-Tick Phenomenon
  • From: Carroll Slemaker <cslemaker1@xxxxxxxx>
  • Date: Wed, 2 Sep 1998 17:32:17 -0400 (EDT)
  • In-reply-to: <199809021656.JAA00264@xxxxxxxxxxxxxx>

PureBytes Links

Trading Reference Links

First, let me say that this problem has not been peculiar to BMI
subscribers.  I use Signal and I've seen it also.

Now, Igor, I disagree with one aspect of your explanation - that the
problem begins with receipt of a bad tick from the data vendor.  I have
seen true bad ticks - they are quite obvious on a tick or 1-minute chart
as a sharp spike up or down.  In my experience, they have been rare.

In the cases recently reported - and which I experienced also - there
was NO such spike observed on the real-time chart, NONE, ANYWHERE.

Secondly, if, as has been reported, Omega's data blocks contain about 60
ticks and if the observed error occurs only when the alleged bad tick
coincides with the initial tick of a block, then we should expect to
see, on average, 60 times as many mid-block bad ticks as initial-tick
bad ticks.  This is FAR from what I have seen, however.  I don't
remember the last time I saw a true bad tick (a real-time spike) in
INDU, but "The-Problem" has occurred on three of the last five market
days and has occurred three or four times on each of those days!

No, this is almost absolutely certainly a bug in the Server.  It is
probably the kind of bug common to the C programming language and other
such non-strongly-typed languages (I'm a Pascal fan, myself) - an
inappropriate value has probably, under some context-dependent
circumstance, been stored in a variable and overflowed, corrupting a
different variable happening to occupy an adjacent memory location.

Such a bug will cause symptoms which appear sometimes, and sometimes
not.  Also, it can, under varying circumstances, cause different
symptoms which appear totally unrelated.  In other words, the actual bug
which is the source of "The-Problem" may be totally unrelated to the
compression and storage of price ticks - that simply happens to be the
function which gets clobbered by it.

Carroll Slemaker