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

Re: Study to Filter out Bad Ticks/Gaps/etc.



PureBytes Links

Trading Reference Links

Earl,
Thanks for the comments. I understand you comments about the volume.
I would appreciate if you could explain your filters a little bit more in
detail.
How do you decide at which point to cut the bar, for instance, if you've
detected abnormal high or low?
I thinks it's easier to do on real time tick data since I can flag  the bad
tick price as it
comes. But EOD, you don't know where to cut it and you can only
approximate?
Val

Earl Adamy wrote:

> I've written such routines in VB to filter EOD stock data. It's not
> difficult and I've never understood why vendors don't implement such
> filters. It's more difficult on futures and indexes where there is no
> related volume component against which to evaluate big price moves. The
> biggest problem I suspect you'll have is EL's 64k limit which could make it
> impossible to front-end a system with the filter code without use of a
> special DLL.
>
> Earl
>
> -----Original Message-----
> From: Val Clancy <valclancy@xxxxxxxxxxxxx>
> To: Omega List <omega-list@xxxxxxxxxx>
> Date: Monday, November 02, 1998 8:16 AM
> Subject: Study to Filter out Bad Ticks/Gaps/etc.
>
> >List,
> >I am writing a little EL routine to filter out bad ticks as
> >well as gaps in data, etc.
> >Bad ticks - spike high, spike low, zero low.
> >Gaps in data - missing data for continuous period of time.
> >This routine will find the above, flag it , tag it, print and file
> >append
> >and the main thing - stop the execution of the system on that
> >bar(s).
> >Anyway, my line of thought is to build bands around the price
> >with the idea of if price falls out of the band - means bad tick.
> >I figured the value of the band could be anything more then
> >5 times the STD of  price, considering log normal distribution.
> >Gaps I figured could be detected through time and volume.
> >If volume is normal and then all of sudden the next bar is missing
> >based on time - gap.
> >If any of the above happens the flag will be used to stop the
> >execution of the system on that bar and in case of the missing
> >data will exit or generate protective stop...
> >Questions:
> >Has anyone wrote anything like this?
> >Any suggestions? Any insights?
> >Val.
> >
> >