filtering bad ticks

Questions about MultiCharts and user contributed studies.
bibulous
Posts: 56
Joined: 16 Apr 2010
Has thanked: 17 times
Been thanked: 2 times

filtering bad ticks

Postby bibulous » 21 Apr 2010

Hi,

is it possible to let MultiCharts automatically filter out bad ticks? For example if there is a price difference of x% or if last, bid or ask are 0?

Cheers,
Marc

User avatar
TJ
Posts: 7740
Joined: 29 Aug 2006
Location: Global Citizen
Has thanked: 1033 times
Been thanked: 2221 times

Re: filtering bad ticks

Postby TJ » 21 Apr 2010

Hi,

is it possible to let MultiCharts automatically filter out bad ticks? For example if there is a price difference of x% or if last, bid or ask are 0?

Cheers,
Marc
not yet...

my understanding is,
this feature has been requested and is scheduled for a future release.

bibulous
Posts: 56
Joined: 16 Apr 2010
Has thanked: 17 times
Been thanked: 2 times

Re: filtering bad ticks

Postby bibulous » 22 Apr 2010

not yet...
my understanding is,
this feature has been requested and is scheduled for a future release.
ok, then I will have to work-around this in my strategie.. The chart is still messed up though..

User avatar
TJ
Posts: 7740
Joined: 29 Aug 2006
Location: Global Citizen
Has thanked: 1033 times
Been thanked: 2221 times

Postby TJ » 22 Apr 2010

you can manually edit the data with QuoteManager.

bibulous
Posts: 56
Joined: 16 Apr 2010
Has thanked: 17 times
Been thanked: 2 times

Postby bibulous » 22 Apr 2010

you can manually edit the data with QuoteManager.
Yes, I know. I am thinking about my automated strategy though..

User avatar
TJ
Posts: 7740
Joined: 29 Aug 2006
Location: Global Citizen
Has thanked: 1033 times
Been thanked: 2221 times

Postby TJ » 22 Apr 2010

...get better datafeed.

I don't know how or when the tick filter will screen the data.
would the bad tick hit your strategy before it is being filtered?

if you are concerned, you should add a filter in your strategy to prevent bad ticks.

User avatar
Bruce DeVault
Posts: 438
Joined: 19 Jan 2010
Location: Washington DC
Been thanked: 2 times
Contact:

Postby Bruce DeVault » 22 Apr 2010

This is actually a general problem, and is tougher than a lot of folks realize, especially for strategies that try to take a lot of trades with relatively small expectancies. The key thing is that data feeds typically provide real-time tick data that's used by the platforms to build minute and other kinds of bars. The data feeds also provide historical minute and daily data, but these historical bars aren't used at all in real-time - what's traded in real-time is the tick data, as combined as needed to make minute/daily bars etc. The data feed often has filters in place on the tick data, and processes in place to handle correction ticks etc. but these are often not the same from tick to minute data, so what happens often is that the minute bars built in real-time from the arriving tick data don't precisely match the minute bars delivered historically because the data feeds typically clean up and apply corrections to the historical data a lot more than they can clean up the real-time arriving tick data.

The best way to address this very broadly, besides having "reasonable" filters in place in the platform and in the strategy where appropriate, is to make sure your strategy has enough "room for error" in its performance statistics that these types of issues can be there, and still safely be in an area of likely profit. It's when you have a strategy that's marginally successful e.g. it has solid winning statistics, but they're very small, that there's the most danger from this type of thing causing real-time discrepancies, and this together with a host of similar sort of "gotcha conditions" such as limit order fills etc. is one reason lots of "scalping-type" strategies with large numbers of small trades fail in real-time at the retail level despite quite impressive looking back-tests.

janus
Posts: 835
Joined: 25 May 2009
Has thanked: 63 times
Been thanked: 104 times

Postby janus » 22 Apr 2010

so what happens often is that the minute bars built in real-time from the arriving tick data don't precisely match the minute bars delivered historically because the data feeds typically clean up and apply corrections to the historical data a lot more than they can clean up the real-time arriving tick data.
Except in the case of IB, their historical volume is not cleaned but actually "corrupted", and they know about it but refuse to fix it. :shock:


Return to “MultiCharts”