Thanks Richard, I'm looking at 5min ES, EX, DAX, SPI. Don't need less than that
- but still concerned that if data reliability is not there, the only solution
is manual live forward testing .... bummer
Cheers, PiF
--- In [email protected], "timekeeper_origen" <timekeeper_ori...@...>
wrote:
>
> --- In [email protected], "pfgrillet" <pfgrillet@> wrote:
> >
> >
> > Thanks Richard
> >
> > Interestingly, Reuters is not listed in the recommended Data providers
> >
> > Also came across Futures Truth, who are professionally forward testing
> > systems (www.futurestruth.com <http://www.futurestruth.com> ) and they
> > are using CSI, Prophet and Tick Data. See results in their magazine.
> >
> > Cheers,
> richard
> >
> >
> > --- In [email protected], "timekeeper_origen"
> > <timekeeper_origen@> wrote:
> > >
> > > --- In [email protected], "pfgrillet" pfgrillet@ wrote:
> > > >
> > > > Backtesting without reliable intraday data is pointless.
> > > > I've had experience comparing intraday backtests with live trading
> > forward testing and having completely different bars.
> > > >
> > > > I am keen to get other users' feedback on their experience with
> > intraday data reliability (tick data preferably) - any advise of a
> > particular data vendor?
> > > >
> > > > Cheers, P-F - Australia
> > > >
> > >
> > > reuters - opting for their scrubbing
> > > good luck
> > > richard
> > >
> >
> hi pif
>
> been there done that with those three and others. there are big differences
> in providers' "integrity" and core competencies (e.g. CSI vs. Tick Data..am
> unfamiliar with Futures Truth)
>
> may i ask what frequency bars you require?
>
> if you're looking at ultra high frequency you'll encounter a problem with
> something called jitter in systems that can - for very specific reasons -
> generate the frustrating differences in the data you're seeing even stemming
> from the same provider.
>
> ugly example: sourcing bars from NASDAQ db itself and taking an exchange
> member colo'd recorded comparison (i've done this, unfortunately) can create
> differences of up to ~10% "error" in some of these numbers when looking at
> every second of a full year's worth of 100 stocks.
>
> even more reason to head off to the pub if you decide to measure sub-second
> lengths -
>
> richard
>
