--- In
amibroker@xxxxxxxxxxxxxxx, "Paul Ho" <paultsho@xxx> wrote:
>
> One problem for yahoo data for Australian stocks is that price is
only
> quoted down to 2 decimal place only. In real life, 3 decimal place
is quoted
> at the exchange.
>
>
> _____
>
> From:
amibroker@xxxxxxxxxxxxxxx [mailto:
amibroker@xxxxxxxxxxxxxxx]
On Behalf
> Of brian_z111
> Sent: Friday, 15 February 2008 11:38 PM
> To:
amibroker@xxxxxxxxxxxxxxx
> Subject: [amibroker] Re: New EOD data feed other than Quotes Plus
>
>
>
> Ara,
>
> For the sake of sharing information - to save people wasted time:
>
> > I know many are willing to to do a lot of work in order to use
free
> >data from Yahoo or add fundamenttal info from Yahoo.
>
> It's no work at all.
> Yahoo and Yahoo current fundamentals are integrated with AmiQuote >
AB
>
> >>I have not tried it, but my impression has been that what you get
> >>is what you pay for.
>
> Yahoo data is not free - it is paid for by Yahoo (bought from
> reputable companies) from revenue from ads - it is distributed for
> free to attract users to the site who then buy things via the ads
(I
> sometimes do).
>
> >I would rather pay a reasonable fee and get a reliable product.
>
> How do we know that our data products are quality? - because the
> provider told us so?
>
> How can we check data and how many do?
> It is not so easy to check data.
>
> As I suggested to Barry the other day - we can go to Yahoo,
download
> raw, or fully adjusted, get dividend info, get split info and do
our
> own checks on Yahoo in the same way we can do that for any other
> provider. I have done one or two here and there and I haven't found
> anything alarming to me. If I did I would then work around that.
>
> What we can't get from Yahoo is event adjusted data without
dividend
> adjustment - people who want that have to get it elsewhere.
>
> Also, sadly, for longer term traders, more error is introduced into
> backtesting by using data without div adjustment than errors within
> Yahoo's data (I only say that to help people).
>
> IMO - first we should know what our trading style is, second we
> should have a strategy in mind and then third we should buy data
that
> will give us the best chance to analyze past performance of that
> strategy (in my case that means a few different types of data).
>
> As you know I could give a couple of pages more in support of
traders
> taking a more flexible and proactive attitude to their data but I
> would be better to do that at the UKB.
>
> One example:
>
> I recently developed a very promising new system on holiday - just
> before walking out the door to catch the plane I copied a complete
AB
> directory from an XP machine and dumped it on a Vista laptop
> (naughty, naughty) - I dumped all the old databases to get the size
> down and only took a copy of Jim's US Yahoo database (I wanted div
> adjusted data for EOD backtesting).
>
> I managed to get some good work done.
>
> When I got back I then carried on towards testing in intraday data
> from another provider and so far it is performing exactly as tested
> on Yahoo (only better) - I am using intraday to finesse the entries
> and exits that I couldn't finesse in EOD data. The trades average 2-
3
> days round trip.
>
> Also when it is ready to trade I will be totally familiar with it
> after thousands of tests etc and I will see straight away if
> something is odd when I start going to the brokers (who also has a
> different source for data).
>
> Also consider that the so-called differences in data is a bit of a
> myth. In Australia there is only one exchange - they have a
monopoly
> on the data. They do some filtering of the tick data themselves
> before releasing it to their providers so how different can it be.
>
> The only differences in data in Australia that can occur are from
a)
> extra filtering or b) event adjustments.
>
> At the prices I don't think too many here are doing to mucn of a)
and
> b) is checkable.
>
> Also, as we have seen in a recent article, bad tick data isn't the
> end of the world at the EOD level.
>
> In fact EOD traders can tolerate quite a bit of error and still win
> the day.
>
> brian_z
>