Pattern searching using candles may be significant with EOD data but EOD
results are very unlikely to apply to Intraday patterns. candles lack
important information (like HL precedence). Intraday patterns are arbitrary
and are subject to timing, timing shifts, data delays, time frames, etc. An
Intraday pattern will not retain its shape when any of these non-price
variables change. I have another demo somewhere that explores that...may look
it up when i find the time. Intraday patterns should really be run in the
smallest timeframe possible.
My preferred approach was to overlay a relative time vs price matrix on the
chart. Matrix dimensions can be referenced to the initial open price of the
first bar in the matrix, and can be expressed in percent change. If a quote
falls in one of the cells it is set (binary). This way you end up with a
binary image consisting of many small binary squares (like a checker board).
This image can be expressed with a unique string or number. You determine
pattern profitability by sliding this matrix over your data and looking for
profit. This is very computational intensive...
However.... this can lead to a very unique and interesting trading systems.
Such a system would capture the current n-bar pattern. You then take this
pattern and slide it over you historical data, calculating the profits of this
pattern as you go along. This way you will get a Profitability rating for your
CURRENT price pattern. This pattern is totally random and not related to
any traditional patterns - however if the system finds that this pattern was
historically profitable the system would make the trade. Thus this system will
trade heretofore unknown patterns that are defined in real time as data
arrives. The advantage is that this system doesn't require a pattern library.
There is much more to it... actually very interesting. Because you can collect
profitable/unprofitable patterns in a file and gradually build an
empirical pattern library, completely automated. Gradually this system would
become more efficient (computationally) as the library grows, and winning
patterns are traded and losing patterns are rejected.
Instead of sampling raw price you can of course sample indicators, or
combinations of indicators, in the same way. Indicators tend to compress price
data and reduce the computational overhead. Nevertheless, It can get quite
complex. Many years ago I spend a significant amount of time on this however
the computing overhead was just too much - perhaps with faster computers this
idea can be revived.
best regards,
herman
For tips on developing Real-Time Auto-Trading systems visit:
http://www.amibroker.org/userkb/
Thursday, February 28, 2008, 8:42:28 AM, you wrote:
> Thanks Herman,
> Just when I was going to go to bed too!
> brian_z
> Please note that this group is for discussion
between users only.
> To get support from AmiBroker please send an e-mail
directly to
> SUPPORT {at} amibroker.com
> For NEW RELEASE ANNOUNCEMENTS and other news always
check DEVLOG:
>
http://www.amibroker.com/devlog/
> For other support material please check
also:
>
http://www.amibroker.com/support.html
>
> Yahoo! Groups Links
> <*> To visit your group on the web, go
to:
> http://groups.yahoo.com/group/amibroker/
> <*> Your email settings:
> Individual Email |
Traditional
> <*> To change settings online go
to:
> http://groups.yahoo.com/group/amibroker/join
> (Yahoo! ID required)
> <*> To change settings via email:
> mailto:amibroker-digest@xxxxxxxxxps.com
> mailto:amibroker-fullfeatured@yahoogroups.com
> <*> To unsubscribe from this group, send an
email to:
> amibroker-unsubscribe@xxxxxxxxxxxxcom
> <*> Your use of Yahoo! Groups is subject
to:
> http://docs.yahoo.com/info/terms/
>