Encoding geometric patterns

Discussions about the testing and simulation of mechanical trading systems using historical data and other methods. Trading Blox Customers should post Trading Blox specific questions in the Customer Support forum.
Post Reply
blueberrycake
Roundtable Knight
Roundtable Knight
Posts: 125
Joined: Mon Apr 21, 2003 11:04 pm
Location: California

Encoding geometric patterns

Post by blueberrycake » Wed Mar 16, 2005 6:27 pm

Has anyone seen any good algorithms for identifying various geometric formations on price charts? In particular, things like spotting minor/major support and resistance levels, identifying consolidation channels, etc. Basically things that are pretty easy to spot on a chart through visual inspection, yet non-trivial to encode in a programmatic way for systematic trading.

-bbc

verec
Roundtable Knight
Roundtable Knight
Posts: 162
Joined: Mon Jun 09, 2003 7:04 pm
Location: London, UK
Contact:

Post by verec » Wed Mar 16, 2005 7:46 pm

Not to be teasing, but ... Isn't the answer contained in the question?

Let's put it that way: it is probably as easy for the eye to spot the patterns you mention, as it is for the same eye to miss the crucial "anti-pattern" that voids the first ...

Not to mention the fact, as Dr Elder puts it, that your eye usually concentrates on "the middle of the chart" whereas the actual trading always takes place at the right edge ...

blueberrycake
Roundtable Knight
Roundtable Knight
Posts: 125
Joined: Mon Apr 21, 2003 11:04 pm
Location: California

Post by blueberrycake » Wed Mar 16, 2005 8:15 pm

verec wrote: Let's put it that way: it is probably as easy for the eye to spot the patterns you mention, as it is for the same eye to miss the crucial "anti-pattern" that voids the first ...
That's very true. Hence the need to give a mechanical definition to some of these concepts so that they can be tested. For things like support/resistance levels, I think it's actually pretty well defined as various local maximums/minimums. One can certainly tweak the parameters of the definition, but it's probably one of the less hazy concepts out there.

bbc

verec
Roundtable Knight
Roundtable Knight
Posts: 162
Joined: Mon Jun 09, 2003 7:04 pm
Location: London, UK
Contact:

Post by verec » Wed Mar 16, 2005 9:00 pm

My point was that the so-called "patterns" are probably deceptive. And if I had an algorithm to spot "double-tops" as easily as the eye does it, my contention is that it would be as ineffective as humans are in actually making use of it.

Assuming this was possible, you would run into all sorts of issues like scale (is a 2 points "flag" as much of a "flag" as a 20 points "flag") and axis distortion: when is the declining slope of the first leg of your W (double-bottom) steep enough to qualify?

What is "obvious" to the eye is also misleading.

JG
Contributing Member
Contributing Member
Posts: 6
Joined: Fri Jan 07, 2005 5:48 pm

Post by JG » Wed Mar 16, 2005 11:13 pm

William Eckhardt stated something along these lines nicely in his MW interview (not verbatim).
-----
The probability of having a reversal day, given the market is at a price extreme, is a very different probability than the probability that the market is at a price extreme, given the fact that a reversal day just occurred.

If 85% of tops/bottoms have property x, but property x occurs very often when the market action is not a top or bottom, then trading property x probably isn't nearly as good as it sometimes appears to the human eye on a chart.
-----

I believe the mistake the human eye and mind make is much as Mr. Eckhardt stated in this interview. It is the difference between prior and posterior probabilities. Many things that look nice on a chart look that way because the pattern has already played out. It didn't look that way while it was playing out.

I do not mean this in a manner that indicates I am against testing chart patters, far from it. I do believe (from experience) that much time can be spent in pursuit of dead ends in this area though. This is especially so, I believe, if one is looking for the patterns they already want (read biased) to find.

ES
Roundtable Fellow
Roundtable Fellow
Posts: 97
Joined: Mon May 05, 2003 1:02 am

THE MODEL

Post by ES » Fri Mar 18, 2005 1:13 pm

The greatest challenge is to understand who the participants in the market are. It is these people that define the price of a stock or a bond or a commodity. Seperating retail and institutional action is my first filter.

I want to see if blind money is coming in or out of the market. I also want to see what type of institutional action is going on and compare these to the trend.

We could basically program a model to mimic the actions of each portfolio manager in the world. everyone discloses what their funds contain and their trading methodolgies. we can easily determine how many trend followers exist and what there weighiting is according to the dollars injected or taken out of the market, daily.

Was there a trend in oil, you better believe it. On simple breakout rules, a ton of trend followers were long. But so may have been the rest of the market. Then there's the contrarian who risks his ..... on the line.
However, perhaps we can agree on one matter, exists are more important than entries. This is what seperates the bigger money makers from the others. And the bigger the account size the more diffilcut this becomes.

Post Reply