Patterns
Since we are pioneers in new trading technologies, I've been considering a new product that I call "Patterns". Yes, yes, patterns themselves are not new, but we always come up with novel ways to use technology for a gain.

Now, patterns in their normal use might have weak predictive power, pretty much indicating 50:50 on market direction one bar afterward. However, at any given moment, numerous patterns are in effect across multiple time-frames and these patterns could be stacked like Tetris blocks, as one good customer calls it, and their combined directional probabilities can be viewed over numerous future bars (or quotes for real-timers) to give yet another view of market direction.

Better yet, rather than applying your typical human-defined cute-named patterns ("Dead cat bounce", "Horned bottom", etc), in our labs we have an auto-clustering adaptive pattern building technology. This technology adaptively clusters, builds and stores real patterns found in the data by taking a data sample, comparing it to a collection of patterns, selects one or more patterns based on similarity then adapts the patterns. If the similarity is too distant, it creates a new pattern. The adaptation algorithm is similar to what you find in Self-Organizing Maps or some image processing. The patterns can be frozen or continue to adapt (cluster) as time goes forward. You get a unique set of patterns that that actually occur in the data with control over how many there are (generalization) as well as the predictive power of each, over time.

Let's take that one step further... the patterns do not have to be on price, as normally done, but could be built on indicators, so patterns of indicator movement can be related to price movement.

The pattern building itself is based on data similarity and not equity so it cannot "over-fit" the data in the sense of curve-fitting data on a performance basis. The use of adaptive equity performance comes later, when we use the patterns...

It would be interesting to look at various sizes of patterns (lengths in bars, or number of quotes), then combine their estimate of the probability of the rise or fall of the markets for multiple bars into the future to create a merged probabilistic "signal", perhaps equity performance weighted. OK, I'll stop there for now. That gives you the idea, I hope.

One reason I think this may work well is because of the sneaky shifting patterns I see in the SP500 futures. There's a variable-length heart-beat in there that might be hard to quantify "in code" or in data modeling. If we have an adaptive pattern that picks up and synchronizes with that, we have money in our collective pockets. Other technologies may have a challenge picking up such a pattern, but this one is built for it.

Maybe it's time to reapproach patterns, with "Patterns". In our forums, there was quite a lot of encouraging discussion from our customers.

What say you?
Email me at cmcook@biocompsystems.com

Carl Cook
President and Chief Technical Officer

Announcements

Want Dakota, Profit, Add-Ins? Modeling Tools? EMS III?
Go Get 'Em !
Grab this 25 Percent Stimulus Deal of our Own!

The earlier store-based code thing worked quite well for all of us. So now you can use the Code ADAPT in the cart in the store for a 25 pct reduction on anything and everything we sell in there... workshops, Dakota, Profit, modeling tools, the NGO, EMS III, anything... but it is for a limited time, and we're not saying how long this is good for this time either, so get tooled-up for Fall and the coming Winter.

Profit 8 Workshop: Houston, October 23, 24, 25, 2009
The Profit 8 workshop in Houston, Texas will be on the above dates at a hotel in the Houston Galleria area. It is three days of intense 1-on-1 brain transfer with Carl Cook, the creator of Profit. Seats are filling, with some yet remaining.
Interested? Make the most of your investment. Sign-up >>

My Pet "Swarm" Achieves Another Equity High

My "Double-Smoothed Z-Score, Median of Signals, 10 yr History, One Day Ahead" trading system clocked in another equity high on Sept 28th at 2966 points. In the last newsletter edition it was at a then-new-high of 2,942 points. In the last few days, with the bad economic news, it's stepped back from the highest equity, but within the normal "jiggle" of the equity curve.

Dakota users can download this trading system (bot) from our forums and it is very easy to setup... click, click... No coding required.

And here is a screen grab of the price, signal and equity curve of the system recently.

Stay tuned...