Reverse engineering "Hive Technology"

Hi,

Something that caught my eye recently was wave59's "hive".

There are downloadable videos for this at http://www.wave59.com/webinar_archives.asp. The March 14th 2010 webinar is the most descriptive.

The first time I watched these videos a few weeks ago I thought that the "hive" was the result of a weighted vote of a huge amount of technical indicators. I also thought that the weighting for each indicator's vote was determined via a genetic algorithm optimising for max profit over a lookback period.

But now I watch the video again and notice a completely different approach- it looks like they're finding the best indicators over a range of different lookback periods, then doing the exact opposite of what those previously optimal indicators want you to.

There's apparently an artificial intelligence step after this, that i suspect to be a genetic algorithm, although I dont know for sure.

I'd love to discuss this with other people, we may be able to replicate the hive or even make something better!

If you have any questions regarding what I've already worked out, please ask away.

There is no march 14th webinar, did you mean may?

Hive? artificial intelligence? genetic algorithm? are we still talking about trading? what are you talkin' about son?

Interesting hive algo: http://www.scholarpedia.org/article/...lony_algorithm

Interesting, I haven't heard of this company before.
but...

I have some worries....
I can't find any performance for this "private fund" that Earik Beann, creator of Wave59, supposedly manages. That bothers me a little.

It all reads very hi-tech...the stuff on the website, but....

Speaking topics at a recent conference:

Intraday Astronumerology

Etheric Planets - The Key to Unscrambling Modern Astrology
-- "Learn about the amazing Lunar cycle, examine profitable backtest
-- results..."

Roulette Systems and Card Cycles
-- "Believe it or not, playing cards actually represent a solar calendar,
-- and can actually be used to predict events in the future in the same
-- way as modern astrology."

...So these excerpts from his presentations taken from a google search of him speaking at a conference in Vegas last year scare me a little. (link below)
http://www.wave59.com/poweruser09.asp


I do love to theorize about the computer science aspects of AI, genetic algos, and such...

I think this particular guy is a nut-case unfortunately.
Just my 2 cents.

More hive literature: http://conference.iproms.org/sites/c.../PID217625.pdf

I'm with intu on this, I suspect this is all just more "fancyware".

MIT can't gaurantee,"follow through," on an entry signal. Inherent limiting factors allow for just so much certainty. The upper limits of certainty are approachable by many of the simplest methods. Is market going down? Up?. Just look at your screen!! Buy when mkt. is going up. Opp. for sell. Smoke...good!! Friend...good!! Wine...very good!!!

It's the March video, sorry I get mixed up!

I don't think he's using a bee's colony algorithm, they're not suited at all for a task like this.

I think the name "hive" refers to a hive mind (in simple terms a vote of a big number of systems). My current guess is that each vote in the hive is weighted by a genetic algorithm.

Basically you start off with an array of voting weights, then randomly crossover the best performing weighting lists to make a new generation of better voters, repeat until performance is acceptable.

on a cave wall.

More hive 'splanations: http://www.wave59.com/About/GeniusFe...echnology.aspx

Looking at how fast those performance reports were generated, I can't beleive it's a true backtest.

I think he's using a cluster, even with parrallel code my version isnt so quick

to calculate vs the ability to think is the first step on the path to enlightened trading...I consider my Promethean mission accomplished. Have a nice day.

I always wonder about companies that seem to be able to give you the keys to trading paradise for the ultra low price under $300/month. Plus you do have to buy the book at $459, I'm sure it's a good read and all.

Personally if I can't understand the concept fairly quickly it's too hard for me to trade.

Be interested in seeing what you guys think about it.

heh, this looks like a joke to me.

Oh well, I understand that some people aren't interested in this but I personally am.

Im not interested in astrology but this genetic stuff is interesting at least.

I've coded my own genetic systems and get roughly 65-75% accuracy with it. With stop losses it's not a bad trading system, although I admit its complicated and I still have a list of extra things to code into it.

Effectively it watches a pool of about 1200 indicators, and pays more attention to the ones that made profits in a given lookback period and less attention to non-profitable systems.

Using my genetic algorithm to se voting weights makes more profit than the best single indicator would do over the same period.

The thing about genetic algorithms is that they don't get trapped in local optima, and hence are able to search a massive solution space without getting trapped in a single condition.

The problem with traditional genetic systems is they are fairly inefficient when compared to some variations of evolving networks.

What other types of evolving newtorks would you recomend?

If there's anything else as good at escaping local optima i'd love to know about them