Patterns and aggregated data as basis for Hidden Markov Models and Bayesian networks

Jim Simons IS the standard for Hedge Funds today. You wouldn't believe some of the amazing models I've seen, and the techniques will never see the light of day, let alone here.
http://www.olsen.ch/center/papers/hedgefuture.pdf

 

First Turtletrader.com is a craps which has been denounced by the true turtletraders.

Secondly I don't see about what one should be amazed ? Renaissance is big and that proves what ? That they have been very good at marketing as for the rest it's much like LTCM like: triviality.

 

I feel embarrassed now! I have never heard about Jim Simpson! But what a concept!!!! Could you be so kind and send me some more links and whatever you have on this subject! Most fascinating stuff! Thank you so much!

 

<IMG SRC=http://www.originalturtles.org/images/pageheadings/scams.gif>
http://www.originalturtles.org/scams.htm

One of the sad realities of the trading industry and futures trading in particular is that there are far more people making money selling systems and “ways to make money trading” than there are people actually making money by trading.

There are many “famous traders” who don’t make money as traders. They make money selling new trading systems, seminars, home study courses, etc. Many, if not all, of these so called “experts” can’t trade and don’t trade the systems that they sell.

Yes, this is also true of those selling the Turtle Trading Rules. Consider the major sellers: the first, a web site, TurtleTrader.com, and the second, a former turtle. Here’s what they won’t tell you:

<B><font color=orange>TurtleTrader.com</font></B> - A web site run mainly by one guy (an admittedly talented web marketer that also has a pharmacy site and a site that sells personality tests), TurtleTrader.com purports to have the actual Turtle Trading Rules and sells them to you for about $1,000. The site is filled with huge amounts of information about trading and bills itself as the “No. 1 Source for Trend Following Worldwide”.

What they don’t tell you is that the site is run by a guy who doesn’t even trade his own rules — or trade at all for that matter — has never been a successful trader, yet purports to be an expert on the “Turtle Trading Rules” and trend following.

His rules, while reasonably close to the actual rules, differ in material ways from those taught by Richard Dennis.

You won’t get any expert advice from the guy who runs TurtleTrader.com. All you will get is the regurgitation of advice from other traders not tempered by the experience of a successful trading career. Paying for advice from this source is a lot like hiring a blind guide.

In the final analysis, TurtleTrader.com is not any better than the other scams and system selling hucksters he warns about. It is a site run by a guy who appears to be more interested in taking his customers money than he is in their success with the system he sells; someone who misrepresents himself as an expert in trend following without mentioning that he doesn’t trade.

 

this is kind of old but this is what got me started on the direction I have chosen http://www.charttricks.com/Articles/ren_tec.pdf

 

if a pattern is what you are looking for, than a pattern is what you will find - even if there isn't one there. the implications thereof i leave to the Popperists amongst us.

 

Didn't you say only 5 minutes ago : "There are no patterns in the Time/Price binary sets Get it?"
Well you are very prompt to change then within 1/4th minute

 

Thanks for the most interesting post. I worked with Bayesian Probability Models many years ago, not because I wanted to, but because I had to. It was THE one course you had to pass in the MS for Operations Research at the College I went to (Union College) and it was a ball buster.

Since my company paid for the school I also had to apply it to the manufacturing processes of heavy duty gas turbines we built in the same town (Schenectady) and I got more experience with it than I had planed.

To summarize - I admire your knowledge and determination, but - I would not try and apply Bayesian Probability Theory to financial markets. One of the readers put it in a capsule - there are no repeatable patterns you can identify with it. There are patterns for sure, but they are much too complex for mathematical solutions like this. Observation and a good eye for patterns ( intelligence helps on occasion) is still the best method.
As for Turtle Traders, don't even bother with it - it's not of any use to 99.999% of us mortals.

Best of luck....

 

which post? on second though, don't answer that

 

Thanks for your views on using AI, but I should point out that I'm not going to do some fancy price predictions, or even trading predictions (at least not yet).

Normally I use tape reading and some classical technical indicators along with a lot of news reading/watching and trying to get a "sense" of the mood in the market by watching reactions to trades happening and technical indicators (which seldom seem to be a firm guidance with the highly volatile ES futures).

When I trade, I get some kind of "pattern" or signal for my trading, I guess, and I also witness trades which happen so fast in reaction to signals/other trades that they have to be automated. It's also a fact that many try to use automated trading - probably with varying success.

I have some background in using rules-based and fuzzy logic from the mid 1980's, and then some NNs with genetic algorithms from 1996 onwards (but for accurately predicting hourly changes 24h ahead for energy markets [my models did that at over 90% success-rate too - around +80% yearly average - that was actual predictions] and other data mining projects).

The "holy grail" as we all know would be having some super-AI system being able to do all kinds of intraday markets; but in my view such a system would encourage changes/adaptation to these kinds of systems in the markets where market players would try to distort signals on purpose to skew markets into their favor. This is the case nowadays as well to some extent.

What I'm looking for is patterns that I use nowadays, and perhaps trying to find some new patterns for me to trade on. Then if I trust a automated system which could use a mix of proprietary generated indicators, I would be willing to try it - after some serious testing.

However, the markets are constantly in flux, so there will always be a need to adapt the models and indicators that used to work, probably don't any more. That's evolution, and I think I'll handle that manually.


So, as long as you understand what I'm trying to do, and what I'm not trying to do, I think trying to identify patterns that are currently working in the markets is not such a futile idea/project. I use my good eye to watch the tape and the other to scan for other news.

Well, I just came back from the beach, and have go out eat some sushi now. I'll answer more about Kohonen Self-Organizing Maps etc. later, and also why I'm interested in HMM and Bayesians (which are not normally good for generalizing - but that's ok in my case I guess).

Still, if anyone are actually using some AI techniques - either standard or proprietary - I'm still interested in hearing about it. I know most of the pitfalls about AI, since I've been pursuing this interest for some time, and have some successes with applying it.

AI also works for things like speech-recognition, NLS and other pattern recognition cases, I'm just not looking for the straightforward patterns in price, volume etc.