Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Gold/Mining/Energy : Daytrading Canadian stocks in Realtime -- Ignore unavailable to you. Want to Upgrade?

To: Vitalsigns who wrote (21653)	9/21/1999 4:17:00 PM
From: The Osprey	Respond to of 62348

Holding my COL.T overnight......Almost like saying I am taking a damn book to bed with me....Hope(another 4 letter word) it is better to-morrow and markets bounce back big time. Later all. The Osprey PS-Forgot to mention I am in COL.T at 15.50,15.00 and 14.95....

To: Vitalsigns who wrote (21653)	9/22/1999 12:53:00 AM
From: Cameron	Read Replies (3) \| Respond to of 62348

(I just finished rereading this message and realized how long it is... get a coffee before you start to read it <gggg>) The little nanos have been breeding away like rabbits and they've come up with the following crystal ball projections for tomorrow: TSE 300 Index - 6900 vs. close of 6917 MCF - $5.50 +/- $.20 vs $6.65 close- yikes!! And I'm long!!! ATY - $17.70 +/- $.40 vs. $16.15 close - I hope it is true!! I'm in at $16.75. IMO - $30.60 +/- $.30 vs. $30.30 close - this is the first time reporting on this one. Number of generations are still quite low but I guess that IMO investors are more predictable than iNut investors?? I've never traded IMO but thought that a few non-tech stocks would be interesting to track. Anyhow, regardless of the specific number, the call would be market flat, ATY up a chunk, MCF down a chunk. They predicted the same for MCF last Thursday and Friday however and were wrong!! The one thing I really want to reiterate is that from my point of view this is really an experiment and/or game and hopefully others will view it this way as well. These little guys can't predict the future. They had no way to knowing about the US trade deficit which hit today so couldn't possibly be expected to predict the decline. If the deficit had been expected and rumoured about, it would have been reflected in trading patterns leading up to today and I'd like to think they'd might pick that up but we'll never know and in any case, it was as much of a surprise to the market as it was to the nanos. If they can do anything, they can hopefully predict a price potential, all else being equal. They only look at historical price and volume patterns (I think of it as cyber tech analysis) although they can, at least theoretically, discern patterns far beyond my simple mental capabilities... so... in brief... they suffer all the same short comings that TA does. To be explicit, one PR from MCF and the nano prediction becomes totally obsolete!! (I had to say that just to make Vital relax). The little guys/gals (actually they're hermaphrodites!!) are adding up cycles on CRY, NNC, COR, CIBC(CM) and TBC.A but the standard errors are still quite large/number of generations still quite small. Should be able to bring them up in a few more days though. Priority (foreground processing) is on NNC.... currently at $40.00 +/- $2.40 after 248 generations (things get stable after about 4,000). How's that for a range!! (This is the part that gets a bit long and tedious) Just to answer a couple questions publicly that I've received from more than 1 individual re. the nanos. a) The nanos are indeed based on genetic programming principles. This style of programming has proven to be highly effective at solving problems where the structure of the question (ie. linear/non-linear) and relevant variables are unknown. The primary weakness of genetic programming however is that a population can get trapped at a local optimum.. ie. there is no guarantee that it has found the "best" solution. My hope is that by running the nanos against a large cross sections of stocks for many, many generations, that a common structure will develop in terms of which variables are relevant and how they inter-relate. Once this is discerned, a neural net could then be built which would in fact guarantee an optimal solution for the pre-defined structure. The latter phase would be months down the road however. b) The system is written in C++ by yours truly. Without providing a resume, I have worked in both systems and marketing research at different times along the way, as well as attending a number of advanced analytics conferences on stuff like neural nets, etc. I also took tons of the stuff in school. I think I was probably Toronto's first commercial C programmer back in 1983/84 when I wrote a linear programming system for the City of Toronto as a consultant. Now it is just a hobby however.. I know... I need a life!! In any event, I have worked with this type of thing quite a bit, although never with an investment perspective before. c) re. giving out copies of the program. The nanos ARE an experiment or "academic exercise" at this point and to be honest, the interface is a bit of a mess and I naturally haven't written a manual or anything. Some of the menu items aren't hooked up, etc. Also... the alterior motive.. if it does in fact work I may decide to sell it to a broker, write a book about the TRUE mechanics behind TA or who knows what and make my millions that way... in which case, having copies floating around wouldn't be prudent. Lastly and to me right now, most importantly, based on back casting, they seem to track well on a limited number of stocks.. but to be honest, I don't know whether this is a viable trading tool or not, or whether it has the potential to become one. Therefore, I think the nanos provide a "fun" focus for chat on the thread but I sure don't want anyone using them seriously at this point. Providing copies implies they might be used this way and I don't want to be responsible for someone losing a fortune. Given all of the above, I'm going to keep them close to home for now. I'll run them on any stocks requested however as long as there is a 2 or 3 year history without any stock splits - although there are only so many I can get up and running at one time and it takes several days of running before a population starts to approach an optimum standard error. Tomorrow should be ... interesting!!! See you all then.