Comments on: Apple Inc, (AAPL) Neural Net Model

By: Tom

Tom — Fri, 02 May 2008 00:24:21 +0000

dc: thanks for your support!

By: dc

dc — Tue, 29 Apr 2008 18:05:55 +0000

one more thing
don’t nit pick if you don’t have anything better to offer PERIOD!
WE WILL BE MORE RECEPTIVE TO INFORMATIVE CRITIIQUE THEN JUST NIT PICKING!
“GO HOME TEAM, YEYYYYYY” my bet is riding on the home team, and by the end of the year into next, i think we should see results.
“CONTRIBUTING IS THE NAME OF THE GAME NOT NIT PICKING”

By: dc

dc — Tue, 29 Apr 2008 17:49:07 +0000

tom
i wouldn’t worry, i think he haid no game so he left
GO HOME TEAM!

By: Tom

Tom — Tue, 29 Apr 2008 12:33:46 +0000

haid: I dont want you to go away, you make valid points. I used to be a simple Buy and Hold investor, and it doing that works great in Bull markets. In Bear markets you watch your gains evaporate. It wasn’t until I went to MBA school that I did quite a bit of personal research into the markets and studied some of the great masters (i.e. Buffet). They do use a Buy and Hold strategy on the surface but what they really do is time the markets. They wait for the market participants to overshoot prices (up or down) due to fear or greed and then go long or short.

Their buy and hold time frame is years, mine is mostly weeks or days.

I look at this way, never pay retail but buy wholesale. I do make some mistakes (if you read through the blog you’ll find them) in trading but that’s why you should always risk management.

Those returns are from actual Forex trades (you can see the ticket #’s) where I did time the market using neural net models. I didn’t trade all the time, as the dates bear out, and I had some clunker trades in there as well. I posted that information because of the skepticism that readers might have when I make a claim like, “I made over 50% trading Forex last year.”

Feel free to bust my chops anytime and thanks for reading.

By: haid

haid — Tue, 29 Apr 2008 11:34:46 +0000

I’m not here to bust your chops. Experience and training are worth something though (a point my teenage son would dispute).

If you want to compare yourself against non-investing, you will do well in a bull trend. As a scientist, it is not very useful. What is wrong with a 4% improvement over random? It would be better than the average fund manager’s performance. When you claim astronomic returns (22% to 30% above), you invite skepticism. If you don’t want such points made, don’t accept feedback.

I’ll go away now.

By: Tom

Tom — Tue, 29 Apr 2008 10:07:27 +0000

haid: If you want to read about my success using neural nets to trade, check out my year Forex year end results @ http://www.neuralmarkettrends.com/2007/12/29/forex-results-for-2007/

By: dc

dc — Tue, 29 Apr 2008 05:25:35 +0000

hey haid don’t be such a haid-er
this neural net product is just something to practice on before the real thing gets under way. In about a year or so,if all goes well we should get the real thing going.
But, if you have something better, lets hear it. We want all the details and white papers and abstracts to download,and if we find them interesting we will be full of questions with emails back and forth for quite a while.WE ARE READY TO BE TOUGHT SOMETHING BY YOU IF YOU CHOOSE TO BESTOW UPON US YOUR GREAT KNOWLEDGE.

By: haid

haid — Tue, 29 Apr 2008 04:08:01 +0000

Tom,

I’ve studied AI as a Ph.D. student and have 20+ years experience. I know its easy to run new experiments. Presumably the point to make money anticipating the future, not showing how cool it is to mimic the past. And obviously in the future you don’t know whether it is going to be a bull, bear, peak, or valley. So if you don’t train on a wide range of market behavior, it is unlikely that you will do well over the long term.

The only point about the buy and hold strategy is to be used as a comparison to see if your training is performing better than random (which should approximate buy and hold). At the beginning of your chart AAPL is around 135, at the end its around 170. That would be a return of about 26%.

If you want to trumpet your success, it should be about making an extra 4%.

By: Tom

Tom — Mon, 28 Apr 2008 23:54:10 +0000

haid: its quite easy to use different different test sets for different time spans, just as easily as it would be to retrain the model with new data when market conditions change.

Buy and hold is a strategy that works for some but not for me, carefully timed purchases can make the difference between having a hum drum rate of return and a spectacular return. I’m not advocating that my strategy is better than yours, it just depends on the individual.

FWIW here are the signals the model would’ve generated during the Feb-March neutral time. As you can see, the signals would’ve reaped you a gross return of 30%.

http://www.neuralmarkettrends.com/wp-content/uploads/Common/snm-AAPL-042508-2.jpg

By: haid

haid — Mon, 28 Apr 2008 23:09:07 +0000

Its not about the stock. You can use AAPL. But choose different types of training sets. Choose a time when the stock is neutral but overvalued (november 8 to Jan 8).
Choose a time when it is neutral but on the bottom (feb through March).
Choose a time when the stock is falling (January).
And of course, your time when the stock is rising.

If your learning system can perform better than the overall “buy and hold” strategy for these scenarios and it can determine when it is in one mode or another then you are on the right path.

This all assumes that the market behaves in patterns; otherwise the recognition of patterns is unuseful.