anticitrade (99.58)

anticitrade's CAPS Blog

Can a consumer sentiment based valuation be automated?

Recs

14

Report

November 04, 2010 – Comments (7) | RELATED TICKERS: CHL , MON , PBR

For the last month or so I have been devoting all my free time to the development of an idea of using specific consumer sentiment to predict price movement for specific stocks. I now have a working model that correlates sentiment individually and specifically to any given stock price.

I evaluated the top 100 largest cap companies using my methodology and then created a portfolio based on the directional call for any company with correlation coefficient above .5 (~20 stocks). The new portfolio is called "anticiment".

I have spent a lot of evenings building this and sleepless nights thinking about it, so I am not planning on divulging my methodology. I think it is a right of passage to suffer "excel nightmares" (where you wake up every few hours from a dream where excel formulas weren't quite working and you become lost in a world of vast sheets of data) before you get a tool like this (or you pay me a lot of money, either one is fine with me).

Anyway, I will use my "anticiment" profile for testing the tool. You should know that my forecast window is about 3 weeks.

The picks are:

CHL      +
MON     +
PBR      +
PFE       -
MO        -
AAPL     -
MCD       -
ABX        -
RIMM      +
GE           +
VALE       -
TM           +
GOOG      -
IBM         -
XOM       -
ABT         -
CVX        -
UTX        -
ABB        -
CVS        -
UN           -

(+/- indicates expected direction.... It is possible that my system has a negative bias that needs to be explored)

7 Comments – Post Your Own

#1) On November 04, 2010 at 12:47 PM, dbjella (< 20) wrote:

Have you tried use MS Access? Excel can get awfully difficult to maintain when functions change. Data imports can get tricky as well. Just curious.

Report this comment

#2) On November 04, 2010 at 1:18 PM, Valyooo (99.82) wrote:

Very interesting and I wish you the best of luck. I like your pbr call. But I can't see goog and ibm going down. Guess we shall see

Ps are these absolute, or compared to the market?

Report this comment

#3) On November 04, 2010 at 2:58 PM, anticitrade (99.58) wrote:

dbjella,

I haven't messed around with MS access.. Sounds like something I should investigate.

Valyooo

I agree with you about your thoughts on GOOG and IBM. Since it's not based on my sentiment, I just pass through what the numbers say. The numbers are absolute. I have also tried adjusting the numbers for movement in the S&P500 (looking at alpha), that will probably be a subsequent test.

I should also mention that the above ticker calls are ranked in order of r^2 values... With the best fits being found at the top.

Report this comment

#4) On November 06, 2010 at 12:48 PM, moneyming (< 20) wrote:

Anticitrade, your sleepless nights in developing these awesome systems are very inspiring to me =)

Can you explain a little more on how you measure consumer sentiments? Specifically, what is the original raw data that you can pull from about these companies?

Report this comment

#5) On November 10, 2010 at 11:19 AM, anchak (99.84) wrote:

How do you define sentiment on a stock - Short position?

BTW : Definitely use Access if you are stuck with Microsoft Office. And keep the models coded in VB - that way - they are much more transparent and auditable for later

Report this comment

#6) On November 10, 2010 at 11:37 AM, anticitrade (99.58) wrote:

Moneyming and anchak,

I define sentiment as an aggregate interest in a specific product, sector, industry, company or action.

For example, by tracking peoples interest in painting a home, I can get some idea how many cans of paint may sale over the next month.

Revealing my data source would make this all much more meaningful and interesting, but I am not ready to divulge that piece of the puzzle yet.

Thanks for the advice anchak, I will learn more about Access.

Report this comment

#7) On November 12, 2010 at 12:52 PM, mechincl (46.96) wrote:

You know, it might not be that your sentiment system has a negative bias so much as it is picking up on some potential downward market risk. I'd say there's a good bit of short term downside risk at the moment. The market has gone up quite a bit over the last couple months without a breather.

In looking at similar ideas myself, I've wondered if you could take such data from an aggregate perspective, and determine potential for total market direction. If you end up with a good quantity of stocks where historically the sentiment has been fairly accurate in predicting movement, and all of a sudden 80% of the stocks are popping downward flags via the sentiment, that might be a very useful risk indicator in my opinion.

Report this comment