| View previous topic :: View next topic |
| Author |
Message |
Doly
Expert
Joined: Dec 03, 2004
Posts: 4041
|
 Posted: Wed Jul 20, 2005 8:41 am Post subject: |
  |
|
| If I get your model right, your table includes little more than a year. The most likely periodicity is quite clearly annual. Wouldn't your model improve if you took into account more years? |
|
| Back to top |
|
 |
khebab
Moderator
![]()
Joined: Sep 27, 2004
Posts: 935
Location: Canada
|
| Posted: Wed Jul 20, 2005 9:28 am Post subject: |
|
|
| BabyPeanut wrote: | | BabyPeanut wrote: | Based on your chart the price will come down in August. Fat Chance!  |
Hey it's the second half of July and the price is coming down. |
Well, the model predicted a price crash down to $52 around July 12th but it stayed around $58 instead! I guess the hurricanes (Dennis, Emily) kept the prices up.
_________________
______________________________________
http://GraphOilogy.blogspot.com |
|
| Back to top |
|
|
khebab
Moderator
![]()
Joined: Sep 27, 2004
Posts: 935
Location: Canada
|
| Posted: Wed Jul 20, 2005 9:34 am Post subject: |
|
|
| Doly wrote: | | If I get your model right, your table includes little more than a year. The most likely periodicity is quite clearly annual. Wouldn't your model improve if you took into account more years? |
Yes, I could include more years. I decided to take only the data since Sept, 02, 2003 because of the stability of the linear trend for this period. Because price volatility has increased since Mar 30, 2004 I finally retained the last 208 business days. The difficulty is to isolate a data section which has a stationnary model for the price trend. The periodicity transform is then applied only on the residuals centered around 0 to capture the periodic fluctuations.
_________________
______________________________________
http://GraphOilogy.blogspot.com |
|
| Back to top |
|
|
khebab
Moderator
![]()
Joined: Sep 27, 2004
Posts: 935
Location: Canada
|
| Posted: Mon Aug 08, 2005 11:16 am Post subject: |
|
|
Ok, time to check the validity of the forecast one month after it was issued. The new data points are in red and the predicted price is the blue curve.
I was predicting a steep fall but hurricane Dennis (Landfall on July, 10) kept the price high. However since July, 14 prices are on steep rise wich seems to follow the prediction. The overall correlation coefficient is 27%.
edit (08/10): update with new data points up to Aug, 10.
_________________
______________________________________
http://GraphOilogy.blogspot.com
Last edited by khebab on Wed Aug 10, 2005 3:08 pm; edited 1 time in total |
|
| Back to top |
|
|
pup55
Expert
Joined: May 26, 2004
Posts: 3853
|
| Posted: Wed Aug 10, 2005 2:24 pm Post subject: |
|
|
This model has correctly predicted this week's pricing excitement!.
It looks as though the peak will be about August 24th. |
|
| Back to top |
|
|
khebab
Moderator
![]()
Joined: Sep 27, 2004
Posts: 935
Location: Canada
|
| Posted: Wed Aug 10, 2005 3:16 pm Post subject: |
|
|
| pup55 wrote: | This model has correctly predicted this week's pricing excitement!.
It looks as though the peak will be about August 24th. |
I just updated the figure above with new data points up to today. Prices should collapsed around the end of August below $55! but I don't think it will, I have the feeling the trend (the blue straight line) is going exponential!
_________________
______________________________________
http://GraphOilogy.blogspot.com |
|
| Back to top |
|
|
khebab
Moderator
![]()
Joined: Sep 27, 2004
Posts: 935
Location: Canada
|
| Posted: Wed Aug 24, 2005 11:14 am Post subject: Re: Price Prediction Based on the Periodicity Transform |
|
|
Here a scatter plot, of the predicted prices versus the actual prices between June, 28th and Aug, 23rd.
The prediction is rather good but we have to wait till more data are available.
_________________
______________________________________
http://GraphOilogy.blogspot.com
Last edited by khebab on Wed Aug 24, 2005 2:20 pm; edited 1 time in total |
|
| Back to top |
|
|
pup55
Expert
Joined: May 26, 2004
Posts: 3853
|
| Posted: Wed Aug 24, 2005 11:50 am Post subject: Re: Price Prediction Based on the Periodicity Transform |
|
|
Once again, this is brilliant!
The question now is: will there be a little market correction in here for a couple of weeks? |
|
| Back to top |
|
|
khebab
Moderator
![]()
Joined: Sep 27, 2004
Posts: 935
Location: Canada
|
| Posted: Wed Aug 24, 2005 2:28 pm Post subject: Re: Price Prediction Based on the Periodicity Transform |
|
|
| pup55 wrote: | Once again, this is brilliant!
The question now is: will there be a little market correction in here for a couple of weeks? |
Thanks! I think a big correction is highly probable because prices have stayed above $60 for almost 24 days now! If price volatility is mainly due to speculation I think we can expect a drop but any disuptions (due to an Hurricane for example) could push prices higher.
_________________
______________________________________
http://GraphOilogy.blogspot.com |
|
| Back to top |
|
|
khebab
Moderator
![]()
Joined: Sep 27, 2004
Posts: 935
Location: Canada
|
| Posted: Sun Nov 27, 2005 3:25 pm Post subject: Re: Price Prediction Based on the Periodicity Transform |
|
|
It has been a while I didn't look at this model. Prices have remained above the trend longer than expected, I guess because of the two major hurricanes:
The correlation is still rather good:
_________________
______________________________________
http://GraphOilogy.blogspot.com |
|
| Back to top |
|
|
|
News
Discussions
Resources
Members
Forum FAQ
Search
Usergroups
Profile
Log in to check your private messages
Log in
