userque & NUGT/DUST

Sort by newest Sort by date
I for one like the additional information. Only time will tell how useful this information is and how it should be used.

userque said:
Right. I trade the forecasted direction. The price forecast and confidence levels are a byproduct the system produces. I thought it would be helpful and interesting for some folks, to reveal a little about how the system is "thinking."

They can be ignored by other folks.

I may stop posting them if they are not desired; if folks prefer less information about a closed system, as that would mean a couple fewer steps for me.

Note also that I've never made a system that's 100% correct. Sooooo it'll be wrong sometimes. The squeamish may want to paper trade NUGT/DUST using their own devices... or at least: something different. :)







_______________________________________________________________
DISCLAIMER: Userque's trading signals are not affiliated with TSP Talk or Buy Low Sell High, Inc. The information is for educational purposes only! The information is not advice or recommendations. The information may be revised at anytime. Userque does not give investment advice. Do not act on this data. Do not buy, sell or trade any stocks, ETF's, indices, funds, or any other instruments/vehicles based on this information. Userque may trade differently than discussed or posted in this forum.
Click to expand...
 
Right!

It matches recent price action patterns to historical data. It decides how to go about it (how many matches, how many days to match, etc.). If all of the matches are up, for example, it'll be 100% confident that the forecasted day will be up. Again, it's just a measure; not a quantum phenomenon :)

The percentage change forecast is merely the average of the percentage changes of the historical days the system found and used.

RealMoneyIssues said:
Confidence in a correct call?
50/50 GDX up or down? Interesting :)


Sent from my (Daughter forcing me to use an) iPhone using Tapatalk...
Click to expand...
 
Like Steiner said, there may be many ways to use data, for example:

If, historically, the price action would result in a down day of 2.28%, but today, it was only down, 0.53%, then one can conclude that today is more bullish than usual, and further, that tomorrow may be more bullish than expected by the system.

This is just an example of an untested theory. I'm just illustrating how creative one can be with simple data...if one tries.
 
Last edited:
Q,

I just want to make certain I am interpreting these numbers correctly. Confidence level = 50% means it is a tie between both NUGT and DUST in regards to which one will close up tomorrow. However, since your system is now trying to be the most profitable as opposed to the most correct, it is choosing NUGT since it has the larger of the two potential % moves up at close. Is my thinking correct?


userque said:
FOR 11-03-15 AT CLOSE:
ENTER NUGT

CONFIDENCE LEVEL: 50%

GDX FORECAST TO CLOSE UP CIRCA 0.54% FOLLOWING TRADING DAY




_______________________________________________________________
DISCLAIMER: Userque's trading signals are not affiliated with TSP Talk or Buy Low Sell High, Inc. The information is for educational purposes only! The information is not advice or recommendations. The information may be revised at anytime. Userque does not give investment advice. Do not act on this data. Do not buy, sell or trade any stocks, ETF's, indices, funds, or any other instruments/vehicles based on this information. Userque may trade differently than discussed or posted in this forum.
Click to expand...
 
Close...I think. :)

50% means that 50% of the historical GDX patterns chosen by the system went in the same direction the system is picking.

The system uses the average of the chosen patterns to pick NUGT, or DUST. Positive, it picks NUGT. Negative, it picks DUST.

For example...

Steiner said:
Q,

I just want to make certain I am interpreting these numbers correctly. Confidence level = 50% means it is a tie between both NUGT and DUST in regards to which one will close up tomorrow. However, since your system is now trying to be the most profitable as opposed to the most correct, it is choosing NUGT since it has the larger of the two potential % moves up at close. Is my thinking correct?
Click to expand...
 
... it decides to use two historical GDX patterns. The day following the first pattern was up 1%. The day following the second pattern was down 2%. The average is -0.5%. The system picks DUST. The output is:

DUST
Confidence: 50%
Percent Change: down 0.5%
 
Looks like you had it right [expect that it only uses GDX data, not NUGT/DUST data]. Just that your wording is specific to an example of only two patterns.

If it chose three patterns, NUGT or DUST would still be chosen by the average percentage change of the three. And whatever percentage of the three were going in the same direction as the average; that would be the confidence percentage.
 
Last edited:
I should search your account talk for this - but, if not proprietary, what software do you use?

Regarding #3 Do I understand your system would predict/estimate whether the close would be higher or lower based on the Open?
 
No problem. Search is difficult to use sometimes.

I use a custom, built from scratch, kNN (k-Nearest Neighbors) machine learning algorithm, in excel w/VBA.

The OPEN is the most recent data it would use. It can use the OPEN, HIGH, LOW, and CLOSE from previous days as well.

I originally built my own to have a custom cost function. That's where the magic happens. I also control the input weights and how it searches for the best parameters. More magic. :)

None of the off-the-shelf machine learning algorithms I've tried seemed to work adequately.

cully said:
I should search your account talk for this - but, if not proprietary, what software do you use?

Regarding #3 Do I understand your system would predict/estimate whether the close would be higher or lower based on the Open?
Click to expand...
 
Last edited:
FOR 11-04-15 AT CLOSE:
HOLD NUGT

SYSTEM BYPRODUCT DATA.
This is data from the system, but not necessarily optimized by the system. Some may find it interesting; if not, please ignore it.

CONFIDENCE RATIO: 2:0
* A ratio of 2:0 means 2 historical patterns match the same direction as the current CLOSING forecast (NUGT, LONG, UP), and 0 don't match (are opposite); which is 100% (2 out of 2 match). The ratio provides more information than just the percentage. The ratio indicates how many historical patterns the system used to draw its conclusions. In this case, 2.

GDX FORECAST TO CLOSE UP CIRCA 1.69% TODAY
GDX FORECAST TO CLOSE UP CIRCA 0.54% FOLLOWING TRADING DAY




_______________________________________________________________
DISCLAIMER: Userque's trading signals are not affiliated with TSP Talk or Buy Low Sell High, Inc. The information is for educational purposes only! The information is not advice or recommendations. The information may be revised at anytime. Userque does not give investment advice. Do not act on this data. Do not buy, sell or trade any stocks, ETF's, indices, funds, or any other instruments/vehicles based on this information. Userque may trade differently than discussed or posted in this forum.
 
Last edited:
1 - Are you tracking forecast vs. actual
2 - Can the system learn from this data

Thanks

userque said:
FOR 11-04-15 AT CLOSE:
HOLD NUGT

SYSTEM BYPRODUCT DATA.
This is data from the system, but not necessarily optimized by the system. Some my find it interesting; if not, please ignore it.

CONFIDENCE RATIO: 2:2
* A ratio of 2:2 means 2 out of 2; which is 100%. The ratio provides more information than just the percentage. The ratio indicates how many historical patterns the system used to draw its conclusions. In this case, 2.

GDX FORECAST TO CLOSE UP CIRCA 1.69% TODAY
GDX FORECAST TO CLOSE UP CIRCA 0.54% FOLLOWING TRADING DAY




_______________________________________________________________
DISCLAIMER: Userque's trading signals are not affiliated with TSP Talk or Buy Low Sell High, Inc. The information is for educational purposes only! The information is not advice or recommendations. The information may be revised at anytime. Userque does not give investment advice. Do not act on this data. Do not buy, sell or trade any stocks, ETF's, indices, funds, or any other instruments/vehicles based on this information. Userque may trade differently than discussed or posted in this forum.
Click to expand...
 
RealMoneyIssues said:
1 - Are you tracking forecast vs. actual
2 - Can the system learn from this data

Thanks
Click to expand...

I only track what I post in the trade logs and graphs here. The system does track percentages and uses them to decide how to adjust the parameters. (I could look at them, of course). But, it only is concerned with getting the direction right. Whether a day will be up big or up small is irrelevant to a system that is only trading at the close.

Yes, it does learn. Each new trading day is added to the system. Those percentages are merely calculations based upon the trading day data: 100*(Close-Previous Close)/Previous Close. It "learns" two ways. It adjusts the parameters based upon how they perform, and;

If/when today's price action matches some day in the future, it will use what it "learns" today to help forecast in the future.
 
Progress Report

Currently testing a model that uses a hybrid cost function. Piggybacking on my previous discussion: This model will use the average of historical patterns only when the number of UP results are equal to the number of DOWN results; otherwise it will choose the direction that the most results are heading.

As a result, if this model outperforms the current model, I will use it instead and will no longer post the byproduct information as it won't always use the AVERAGE.

And, I may post a FORECAST for the OPENING, if the new model is better and believes it should be short instead of in nugt.

Also, after giving it much thought, I believe a different weighting scheme may do better. I've also applied the new scheme to the new model.
 
Progress Report Update:

Several configurations tested.

Model updated to best performing configuration. This model outperforms all prior models. :)

Preliminary forecast will be circa 15:40
Final forecast will be circa 15:55
 
Doesn't matter if it outperforms if no signals are produced...

userque said:
Progress Report Update:

Several configurations tested.

Model updated to best performing configuration. This model outperforms all prior models. :)

Preliminary forecast will be circa 15:40
Final forecast will be circa 15:55
Click to expand...
 
FOR 11-05-15 AT CLOSE:
HOLD NUGT

* MY APOLOGIES. MISSED FORECAST DUE TO PERSONAL REASONS. AS SUCH, THE SYSTEM REMAINS IN NUGT. I'VE NEVERTHELESS POSTED THIS LATE FORECAST FOR THOSE INTERESTED IN WHAT IT WOULD HAVE BEEN.




_______________________________________________________________
DISCLAIMER: Userque's trading signals are not affiliated with TSP Talk or Buy Low Sell High, Inc. The information is for educational purposes only! The information is not advice or recommendations. The information may be revised at anytime. Userque does not give investment advice. Do not act on this data. Do not buy, sell or trade any stocks, ETF's, indices, funds, or any other instruments/vehicles based on this information. Userque may trade differently than discussed or posted in this forum.
 
Progress, More Detail:

The latest model performed much better than the previous with only the OPEN (from the current day). But, I tested it with High, Low, and Close as well; it did a little better. So, I built in the option to give it that information, or not...just like before, but with an option this time.

(FYI only and FWIW, Out of curiosity, I ran the new model "for today." It disagreed with the prior model in that it called for today as being a down day.)
 
You must log in or register to reply here.

Similar threads

userque
  • Article Article
Introduction
Replies
0
Views
2K
userque
userque
userque
  • Article Article
userque & NUGT/DUST
Replies
0
Views
6K
userque
userque
E
Neural Network Artificial Intelligence System (No Chat)
4 5 6
Replies
104
Views
14K
Birchtree
Birchtree
350zCommTech
Citizens For Financial Responsibility Petition
Replies
9
Views
2K
350zCommTech
350zCommTech
mlk_man
Investment Trading System Scams Exposed
Replies
4
Views
11K
remmy
R
Back
Top