userque & NUGT/DUST

Did a whole lot of experimenting over the weekend. Didn't find a better configuration however. But, I didn't get a chance to finish the Index Feature.

Currently optimizing the last of the four GDX models, and after the opening business is handled shortly, I'll try to finish upgrading the TSP model.

I use the Nasdaq composite for TSP Equities because it goes back to 1984. With the new found excel speed, I'll start using the Nasdaq Model along with the individual equity fund models. For example, Only enter when they both agree; exit when either gives a sell signal.

The F fund model will still stand alone.

EDIT: The new intraday GDX models are still young in data as there is a gap in my intraday historical data. The models are subject to be more erratic as a result; until they build up more data points. Traditionally, it should be fine in a couple of weeks.
 
Holding DUST










_________________________________________________________________________
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.
 
Holding DUST










_________________________________________________________________________
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.

Currently updating model with today's opening price. I'll post if there are any changes.
 
Ran two models so far, they indicate a higher high tomorrow and a higher low tomorrow. Running the third model now--looking for an exit. If the [forecast] low of the day has passed, I'll run the final model (high time for the day) to try to avoid exiting at the high of today. Will post the results.
 
Last edited:
ENTERING NUGT CIRCA 13:20 ET.







_________________________________________________________________________
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.
 
EXIT DUST AT 20.72

ENTER NUGT AT 23.75




_________________________________________________________________________
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.
 
Still haven't gotten to the Index Feature yet (haven't finished optimizing the models yet). Had to redo those experiments I mentioned yesterday. One configuration was actually an improvement to the model's accuracy.

That'll be it for this model. Next I'll test a different type of model. The current model can't adjust individual input weights, it's either all or nothing. Also, the current model can't pick and choose inputs individually, only in a contiguous block that must start with the most recent history. This model also can't find hidden patterns.

The next model will search through all or most of the possible input/weight combinations to a specified precision, and be able to find hidden patterns. Again, I can attempt this now that I fixed my excel speed issue. I'll have to re-code most of the VBA, but I hope to have it up and running before tomorrow.

I'll stick with the most accurate model, of course.
 
HOLDING NUGT FOR NOW. HIGHER HIGH FORECAST FOR TOMORROW

I'll add today's optimizations to the database and will post if the forecast changes (or, if it is fine...either way). The database is not large yet, so adding only one day could affect things.





_________________________________________________________________________
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.
 
ENTERING DUST CIRCA 0945 ET.

EDIT:

EXIT NUGT AT 24.90
ENTER DUST AT 19.48








_________________________________________________________________________
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.
 
FOR 11-27-15
HOLDING DUST








_________________________________________________________________________
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.
 
EXIT DUST AT 18.90
ENTER NUGT AT 25.46


SP500 2082





_________________________________________________________________________
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.
 
Just curious.... what is the significance of listing the SP500 at 2082.

EXIT DUST AT 18.90
ENTER NUGT AT 25.46


SP500 2082





_________________________________________________________________________
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.
 
Just curious.... what is the significance of listing the SP500 at 2082.


I list the SP500 on the charts/trading-logs as a benchmark. Posting it along with my trades makes it so I don't have to look it up later, when I get around to updating the charts. :)
 
Update

Finished the new models. I've just figured out how best to use them. I likened it to how they must have thought when they first built the computer; or how families thought when they bought their first computer ever: "I know it's powerful, but what can I do with it?"

Though they perform better than the others I've built, they are still young. Firstly, they are optimizing on only three weeks or so of data right now. They use all of the historical daily GDX data that I have to do it (from 2006), but three weeks of optimization data is a tad short of best-practices. Each day, a new day is added to that total; so they'll eventually get to where they should be in a few to several weeks.

Secondly, I haven't started the feature creation yet.

Typically, daily historical data from the internet consists of Date, Open, High, Low, Close, and Volume. Most system developers add at least one created feature:

The daily price change percentage. ((Close-Previous Close)/Previous Close)*100

But, there are tons of possible features that can be created from the initial set.

High - Low. (a.k.a Range)

So now, in this example, we've added two brand new features. Two new columns to the spreadsheet. These added features may help the algorithm perform better. In fact, researchers have seen algorithms go from 80% accuracy to 100% accuracy by finding "the right" feature to add. These features can reveal hidden correlations and/or relationships in the data set.

The new models pick and choose which features to use from the lot they have been given. I note which features they like, and which they don't use. I keep the ones they like, and try new ones in place of the ones they don't.

The permutations are endless, even for the simple example above. Now throw in data from the first previous day, for example:

Previous High-High
Close - Previous Low
Etc.

Now throw in data from the third previous day...etc.
Now throw in genetic splicing (splicing the two examples above):

Previous High - (Close - Previous Low)

And genetic mutation:

Previous High - (Open - Previous Low)

Rinse and repeat.

This is genetic programming. I don't think I'll go so far as to automate this process; but I will manually do something of the sort. It'll be a slow, but deliberate process.

Additional to feature creation, I can now add data from other instruments/sources. Currencies, Labor statistics, Moon phases, Lottery Numbers, whatever. Regardless of what I throw at it, the algorithm decides if the dog hunts or not. If it doesn't, it ignores the feature/input. If it helps a little, it assigns it a little weight. If it helps a lot, it assigns it a lot of weight. If it helps inversely, it assigns it a negative weight.

Fun.

But first, I have to change the TSP models over to the new system. Not fun.

Will someone wake nnuut up, he likes to post those video links immediately after my posts. I guess I could post them for him.
 
Last edited:
LMAO...what about the rest of them :)

HOLDING NUGT TODAY




_________________________________________________________________________
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.
 
Back
Top