userque & NUGT/DUST

[FAAM]

Thanks Stoplight,

I appreciate that. I think RMI is trying to help me in some type of way; maybe we're just not understanding what the other is saying right now.
_______________________

Proverbs 27:17 Iron sharpeneth iron; so a man sharpeneth the countenance of his friend. (KJV)

I like that reference from the Good Book. I "third" that agreement with Stoplight's comments, and thank you Q again for your efforts, sharing and openness.

 

[userque]

I like that reference from the Good Book. I "third" that agreement with Stoplight's comments, and thank you Q again for your efforts, sharing and openness.

Thanks FAAM

 

[k0nkuzh0n]

...

Thank you. I took the move to a cash as a safe move you personally doing because the system didn't have a chance to give an official signal on time. Which is certainly a smart decision if that was the case.

Then you seemed to finish the run and determined that had the system had enough time to finish, it would have stayed NUGT. In which case I assume you would have calculated performance based on that, and not the move to cash, since the cash move was personal choice (Again, based on my initial assumption). But then that result was invalidated, so it wasn't entirely clear to me, which would be the right way to go.

What language did you end up going with anyway? Python?

 

[userque]

Thank you. I took the move to a cash as a safe move you personally doing because the system didn't have a chance to give an official signal on time. Which is certainly a smart decision if that was the case.

Then you seemed to finish the run and determined that had the system had enough time to finish, it would have stayed NUGT. In which case I assume you would have calculated performance based on that, and not the move to cash, since the cash move was personal choice (Again, based on my initial assumption). But then that result was invalidated, so it wasn't entirely clear to me, which would be the right way to go.

What language did you end up going with anyway? Python?

I understand. I try keep it about the system (regarding this thread), and not about me personally. Just so happens, I do follow my system's signals.

Correct. I moved to cash because the system was unable to forecast.

When I finished the run, I indicated what the system would have done in case someone wanted to know. Some may trade intraday in demo accounts based upon the system. Some may just be curious. Some may be keeping their own types of stats and may want to know the actual call, etc.

I don't think it would be right to make the call retro-active for the purposes of this thread. I want to show what the system can really do--in a real world situation; not what it theoretically could do. In real life, we had no signal to trade with. That should be reflected in this thread.

Personally, I went to cash. But I could have resorted to old school T/A and traded without my system, and that would have had nothing to do with my system, or the thread.

Which is right? I think making the call retroactively is fine for non-public analysis of a system, if it is done fairly. But publicly, I don't think such retro-active moves are fair. How is the public to confirm whether one is "cheating." That's why I don't do retro-active moves here.

Computer Language?

After transferring ALL calculations from the spreadsheet into VBA, I found that there wasn't much difference in the speed. (I put everything back the way it was. It's easier to update that way). So I decided that Python, not being a compiled language, may not offer much speed improvement to justify the effort. So I considered Java. I like that idea as I would also like to do Android coding.

I did some tricks with the sheet to speed up the process. It's fast "enough" for now.

So, I think I'll knock out javascript first to help quickly automate some unrelated things I do online; then java.

 

[userque]

FOR 10-23-15 AT CLOSE:
ENTER DUST

NEXT TRADING DAY AT OPEN:
HOLD DUST

* Partial optimization

_______________________________________________________________
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.

 

[userque]

will post later today.

 

[userque]

...

 

[userque]

FOR 10-26-15 AT CLOSE:
ENTER NUGT

FOLLOWING TRADING DAY AT OPEN:
HOLD NUGT


_______________________________________________________________
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.

 

[userque]

FOR 10-27-15 AT CLOSE:
HOLD NUGT

FOLLOWING TRADING DAY AT OPEN:
HOLD NUGT





_______________________________________________________________
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.

 

[userque]

FOR 10-28-15 AT CLOSE:
HOLD NUGT

FOLLOWING TRADING DAY AT OPEN:
HOLD NUGT



_______________________________________________________________
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.

 

[userque]

Had a chat with someone. Thought I'd post one paragraph in case others think as that person does. Hopefully, this helps:

Why doesn't your system give a direction?

"When the system says NUGT at open; that means that it expects the CLOSE to be higher than the OPEN. The direction is UP in this example. Think about it. That is DIRECTION. It only predicts the relative direction between the OPEN and the CLOSE. If a price is below the open, then logic dictates the direction is UP. If a price is higher than the OPEN then there is no way it knows the direction from that point because it can go either UP or DOWN and still CLOSE higher than the OPEN. I'm going to post this explanation to the thread in case others are thinking like you."

Additionally, the system doesn't intraday trade; so it only cares about directions between the OPEN and the CLOSE. If I were intraday trading, I'd build a system to do just that; but I'm not intraday trading.

 

[userque]

FOR 10-29-15 AT CLOSE:
HOLD NUGT

FOLLOWING TRADING DAY AT OPEN:
ENTER 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.

 

[userque]

FOR 10-30-15 AT CLOSE:
ENTER NUGT

FOLLOWING TRADING DAY AT OPEN:
HOLD NUGT

* System updated. Optimization 90% complete.


_______________________________________________________________
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.

 

[userque]

Major upgrades and testing underway.

System now will act under a profit motive rather than a "pattern-match motive." (My original concern with a profit motive was over-fitting. However, my system has since been forecasting based upon actual price-action history and not black-box calculations.)

Overfitting Definition | Investopedia

Will test/compare 3 configurations: #1 Current (Trading open and close based on current day open,high,low,close, and other information, of course.); #2 Close (Trading only the close based on same information as above.); and #3 Close (Trading only the close based on only the current day open, and other information, of course.)

I'm testing #1 because it would seem it has the greatest profit potential;
#2 Because it would seem the odds are better when only trying to predict the close, rather than predicting both the open and the close; and #3 Because it would be so convenient to be able to enter a MOC order right after the open.

If #3 is "close enough" in profit when compared to the other two, I'd love to go with #3.

Finally, I've finally discovered/developed a way to make the parameters dynamic. While there may be one "setting" that works all the time; it may be more profitable to have a setting than can change with the "weather." I now have a way to test this.

 

[userque]

Forgot to mention:

There are additional major improvements under the hood regarding input weighting as well.

And, the system will now only look one day ahead. I suppose the shorter you look ahead; the more probable the prediction is correct. The system is more conducive to discount brokerage accounts geared towards active traders like Interactive Brokers or deep discount brokers like Robinhood. That matters less with "large" accounts.

While we're here: It trades for the next day only. It doesn't use some sort of "trend." It doesn't calculate some sort of "trend."

 

[userque]

Ran a sample run on all three models. The model that trades the open and the close did the worst. The model that would "run" anytime after the open did slightly better than the model that runs at the close. Too close to call those two with such a small sample. Running a longer optimization now in order to decide. Then it's on to modifying the TSP models.

 

[RealMoneyIssues]

Just curious on the TSP model... are you using closing prices or prices at 1130 EST/EDT since the IFT has to be made prior to noon?

 

[userque]

Just curious on the TSP model... are you using closing prices or prices at 1130 EST/EDT since the IFT has to be made prior to noon?

I use the opening price that morning.

(Of course, the system uses much more than that, that's just the most recent data it uses.)

 

[RealMoneyIssues]

Sorry, also meant to ask if the TSP model would give a signal at the same time, 1130 EST, so the IFT could be made that day or is it designed to determine the move EOD and the IFT is done the following day?

 

[userque]

For example, Monday, anytime after the open, I plug in the opening price. The system outputs the forecast. That forecast is "for what to do, that morning, before noon."

It takes into account that the trade doesn't actually take place until COB in its calculations.

I've pretty much neglected the TSP models in lieu of perfecting the GDX model. Looks that'll be changing soon based upon the numbers I saw earlier tonight with the updated model. We'll see.

Sorry, also meant to ask if the TSP model would give a signal at the same time, 1130 EST, so the IFT could be made that day or is it designed to determine the move EOD and the IFT is done the following day?