TSPCALC Strategy Question
Moderator: Aitrus
TSPCALC Strategy Question
In addition to this site I follow & pay into the TSPCALC site. There are a vast number of seasonal strategies on tspcalc site...
In the Results table for a chosen seasonal strategy returns are listed by month from Jan 2004 to the present month.
I've been following a strategy that has done well over the past 6 or 7 years...not so much this year however...
My question is for example if seasonal strategy 2000 had the best January return over the 2004 to 2019 time frame & strategy 3000 had the best February over the 2004 to 2019 time period and a different strategy had the best March over the same time period...etc.
Is there a seasonal strategy number that follows strategy 2000 for January then follows strategy 3000 for February and then switches to the best returning strategy each successive month? If that is the case what is that particular strategy number?
Thanks for any help…
R/
Heckrules
In the Results table for a chosen seasonal strategy returns are listed by month from Jan 2004 to the present month.
I've been following a strategy that has done well over the past 6 or 7 years...not so much this year however...
My question is for example if seasonal strategy 2000 had the best January return over the 2004 to 2019 time frame & strategy 3000 had the best February over the 2004 to 2019 time period and a different strategy had the best March over the same time period...etc.
Is there a seasonal strategy number that follows strategy 2000 for January then follows strategy 3000 for February and then switches to the best returning strategy each successive month? If that is the case what is that particular strategy number?
Thanks for any help…
R/
Heckrules
Re: TSPCALC Strategy Question
if i am not mistaken, tspcalc allows members to calculate their own custom strategies. use the parameters you just specified and see what comes out. because you are very specific to the results ofheckrules wrote:In addition to this site I follow & pay into the TSPCALC site. There are a vast number of seasonal strategies on tspcalc site...
In the Results table for a chosen seasonal strategy returns are listed by month from Jan 2004 to the present month.
I've been following a strategy that has done well over the past 6 or 7 years...not so much this year however...
My question is for example if seasonal strategy 2000 had the best January return over the 2004 to 2019 time frame & strategy 3000 had the best February over the 2004 to 2019 time period and a different strategy had the best March over the same time period...etc.
Is there a seasonal strategy number that follows strategy 2000 for January then follows strategy 3000 for February and then switches to the best returning strategy each successive month? If that is the case what is that particular strategy number?
Thanks for any help…
R/
Heckrules
each month of many existing strategies, it is going to take a bit of time to get to the answer. once you are done, let us know!
Anger and intolerance are the enemies of correct understanding.
― Mahatma Gandhi
If it's a choice between a difficult truth and a simple lie, people will take the lie every time. Even if it kills them.
― Paul Murray
― Mahatma Gandhi
If it's a choice between a difficult truth and a simple lie, people will take the lie every time. Even if it kills them.
― Paul Murray
Re: TSPCALC Strategy Question
heckrules wrote:In addition to this site I follow & pay into the TSPCALC site. There are a vast number of seasonal strategies on tspcalc site...
In the Results table for a chosen seasonal strategy returns are listed by month from Jan 2004 to the present month.
I've been following a strategy that has done well over the past 6 or 7 years...not so much this year however...
My question is for example if seasonal strategy 2000 had the best January return over the 2004 to 2019 time frame & strategy 3000 had the best February over the 2004 to 2019 time period and a different strategy had the best March over the same time period...etc.
Is there a seasonal strategy number that follows strategy 2000 for January then follows strategy 3000 for February and then switches to the best returning strategy each successive month? If that is the case what is that particular strategy number?
Thanks for any help…
R/
Heckrules
I don't use tspcalc, but from what I understand, the problem is a little more complicated than you [@heckrules] suggest. For example:bloobs wrote:
if i am not mistaken, tspcalc allows members to calculate their own custom strategies. use the parameters you just specified and see what comes out. because you are very specific to the results of
each month of many existing strategies, it is going to take a bit of time to get to the answer. once you are done, let us know!
Suppose the "best" January strategy has it's second and final IFT into the F-fund. Now also suppose the best February strategy assumes that Feb. must start already in the S-fund, and will make two IFT's later in that same month, Feb.
My point is that the months can't truly be considered independently, as you are suggesting. In this example, Jan. wants to end in the F fund, and Feb. wants Jan. to end in the S fund.
So the "best" Jan.-Feb. period may not be made up of the best Jan., nor the best Feb. strategy.
So, once you've finally found the best Jan.-Feb. period, you'll run into the same exact paradox when you try to add the "best" March period. So on and so forth. Even Dec. into Jan. is affected.
In conclusion: (As I've noted months ago:) The best solution will likely require a Machine Learning and/or an Optimization attack as all months ought be considered simultaneously, as they are all interdependent.
Good luck! and do keep us posted!
"In the land of idiots, the moron is King."
Re: TSPCALC Strategy Question
Thanks for the input...I’m not skilled enough with the TSPCALC strategy calculator to figure out all the details...it was more wondering if someone had already done the grunt work and how that new seasonal strategy was fairing this year.
Re: TSPCALC Strategy Question
The Seasonal Builder already does this automatically. The "best fit" feature of the Seasonal Builder has been solving this specific issue for about a year now.userque wrote:
I don't use tspcalc, but from what I understand, the problem is a little more complicated than you [@heckrules] suggest. For example:
Suppose the "best" January strategy has it's second and final IFT into the F-fund. Now also suppose the best February strategy assumes that Feb. must start already in the S-fund, and will make two IFT's later in that same month, Feb.
My point is that the months can't truly be considered independently, as you are suggesting. In this example, Jan. wants to end in the F fund, and Feb. wants Jan. to end in the S fund.
So the "best" Jan.-Feb. period may not be made up of the best Jan., nor the best Feb. strategy.
So, once you've finally found the best Jan.-Feb. period, you'll run into the same exact paradox when you try to add the "best" March period. So on and so forth. Even Dec. into Jan. is affected.
In conclusion: (As I've noted months ago:) The best solution will likely require a Machine Learning and/or an Optimization attack as all months ought be considered simultaneously, as they are all interdependent.
Good luck! and do keep us posted!
Owner/creator of TSPcalc.com - "Know your numbers"
Re: TSPCALC Strategy Question
Yes. You want to use the Seasonal Builder, not the Seasonal Calculator.heckrules wrote:In addition to this site I follow & pay into the TSPCALC site. There are a vast number of seasonal strategies on tspcalc site...
In the Results table for a chosen seasonal strategy returns are listed by month from Jan 2004 to the present month.
I've been following a strategy that has done well over the past 6 or 7 years...not so much this year however...
My question is for example if seasonal strategy 2000 had the best January return over the 2004 to 2019 time frame & strategy 3000 had the best February over the 2004 to 2019 time period and a different strategy had the best March over the same time period...etc.
Is there a seasonal strategy number that follows strategy 2000 for January then follows strategy 3000 for February and then switches to the best returning strategy each successive month? If that is the case what is that particular strategy number?
Thanks for any help…
R/
Heckrules
The difference being..
In the Seasonal Calculator, you input up to 36 trades (12 months x (2+G IFTs)), and it spits out the back-test data. That's pretty much all it does. The scope is always a whole year.
In the Seasonal Builder, all of the ways to trade EACH month have been calculated in advance. So you can sort through each month based on some criteria that you input (keep in mind you are entering criteria for each MONTH, not annual criteria). Then you can put each month together however you see fit. You will run into the issue that userque discussed, where one month's END FUND must align with the next month's START FUND, but you'll find that the Seasonal Builder has colorful fund-alignment bubbles that make it easy to see fund misalignments.
But you don't HAVE to do it by hand, because the Seasonal Builder includes a "best fit" tool that will figure out the best way to combine each month for you.
All you have to do is figure out what "best January" or "best February" means to YOU. Then input it on the Seasonal Builder. Below is an example if you're just looking for the highest mean 2004-2018:
1. Login and browse to the Seasonal Builder. On a desktop computer, it is a subset of the Seasonal menu. Hover your mouse over Seasonal and then drop to Seasonal Builder.
2. Set Start Year and End Year to your desired data range. If you select 2019 as end year, the data will end in 2019 when available (currently Jan-July), and end in 2018 for the rest of the months.
3. Set "Sort By: Mean Descending" (which is the default). This will return the top "monthly strategies" based on the highest mean for your data range (2004-2018 or 2004-2019).
4. Leave everything else blank. Click "Sort & Filter". The webpage will contact the server in the background and update the lists for each of the 12 months.
5. Once "Best Fit" quits spinning, click the "Apply" button. This will make 12 selections for you that all align END FUND to START FUND, and it will be optimized for your Sort By criteria (highest mean).
6. Now just scroll to the bottom, since all 12 months of selections are already made, and click "Submit to Calc".
This will send the strategy to the Seasonal Calculator in a new tab. The strategy that you are looking at has been built month-by-month based on your criteria.
So, for 2004-2018, the highest mean strategy is 68355.
For 2004-2019, the highest mean strategy is CURRENTLY 70568. I say currently because we are using current-year data. Once August is over, then August 2019 data will be added to the Seasonal Builder, so the 2004-2019 best-fit strategy will change.
Thanks
Matt
Owner/creator of TSPcalc.com - "Know your numbers"
Re: TSPCALC Strategy Question
userque wrote:I don't use tspcalc, but from what I understand, the problem is a little more complicated than you [@heckrules] suggest. For example:
Suppose the "best" January strategy has it's second and final IFT into the F-fund. Now also suppose the best February strategy assumes that Feb. must start already in the S-fund, and will make two IFT's later in that same month, Feb.
My point is that the months can't truly be considered independently, as you are suggesting. In this example, Jan. wants to end in the F fund, and Feb. wants Jan. to end in the S fund.
So the "best" Jan.-Feb. period may not be made up of the best Jan., nor the best Feb. strategy.
So, once you've finally found the best Jan.-Feb. period, you'll run into the same exact paradox when you try to add the "best" March period. So on and so forth. Even Dec. into Jan. is affected.
In conclusion: (As I've noted months ago:) The best solution will likely require a Machine Learning and/or an Optimization attack as all months ought be considered simultaneously, as they are all interdependent.
Good luck! and do keep us posted!
Thanks for the reply. I've read the information at https://tspcalc.com/seasonalresearch.php again and re-watched your training video.mjedlin66 wrote:The Seasonal Builder already does this automatically. The "best fit" feature of the Seasonal Builder has been solving this specific issue for about a year now.
Based upon my interpretation of your information, you seem to be performing a secondary/separate optimization to stitch adjacent months together, applied to an initial optimization of individual months. The first optimization seems to be a brute force over the individual months; while the latter optimization seems to be a more limited brute force over the junction between the individual months; again, in all cases, individually.
Assuming I'm correct (you don't exactly explain your approach), your approach is a creative work-around!
I not only posted about the stitching together of months problem; I also wrote about the notion of optimizing over the whole search space (all months and junctions) simultaneously...rather than piece-meal (individual months, then an individual best-fits optimization).
As you've noted in the past, optimizing over the whole search space simultaneously can't be done reasonably with brute force; hence, my reference to machine learning/optimization.
My current work is based upon price-action as well as time (seasonality). (IFT's aren't predetermined).
But I eventually want to get around to building the model I describe here (seasonality only) in order to compare the two to see which is better. My educated guess is that the model using price action and time will outperform the model using just seasonality (time)--this is why I started building the PA and time model first. But experimenting is the only way to be sure which model will be most effective.
If I weren't building my own; I'd certainly be utilizing your premium service!
"In the land of idiots, the moron is King."
Fund Prices2024-04-16
Fund | Price | Day | YTD |
G | $18.19 | 0.01% | 1.24% |
F | $18.58 | -0.32% | -3.33% |
C | $79.08 | -0.21% | 6.34% |
S | $76.95 | -0.41% | -0.18% |
I | $40.73 | -0.98% | 1.37% |
L2065 | $15.67 | -0.50% | 3.66% |
L2060 | $15.67 | -0.50% | 3.67% |
L2055 | $15.68 | -0.50% | 3.67% |
L2050 | $31.50 | -0.44% | 2.93% |
L2045 | $14.38 | -0.41% | 2.81% |
L2040 | $52.59 | -0.38% | 2.72% |
L2035 | $13.91 | -0.35% | 2.60% |
L2030 | $46.37 | -0.32% | 2.50% |
L2025 | $12.95 | -0.18% | 1.90% |
Linc | $25.31 | -0.14% | 1.64% |