Attempting to create a new monthly strategy...

[Chulke]

But, I am having incredible difficulty with the Months of June, August and September.

These 3 month are horrible for the markets and TSP in general.

June is a total crapshoot!
August is just a month to stay in the G Fund!
and September is a crapshoot as well!


But while I was typing this I just had another Idea...Wish me luck!

[Thomaneka]

Looking forward to seeing your results!

[12squared]

But, I am having incredible difficulty with the Months of June, August and September.

These 3 month are horrible for the markets and TSP in general.

June is a total crapshoot!
August is just a month to stay in the G Fund!
and September is a crapshoot as well!


But while I was typing this I just had another Idea...Wish me luck!

Chulke,

Have you looked at the Charts & Returns by Calendar Month?
After looking at the November dates in conjunction with the monthly graphs for various years it occurred to me that the market might sometimes be driven by floating holidays or events, e.g. Fed meetings and multiple options expirations.

In November, the best time to get into S fund appears to be about 7 to 8 days before Thanksgiving. In June, the market seems to go up the week before Father's Day, and often peaks the Monday after. Unfortunately, TSPcalc only counts by trading days. To handle June and November would require an option to use the day of the nth week, e.g. the 3rd Wednesday of the month.

12²

[Chulke]

Yeah that's kinda what I'm looking at...

Trying to meld the monthly strategy with the daily trading strategy from TSP calc.

Thus creating a strategy that will take advantage of the months of the year that have an easily predictable solid return with a high degree of certainty and culling out the months that have are harder to predict return and inserting the daily strategy that best fits and provides me with a higher level of certainty mixed with a solid return for that given month. If you can follow my ramblings here.

Basically you get something like this, just as an example:
January= F Fund
February= S Fund
March= S Fund
April= I fund
May= C Fund
June= Follow the trading guideline set forth in Strategy#17842
July= I Fund
August= Follow the trading guideline set forth in Strategy#17842
September= Follow the trading guideline set forth in Strategy#17842
October= C Fund
November= S Fund
December= C or S Fund which both perform reliably and produce good returns in December.

You get the picture I hope.


And I'm sure there is someone on here who has thought of this already.

idk...I was just thinking out loud I guess...


lol!

[phalanx]

#go-dailystrategy

[12squared]

Trying to meld the monthly strategy with the daily trading strategy from TSP calc.

Thus creating a strategy that will take advantage of the months of the year that have an easily predictable solid return with a high degree of certainty and culling out the months that have are harder to predict return and inserting the daily strategy that best fits and provides me with a higher level of certainty mixed with a solid return for that given month. If you can follow my ramblings here.

Basically you get something like this, just as an example:
January= F Fund
February= S Fund
March= S Fund
April= I fund
May= C Fund
June= Follow the trading guideline set forth in Strategy#17842
July= I Fund
August= Follow the trading guideline set forth in Strategy#17842
September= Follow the trading guideline set forth in Strategy#17842
October= C Fund
November= S Fund
December= C or S Fund which both perform reliably and produce good returns in December.

I think 19226 describes this. It is better in January than most dailies, but hardly a "predictable solid return with a high degree of certainty" on an annual basis.

[Chulke]

I see what you did there...


And yes I realize I could have just plugged in my trade days into the TSPCalc website and pluck out a strategy like what you did...But TSP Calc only has 13 years of data....


I was attempting to extrapolate a strategy based on ALL the data available to me and forego the months with consistently bad returns on the monthly strategies (like June, August and September) and inject a bit of daily to bring up the numbers a bit.

See, that's what I'm having a hard time understanding all the hype with the Daily strategies, because it has such a limited data set when there are actually nearly 30 years worth of total data for TSP.

But, you can't blame a guy for trying to make something better...

Which, is basically why I mostly stay away from posting on here because everyone thinks they have found the best way already.


So good luck to you!

[12squared]

...But TSP Calc only has 13 years of data....

See, that's what I'm having a hard time understanding all the hype with the Daily strategies, because it has such a limited data set when there are actually nearly 30 years worth of total data for TSP.

30 years of data only reduces the margin of error by 1/3 compared to 13 samples. Quadrupling the number of data samples reduces the margin of error by half. E.g. you need 52 years of data to be twice as certain than for 13 years.

Experiment for yourself.
Confidence Interval Calculator

[phalanx]

...But TSP Calc only has 13 years of data....

See, that's what I'm having a hard time understanding all the hype with the Daily strategies, because it has such a limited data set when there are actually nearly 30 years worth of total data for TSP.

30 years of data only reduces the margin of error by 1/3 compared to 13 samples. Quadrupling the number of data samples reduces the margin of error by half. E.g. you need 52 years of data to be twice as certain than for 13 years.

Experiment for yourself.
Confidence Interval Calculator

Math. It's amazing.

[Timber82]

...But TSP Calc only has 13 years of data....

See, that's what I'm having a hard time understanding all the hype with the Daily strategies, because it has such a limited data set when there are actually nearly 30 years worth of total data for TSP.

30 years of data only reduces the margin of error by 1/3 compared to 13 samples. Quadrupling the number of data samples reduces the margin of error by half. E.g. you need 52 years of data to be twice as certain than for 13 years.

Experiment for yourself.
Confidence Interval Calculator

12squared is a witch!!Burn him/her!

[Aitrus]

Wait! We can't burn 12^2 yet. We have to see if he/she weighs the same as a duck first.

[12squared]

Wait! We can't burn 12^2 yet. We have to see if he weighs the same as a duck first.

I am not made of wood - and my ability to float is marginal.

[robika]

Ugh, all the people who keep saying we need 30 samples.

30 is not a rule, it's a guideline. 30 samples is used because it is typically considered large enough for the sampled means to be approximately normally distributed.

As 12squared pointed out above, the confidence interval calculator is going to tell you in what range where the actual mean of your samples falls.

For me, I follow 17620, using the calculator and choosing a 99.9% confidence level, the actual mean of the 13 years falls between 27.45% and 32.25%. Assuming absolute worst case scenario with this confidence level let's say that the actual mean is equal to 27.45%, 6 standard deviations from the mean puts the return at 11.6%

I can live with that.

If you want to wait another 16.1 years until there are "30 samples" so that you can use this data to follow a strategy, by all means wait. But to me, statistically, 13 samples is good enough to start with.

And before anyone goes nuts, I know that the market is completely driven off of people's kneejerk reactions and that "past performance does not guarantee future returns" blah blah blah... that's why I said 13 samples is good enough to start with.

[evilanne]

I see what you did there...


And yes I realize I could have just plugged in my trade days into the TSPCalc website and pluck out a strategy like what you did...But TSP Calc only has 13 years of data....

I was attempting to extrapolate a strategy based on ALL the data available to me and forego the months with consistently bad returns on the monthly strategies (like June, August and September) and inject a bit of daily to bring up the numbers a bit.

See, that's what I'm having a hard time understanding all the hype with the Daily strategies, because it has such a limited data set when there are actually nearly 30 years worth of total data for TSP.

But, you can't blame a guy for trying to make something better...

Which, is basically why I mostly stay away from posting on here because everyone thinks they have found the best way already.

So good luck to you!

I think your intent is the same as what many have done with the daily strategies. Regarding concerns for limited data, consider the following.

Most of the daily strategies correlate somewhat to the monthly strategies. Many of the strategies started with the monthly strategies but were tweaked to improve the RoR, generally by trial and error by different people.

Below is a comparison of 12x12’s Daily Strategy http://tspcalc.com/seasonal.php?ID=19255&years=04-16 to top 5 monthly strategies (see atch). Remember that trading days in a month are about 20-22 days

Jan F Fund - 19 Days (All but 1 day) vs 4 out of 5 are in F Fund (1 in S)
Feb S Fund - 16 Days (All but 3 days) vs 3 out of 5 are in S Fund (1 in S & I)
Mar S Fund - 16 Days (All but 7 days) vs 2 out of 5 are in S Fund (3 in C)
Apr F Fund - 13 Days (All but 6 days) vs 2 out of 5 are in C Fund (2 in I)
May Not Comparable - 2 Days in C & 6 days in S vs Monthly in C or S
Jun F Fund - 11 Days (Rest in G) vs All in F Fund
Jul C Fund - 11 Days (3 days in S & Rest in G) vs 3 out of 5 are in C Fund (2 in I)
Aug F Fund - 11 Days (6 days in S & Rest in G) vs All in F Fund
Sep Not Comparable - 2 Days in C & 6 days in S vs All in F Fund
Oct C/S Fund - 7/6 Days (Rest in G) vs All in C Fund
Nov C/S Fund - 6/9 Days (Rest in F) vs 3 out of 5 in C Fund (2 in S)
Dec S/I Fund - 6/9 Days (Rest in F) vs 4 out of 5 are in S Fund (1 in I)

9 out of the 12 months are generally comparable with 1/2+ of the days in one or more of the same funds as the 5 most popular strategies (October is split between C & S, which is < 1/2 but still in stock > 50% of time).

12X12’s strategy is one of the later daily strategies that focuses on lower standard deviation, which may explain those non-comparable months but in general the monthly strategies are still very similar to to monthly strategies that rely on more data. I think the trial & error process discovered some shorter seasonal sweet spots that were not available using only monthly data.

It is the fidelity of the data for the shorter 13-14 years and utilizing more IFTs that give daily strategies the advantage of greater returns if you can get past the limited time period but as many have pointed out, more recent data is generally more relevant given changes in economic conditions.

Many of the dailies also reduce the risk by limiting the time in the stock funds that was also one of the underlying premises in the creation of the monthly strategies. Finally, when the monthly TSP monthly strategy was created there was less data. The' data goes back to 1988 or 29 years, so if you subtract 9 years (I'm guessing that the TSP seasonal strategy was 1st developed at the time this site was created or in 2008 during the financial crises) that would have been only 20 years of data. It was about 5 years ago that Aitrus & started his analysis and many other mixes were developed.

[12squared]

Below is a comparison of 12x12’s Daily Strategy http://tspcalc.com/seasonal.php?ID=19255&years=04-16 to top 5 monthly strategies (see atch). Remember that trading days in a month are about 20-22 days

19255 is the same as 18390, except for January, which was changed to improve the Mean/StdDev for that month. Ironically, the more consistent performance in January (only one negative year vs. 4 for 18390) caused the annual Mean/StdDev to increase sharply.