The daily seasonal strategy

[mjedlin66]

This thread is dedicated to my seasonal strategy role model, Aitrus.

Greetings. Here is the "daily seasonal" strategy that I came up with at the end of 2015 and utilized effective January 1st, 2016. My fantasyTSP does not match this strategy 100% because sometimes I miss IFTs. So here are the 2016 results- Daily Seasonal: 12%. S-Fund: 16.3%.

So I lost to the S fund, but I beat the other core funds. I am hoping it beats all of the core funds this year. But even if it doesn't I may stick with it. For 2016, I was only 3 points behind the leading core fund, and I did that with less risk because my strategy is only in equities roughly 50-75% of the time. If a market crash occurs when I am in G or F, I will be much better off.

This strategy was made in Excel. I analyzed all of the daily prices from 06/2003 - 06/2015. First, I converted each calendar day to a trading day. The 5th trading day may fall on the 7th of the month one year and the 5th of the month the next year. In order to compare the 5th trading day of one year to the 5th trading day of the next, I converted every calendar day to a trading day by stripping out the weekends and holidays.

Then, I found the daily change for every trading day by comparing it to the previous trading day. Once I had the daily change as a percentage, I divided them by 100 to come up with a multiplicative factor. Finally, I averaged these multiplicative factors across every year 2003-2015.

The result was one full year worth of average gains for each trading day of the year.

I took that 1 year of averages and ran it through an IFT simulator. The IFT simulator was just a bunch of Excel macros which copied all of the average gains for one month and put it through every possible IFT combination. There are 1794 possible IFT combinations for a 23-day period. Remember we are using trading days rather than calendar days, and the maximum number of trading days for any month is 23. So excel ran all 1794 possible IFT combinations for the average gains and then spit out the maximum and which strategy came up with it.

For example, the biggest monthly gain the simulator could come up with for February was 3.35%, and this occured when- Starting: S fund, IFT to I Fund at COB on the 9th trading day, and then IFT to F fund at COB on the 13th trading day, and IFT to G at COB on the 20th trading day and stay there until the end of the month.

It was extremely specific. Notice that the simulator even took into account the "3rd IFT", which is the possibility of going to G fund ONLY as a 3rd IFT.

Once I had the IFT patterns for every month, I had to make them line up. That means that the starting fund of one month had to match the ending fund of the previous month. I did all that by hand, but there were remarkably few adjustments required.

This is the result:
Starting Fund, 1st IFT Fund:trading day, 2nd IFT Fund:trading day, 3rd IFT trading day (G only)

JAN- I, F:1, S:14
FEB- S, I:9, F:13, G:19
MAR- G, S:6
APR- S, F:3, I:6
MAY- I, F:2, S:14, G:21
JUN- G, S:9, G:12
JUL- G, C:1, I:12
AUG- I, F:1, S:17
SEP- S, F:11, C:18, G:21
OCT- G, S:7
NOV- S, F:4, S:13
DEC- S, I:15

For example, F:2 means put in your IFT before noon on the 2nd trading day of the month, so that you buy F at COB (close of business) on the 2nd trading day. If the G trading day doesn't exist in the month of interest, then execute it on the last trading day of that month.

Here are the returns-
2004: F=4.3%, C=10.8%, S=18.0%, I=20%, Mine=25.6%
2005: F=23.9%, C=4.95%, S=10.45%, I=13.6%, Mine=21.2%
2006: F=4.4%, C=15.8%, S=15.3%, I=26.3%, Mine=31.75%
2007: F=7.1%, C=5.5%, S=5.5%, I=11.4%, Mine=20.5%
2008: F=5.1%, C=-38%, S=-39.5%, I=-42.9%, Mine=45.8%
2009: F=5.99%, C=26.7%, S=34.8%, I=30%, Mine=82.6%
2010: F=6.7%, C=15%, S=29%, I=7.9%, Mine=56.5%
2011: F=7.9%, C=2.1%, S=-3.4%, I=-11.8%, Mine=35.8%
2012: F=4.3%, C=16%, S=18.6%, I=18.6%, Mine=41%
2013: F=-1.8%, C=32.8%, S=37.7%, I=22.5%, Mine=30.5%
2014: F=6.6%, C=14.5%, S=8.4%, I=-4.7%, Mine=13.6%
2015: F=0.91%, C=1.46%, S=-2.92%, I=-0.51%, Mine=21.71%
2016: F=2.9%, C=12%, S=16.3%, I=2.1%, Mine=12% (First year of use)
2017 YTD: F=1.88%, C=4.6%, S=2.3%, Mine=3.7% (As of 4/16/17)

Past data does not guarantee future returns. Use at your own risk.

*** Courtesy of nrialto, these are the 2017 trading dates-

F: May 2nd
S: May 18th
G: May 30th

S: June 13th
G: June 16th

C: July 3rd
I: July 19th

F: August 1st
S: August 23rd

F: September 18th
C: September 27th
G: September 29th

S: October 10th

F: November 6th
S: November 17th

I: December 21st

[crondanet5]

A couple of quick thoughts come to mind.

1. How would 2017 ytd look?

2. Remember that the third IFT can be done incrementally rather than a lump sump IFT to the G Fund. Were there months where multiple G Fund IFTs might have worked better and what percentages per G Fund IFT historically worked best?

[mjedlin66]

A couple of quick thoughts come to mind.

1. How would 2017 ytd look?

2. Remember that the third IFT can be done incrementally rather than a lump sump IFT to the G Fund. Were there months where multiple G Fund IFTs might have worked better and what percentages per G Fund IFT historically worked best?

1. I will have to calculate the 2017 YTD. I am not sure.

2. Well, being split between G and some other fund isn't going to net you more money than just being in the fund that performs better than the other. Going 50% from C to G and then going moving the rest to G three days later isn't going to beat full swing C or G.

[Mad Maxx]

Intersting stats. One question though. For this year (2017) there isn't as 21st trading day in September, only 20. Maybe I missed the floating factor in your original message, but what do you do in this case where your stats came out to change to G on the 21st trading day, but there isn't a 21st? Good on you for thinking "out of the box."

[mjedlin66]

If the strategy says to go to G on the 21st day, but there aren't 21 trading days, then you go to G at COB on the last day of the month.

2017 YTD: F=1.88%, C=4.6%, S=2.3%, Mine=3.7%

[crondanet5]

When I wrote of incremental IFTs to the G Fund instead of one 100% IFT to the G, how about 10 IFTs of 10% moves to the G Fund over 10 trading days?

[mjedlin66]

When I wrote of incremental IFTs to the G Fund instead of one 100% IFT to the G, how about 10 IFTs of 10% moves to the G Fund over 10 trading days?

Same story. The only reason to do that is for piece of mind. The aim of this analysis was to find the highest returns. The law of averages works against split investments if you are only looking for the highest returns.

[cswift01]

Very nice work! This was what I was hoping to find out some time ago. I began at asking the TSP when they would pay dividends (i.e. increase the value of the share price) to eventually hope to complete the exact work you've done. Thanks for saving me time!

Oh and they never gave any date for dividends. They said they get included from time to time, whatever that means.

Thanks,

Me

[Aitrus]

That's some good work, MJ! Kudos on taking and idea and running with it to see where it goes.

Any chance of going back further than 2004? What are your next steps in your plans to refine your program from here?

[Alaskansto]

Very impressive work. I would suggest looking at JTH'S blog over on tsptalk to get some ideas on how to present the data. He would break down daily moves monthly for each starting fund monthly in a chart. ex comb1 start s 1-14 ift 15th c 15-26 g 26-30. His blog is still under the best blogs section if you want to check it out. He quit posting about a year ago, I really miss those charts.

[crondanet5]

Not the same story. Somebody actually did this back in 2008 so they received the gains from the early part of the month while incrementally reducing their exposure and when the market crashed they had exited the day before. That's where I'm wondering if incrementally IFTing to G would produce better historical returns rather than a single G Fund IFT.

[hytid]

Not the same story. Somebody actually did this back in 2008 so they received the gains from the early part of the month while incrementally reducing their exposure and when the market crashed they had exited the day before. That's where I'm wondering if incrementally IFTing to G would produce better historical returns rather than a single G Fund IFT.

And would he have made more if he'd stayed 100% in C until the day before the crash? What about the day before that?

In no way can an incremental exit earn more than a perfectly timed exit; it only reflects some uncertainty. Nor will a split in S/C, or F/G for that matter (there will only be one best fund for any duration). If you have the numbers that show a period of time where some ideal split earned more than any single fund, please share.

Acting under the assumption that these timing models provide an "advantage" over others (which is why we're all here, I guess) one's best option is to move 100% into the desired fund and over the long term assuming that these models do provide an advantage, you will have earned more.

OP, have you weighted more recent data more heavily? I've always been curious about something like this.

[Rance]

What about 2015?

August was a disaster, but October was great.

2015 has to be accounted for - good or bad.

[anomadtoo]

2015 would be great....and thank you for sharing your data and insight!

[mjedlin66]

Man, you guys are needy.