Page 1 of 2

September - October seasonality

Posted: Tue Sep 17, 2024 8:48 am
by 12squared
Bob Pisani wrote:The most intense panics on Wall Street have tended to occur during the late summer and early autumn months. This can be traced all the way back to the 1800s...
Here’s why September and October are historically weak for stocks

Re: September - October seasonality

Posted: Tue Sep 17, 2024 9:34 am
by bloobs
12squared wrote: Tue Sep 17, 2024 8:48 am
Bob Pisani wrote:The most intense panics on Wall Street have tended to occur during the late summer and early autumn months. This can be traced all the way back to the 1800s...
Here’s why September and October are historically weak for stocks
To add to this, I saw a chart the other day indicating that after a fed cut announcement results in a negative market 100% outcome after the first 5 days. Pretty convincing, which has my finger on the sell trigger right about now, in spite of what my seasonal schedule dictates.

Re: September - October seasonality

Posted: Tue Sep 17, 2024 9:45 am
by mrwawa
This is what I was wondering about in another thread. My seasonal strategy has me leaving the market today. I was wondering if the rate cut would result in positive gains or not. We will see.

Re: September - October seasonality

Posted: Tue Sep 17, 2024 10:15 am
by bloobs
mrwawa wrote: Tue Sep 17, 2024 9:45 am This is what I was wondering about in another thread. My seasonal strategy has me leaving the market today. I was wondering if the rate cut would result in positive gains or not. We will see.
Wawa - I wanted to be sure if the 100% stat I quoted was correct, and I may have overstated it. SORRY.
It is not 100% at least not for the S&P500.

Apparently, the actual outcomes after a rate cut are:

42% positive after 5 days
50% positive after 10 days
83% positive after 21 days
73% after 63 days.

No where near as bad as I thought I heard it.

So I am staying in the market through this week per my seasonal schedule.

Re: September - October seasonality

Posted: Tue Sep 17, 2024 11:15 am
by 12squared
bloobs wrote: Tue Sep 17, 2024 9:34 am To add to this, I saw a chart the other day indicating that after a fed cut announcement results in a negative market 100% outcome after the first 5 days. Pretty convincing, which has my finger on the sell trigger right about now, in spite of what my seasonal schedule dictates.
The seasonals I have been tracking move to F or I today. However, my own assessment of TSP data from 2004-23 reveals that F has been positive (mean and median) from the 12th to 15th trading days, and I positive from the 11th to 14th trading days. C and S go negative over the same period. Accordingly, I am planning to fade to G by the end of this week.

12²

Re: September - October seasonality

Posted: Tue Sep 17, 2024 11:39 am
by mrwawa
Thanks for the update. I am now curious as to how it is going to play out.

Re: September - October seasonality

Posted: Wed Sep 18, 2024 3:49 am
by Bubba
I'm 5% in C and 5% in S with the plan to sell on Friday (per my own system). It could be that this 10% will be very painful, but even a 10% drop there wouldn't be as painful as a larger drop. I too am expecting first euphoria at the cuts...but most likely either the same day or a couple days afterwards a big sell off. We'll see though. Sept seasonality doesn't always hold. Having said that, I would rather miss a few percentage points increase than receive a few percentage points loss. Just me.

Here's a great one for Sept seasonality. In reality if you you can buy ETFs, then holding tech software for the first 2 weeks of September is almost always great. Afterwards its just downhill...

https://jayonthemarkets.com/2018/08/28/ ... -strategy/

Re: September - October seasonality

Posted: Wed Sep 18, 2024 9:25 am
by bloobs
Bubba wrote: Wed Sep 18, 2024 3:49 am I'm 5% in C and 5% in S with the plan to sell on Friday (per my own system). It could be that this 10% will be very painful, but even a 10% drop there wouldn't be as painful as a larger drop. I too am expecting first euphoria at the cuts...but most likely either the same day or a couple days afterwards a big sell off. ... I would rather miss a few percentage points increase than receive a few percentage points loss. Just me.
Exactly my thoughts....but i committed this year to faithfully follow my chosen strategy's move (the strat i supposedly carefully curated last year) all the way through (down the rabbit hole it leads me :shock: )

The way I see it, my commitment to base my TSP investment decisions on seasonal methodologies was so I dont have to worry nonstop about the daily/weekly nonsensical movements of the market.

Eh, I guess I can blame it on the strat and Matt if it doesnt work out HAHAHAHA.

Re: September - October seasonality

Posted: Thu Sep 19, 2024 3:40 am
by Bubba
I've come across many neat strategies. I'm still working (seems like a never ending thing here) on finding the right strat for me with the hopes that I can just do it and call it a day (as you say and follow it down the rabbit hole). I'm not entirely committed to whatever decision I made in December, because every day I learn new things.

In any case, I recently came across this puppy which looks very, very interesting https://tspcalc.com/seasonal.php?ID=98482. It's made before 2020 (can't see exact date) and it's been doing quite well with a lower SD. I'll be mixing this in with other chosen strats to see where I can get from there.

Good luck!

Re: September - October seasonality

Posted: Thu Sep 19, 2024 7:06 am
by Aitrus
Bubba,

Take a look at 85660. Also created before 2020, and the biggest differences seem to be in Jan and April. Everything else is pretty minor. But those differences have combined to make a 3.5% difference in CAGR from 2004-2023. 98482 is doing better this year and lower SD, but 85660 has better long term mean.

But both seem to closely follow each other in terms of when to be in, when to be out, and which Fund to use when in.

Re: September - October seasonality

Posted: Thu Sep 19, 2024 7:41 am
by Aitrus
Ok, I just experimented with something, and noticed an oddity.

I took the best individual monthly performance from 2004-2023 and compared them to each other. Here's the results:

Jan - 85660 (1.25% vs 0.77%)
Feb - tie
Mar - tie
Apr - 85660 (2.48% vs 1.43%)
May - 98482 (1.80% vs 1.63%
Jun - 85660 (1.74% vs 0.92%)
Jul - tie
Aug - tie
Sep: 85660 (1.23% vs 1.16%)
Oct: 85660 (1.69% vs 1.64%)
Nov: 98482 (3.09% vs 2.90%)
Dec: 98482 (2.80% vs 2.73%)

So I took the best month from each one and used it to create a new Mix - 252916.

But the oddity was January. I left January as 85660, but changed Dec to 98482. This changed the first day of Jan from I Fund to G Fund, because 98482 has the Mix going to G on the 20th day, thus being in G on the first day of Jan. The result was that even though the new Mix has 85660's allocations, it returns 0.77% - the same as 98482. The rest of Jan the two Mixes are identical.

So the difference of CAGR by being in G vs I for Jan is a 0.48%. That's a huge amount of impact for a single day of investing. It tells me that even though 98482 has better long-term performance in Dec (2.80% vs 2.73%), it's outweighed by that Jan takes due to being in the wrong Fund at the beginning of the year - even for just a single day.

The other thing I noticed is that even if you take the best individual month from each Mix, it still underperforms 85660 somehow. 252916 (and 252919 is what it looks like when you leave Dec looking like 85660 to get back that 0.48% in Jan) underperforms 85660 by about 1% CAGR.

That's because the same thing happens in June as what happens in January, when the first three of days of the month have the Mix in G instead of S, and the difference has a difference of 0.92% (using G) vs 1.74% (using S).

It's interesting to see, and flies against conventional logic which assumes that each month's best allocation should win out. It doesn't take into account the effect it has on the early days of the next month. As a result, 85660 is the best performer out of all of them, even though 98482 outperforms in May, Nov and Dec on paper.

Re: September - October seasonality

Posted: Thu Sep 19, 2024 8:14 am
by Bubba
Aitrus wrote: Thu Sep 19, 2024 7:41 am Ok, I just experimented with something, and noticed an oddity.

I took the best individual monthly performance from 2004-2023 and compared them to each other. Here's the results:

Jan - 85660 (1.25% vs 0.77%)
Feb - tie
Mar - tie
Apr - 85660 (2.48% vs 1.43%)
May - 98482 (1.80% vs 1.63%
Jun - 85660 (1.74% vs 0.92%)
Jul - tie
Aug - tie
Sep: 85660 (1.23% vs 1.16%)
Oct: 85660 (1.69% vs 1.64%)
Nov: 98482 (3.09% vs 2.90%)
Dec: 98482 (2.80% vs 2.73%)

So I took the best month from each one and used it to create a new Mix - 252916.

But the oddity was January. I left January as 85660, but changed Dec to 98482. This changed the first day of Jan from I Fund to G Fund, because 98482 has the Mix going to G on the 20th day, thus being in G on the first day of Jan. The result was that even though the new Mix has 85660's allocations, it returns 0.77% - the same as 98482. The rest of Jan the two Mixes are identical.

So the difference of CAGR by being in G vs I for Jan is a 0.48%. That's a huge amount of impact for a single day of investing. It tells me that even though 98482 has better long-term performance in Dec (2.80% vs 2.73%), it's outweighed by that Jan takes due to being in the wrong Fund at the beginning of the year - even for just a single day.

The other thing I noticed is that even if you take the best individual month from each Mix, it still underperforms 85660 somehow. 252916 (and 252919 is what it looks like when you leave Dec looking like 85660 to get back that 0.48% in Jan) underperforms 85660 by about 1% CAGR.

That's because the same thing happens in June as what happens in January, when the first three of days of the month have the Mix in G instead of S, and the difference has a difference of 0.92% (using G) vs 1.74% (using S).

It's interesting to see, and flies against conventional logic which assumes that each month's best allocation should win out. It doesn't take into account the effect it has on the early days of the next month. As a result, 85660 is the best performer out of all of them, even though 98482 outperforms in May, Nov and Dec on paper.
This is really neat and exactly what I do. When I find some comparable strats, then I compare them monthly and then I update them based on monthly "win" when compared. Sometimes it's super easy, one month is clearly better than the other. Other times, I really have to play with the data. Nonetheless, the strat you're presenting is quite good...except for the fact that the SD is too high for my typical investing system. I try to keep it under 8, preferably even lower (much more in line with a long-term bond fund).

Since I believe in seasonal investing and cycles and many other things, I've been working on creating a best investment system per presidential cycle using strats with low SDs and that were created before 2021 (preferably much older for reliability). I'm still working on them and do some updating with this information: https://x.com/WayneWhaley1136/status/18 ... 39/photo/1. I'll see where that leads. Either way, nice going! Thanks.

Re: September - October seasonality

Posted: Thu Sep 19, 2024 12:38 pm
by TSP0nlinee
Appreciate you sharing your insights. I'm new to TSPCalc and use the comments here as a starting point. Thanks? KSH

Re: September - October seasonality

Posted: Fri Sep 20, 2024 7:13 am
by Bubba
Aitrus wrote: Thu Sep 19, 2024 7:41 am Ok, I just experimented with something, and noticed an oddity.

I took the best individual monthly performance from 2004-2023 and compared them to each other. Here's the results:

Jan - 85660 (1.25% vs 0.77%)
Feb - tie
Mar - tie
Apr - 85660 (2.48% vs 1.43%)
May - 98482 (1.80% vs 1.63%
Jun - 85660 (1.74% vs 0.92%)
Jul - tie
Aug - tie
Sep: 85660 (1.23% vs 1.16%)
Oct: 85660 (1.69% vs 1.64%)
Nov: 98482 (3.09% vs 2.90%)
Dec: 98482 (2.80% vs 2.73%)

So I took the best month from each one and used it to create a new Mix - 252916.

But the oddity was January. I left January as 85660, but changed Dec to 98482. This changed the first day of Jan from I Fund to G Fund, because 98482 has the Mix going to G on the 20th day, thus being in G on the first day of Jan. The result was that even though the new Mix has 85660's allocations, it returns 0.77% - the same as 98482. The rest of Jan the two Mixes are identical.

So the difference of CAGR by being in G vs I for Jan is a 0.48%. That's a huge amount of impact for a single day of investing. It tells me that even though 98482 has better long-term performance in Dec (2.80% vs 2.73%), it's outweighed by that Jan takes due to being in the wrong Fund at the beginning of the year - even for just a single day.

The other thing I noticed is that even if you take the best individual month from each Mix, it still underperforms 85660 somehow. 252916 (and 252919 is what it looks like when you leave Dec looking like 85660 to get back that 0.48% in Jan) underperforms 85660 by about 1% CAGR.

That's because the same thing happens in June as what happens in January, when the first three of days of the month have the Mix in G instead of S, and the difference has a difference of 0.92% (using G) vs 1.74% (using S).

It's interesting to see, and flies against conventional logic which assumes that each month's best allocation should win out. It doesn't take into account the effect it has on the early days of the next month. As a result, 85660 is the best performer out of all of them, even though 98482 outperforms in May, Nov and Dec on paper.
Another thought is to consider the years of returns. I like 85660 (which I'm assuming is your "permanent" pick??). A different option would be to go by presidential years. 85660 has a leg up on all of them with the exception of midterm years. So, maybe 85660 in all years, midterms 98482?

More work would be to compare midterm years to each other month by month and see which one's strongest and then go from there.

Re: September - October seasonality

Posted: Fri Sep 20, 2024 7:35 am
by Aitrus
Bubba wrote: Fri Sep 20, 2024 7:13 am Another thought is to consider the years of returns. I like 85660 (which I'm assuming is your "permanent" pick??). A different option would be to go by presidential years. 85660 has a leg up on all of them with the exception of midterm years. So, maybe 85660 in all years, midterms 98482?

More work would be to compare midterm years to each other month by month and see which one's strongest and then go from there.
There's several I watch: 85660, 119415, 130870, 163463, 186197. They all follow the same general cycle, with minor differences in spots: G vs F when stocks are bad, using I for a few more or less days in April, that sort of thing.

The first four are the most similar, with 186197 being the one with the most variation from the others. The '04-'23 numbers are pretty good, with 186197 having a SD down in your comfort zone, but it's doing poorly this year.

https://tspcalc.com/seasonal.php?ID=856 ... ears=04-23

It's not that I follow any specific one perfectly, more like I follow a group that have a similar pattern, and use the small differences in changes to give me the most likely options when it comes time to make a decision in conjunction with whatever trend the market is in at the moment.

Ex 1: They all go to S for three days in the middle of October and then to C for the rest of the month? Got it, that's what I'm doing without hesitation.

Ex 2: Back half of July is a mishmash of F / G / I? Well, I guess I have some careful thinking to do, but I'm probably leaning towards F since that's what most of them do (and G for the final couple of days).