Never ending fun

[Bubba]

Recently I've been reading a lot about lookback periods in investing. One source I found suggested that a lookback period of 5 years is a great way to project the next year's returns. So, I played with that on TSPCalc. I tried finding strategies with an SD lower than 8 (so that I could weed out the odd years over 100+ returns). It gave me this strategy...https://tspcalc.com/seasonal.php?ID=267333 which is interesting.

My question to everyone is, has anyone tried doing something like that? You look at a strategy that's created a year prior and then used that strategy for the following year (assuming the 5 prior years were good)?

Thanks!

P.S. If you have a different system, i would be curious to see what you think of using. Many of us have tried different options. Some have systems that work well.

[imktwo]

169608 is the only one I follow that meets the criteria you are looking: 5 year positive returns and SD less than 8.

[Bubba]

169608 is the only one I follow that meets the criteria you are looking: 5 year positive returns and SD less than 8.

Nice! have you tried doing this strategy (only going to a look back period) last year and the year before?

Thanks!

[gbnooks207]

My look back time frame is 2013 to present day. That is when I began using my smartphone to quickly create my IFTs during my work breaks.
With only three months left until I retire, the relevance of SD is less critical to me compared to the high percentage of net returns (PNR) accumulated over the months during that time frame.
For instance, #279140 boasts a PNR of 96%, achieving 148 positive monthly returns out of a total of 153 months. Meanwhile, #279123 has an impressive PNR of 98% (150 out of 153). Could someone clarify why a lower SD is deemed more valuable than a higher PNR? Additionally, is it possible for anyone to develop a model with an SD under 8 that surpasses the returns of #279140 and #279123? I would like to use it. Doesn't a lower SD suggest a higher likelihood of negative monthly returns?
#279123's SD is already 10.9.
https://tspcalc.com/seasonal.php?ID=279 ... 120-279121

[Bubba]

My look back time frame is 2013 to present day. That is when I began using my smartphone to quickly create my IFTs during my work breaks.
With only three months left until I retire, the relevance of SD is less critical to me compared to the high percentage of net returns (PNR) accumulated over the months during that time frame.
For instance, #279140 boasts a PNR of 96%, achieving 148 positive monthly returns out of a total of 153 months. Meanwhile, #279123 has an impressive PNR of 98% (150 out of 153). Could someone clarify why a lower SD is deemed more valuable than a higher PNR? Additionally, is it possible for anyone to develop a model with an SD under 8 that surpasses the returns of #279140 and #279123? I would like to use it. Doesn't a lower SD suggest a higher likelihood of negative monthly returns?
#279123's SD is already 10.9.
https://tspcalc.com/seasonal.php?ID=279 ... 120-279121

Great strategies! Thanks for posting. I'm trying to find something that works from 2004 until now that was created before 2022 (to give it some time to be applicable to later times). Right now I'm playing with ChatGPT (I used to do this all by hand!) to get this organized. My rule is a SD of at least SPX, no negative years and at least a 20% CAGR+ every year. Chatgpt doesn't like these rules.

[gbnooks207]

My look back time frame is 2013 to present day. That is when I began using my smartphone to quickly create my IFTs during my work breaks.
With only three months left until I retire, the relevance of SD is less critical to me compared to the high percentage of net returns (PNR) accumulated over the months during that time frame.
For instance, #279140 boasts a PNR of 96%, achieving 148 positive monthly returns out of a total of 153 months. Meanwhile, #279123 has an impressive PNR of 98% (150 out of 153). Could someone clarify why a lower SD is deemed more valuable than a higher PNR? Additionally, is it possible for anyone to develop a model with an SD under 8 that surpasses the returns of #279140 and #279123? I would like to use it. Doesn't a lower SD suggest a higher likelihood of negative monthly returns?
#279123's SD is already 10.9.
https://tspcalc.com/seasonal.php?ID=279 ... 120-279121

Great strategies! Thanks for posting. I'm trying to find something that works from 2004 until now that was created before 2022 (to give it some time to be applicable to later times). Right now I'm playing with ChatGPT (I used to do this all by hand!) to get this organized. My rule is a SD of at least SPX, no negative years and at least a 20% CAGR+ every year. Chatgpt doesn't like these rules.

Thanks, Bubba! If you discover a winning idea, please share it here. I might also explore ChatGPT as I expect to have more free time soon. I may need to educate myself about market funds and dabble in that area also. But I keep hearing "Safety First!" since my knowledge is limited to TSP.

[Bubba]

My look back time frame is 2013 to present day. That is when I began using my smartphone to quickly create my IFTs during my work breaks.
With only three months left until I retire, the relevance of SD is less critical to me compared to the high percentage of net returns (PNR) accumulated over the months during that time frame.
For instance, #279140 boasts a PNR of 96%, achieving 148 positive monthly returns out of a total of 153 months. Meanwhile, #279123 has an impressive PNR of 98% (150 out of 153). Could someone clarify why a lower SD is deemed more valuable than a higher PNR? Additionally, is it possible for anyone to develop a model with an SD under 8 that surpasses the returns of #279140 and #279123? I would like to use it. Doesn't a lower SD suggest a higher likelihood of negative monthly returns?
#279123's SD is already 10.9.
https://tspcalc.com/seasonal.php?ID=279 ... 120-279121

Great strategies! Thanks for posting. I'm trying to find something that works from 2004 until now that was created before 2022 (to give it some time to be applicable to later times). Right now I'm playing with ChatGPT (I used to do this all by hand!) to get this organized. My rule is a SD of at least SPX, no negative years and at least a 20% CAGR+ every year. Chatgpt doesn't like these rules.

Thanks, Bubba! If you discover a winning idea, please share it here. I might also explore ChatGPT as I expect to have more free time soon. I may need to educate myself about market funds and dabble in that area also. But I keep hearing "Safety First!" since my knowledge is limited to TSP.

So, if i can help you avoid some time wasting here are a few key things I've noticed from ChatGPT that are worth considering:
1. Never use simulations. Only use raw data. Every time that I had it simulate stuff (for a quick and dirty idea of something) and compared it later to the actual data...it was a massive disappointment! I had constructed an investing system that received a 29% CAGR from 2001 to 2025. Then, I had a few additional constraints added with actual data...that went down to 9%. That was a few hours of my time wasted.
2. You set the rules. Be very, very specific. If I let it do whatever, then it finds some random rules and applies them. Then we get junk.
3. Provide the data sets for ChatGPT. yes, you can provide some websites that have data...but the program is pretty inept at getting the information.

What I'm considering now is to just have it create Python code that I run in a separate program to compare what ChatGPT says and Python. We'll see from there. Right now the best CAGR I can get is 16.54%...but that might change...ChatGPT keeps changing things on me.

[Wind]

see below,..

[Wind]

"My question to everyone is, has anyone tried doing something like that? You look at a strategy that's created a year prior and then used that strategy for the following year (assuming the 5 prior years were good)?"

If you do this with 100 different strategies, and find say 1/2 dozen that got that next year 'right', all you've done is found 6 strategies that not only fit the old data but also happen to 'fit' that additional 'new' old data too. OK, let's say you found strategies that worked on a 5 year data set from 7 years ago, and a few of those that also worked on that next year - how many of those would you need to have one that worked on that most recent of past 7 years too, and would that one actually have any chance of predicting the future year?

Hopefully, you see where this is going - You are selectively choosing strategies that happen to 'fit' old data that is badly skewed by non-seasonal effects; new regulations, banks collapse, companies bringing out new pat'd tech, changes in mgt, big lawsuits, 911, bad publicity etc that is hiding any 'seasonal' patterns, and Not discovering hidden seasonality or predicting future data which will be skewed by different non-seasonal effects.

fwiw - shifting money from one fund to another is probly going to be less unfortunate when markets overall are increasing in value, even if not optimum,... enjoy!

But what I like is the diversification strategy of the L funds. To some extent, they are tied into success of the economy at a pretty large scale. I suppose if that economy collapses, money will be lost but you'll probably be more concerned about finding toilet paper and food anyway. I expect mathematicians have poured through information, statistics, knowledge to arrive at the allocation percentages. They give up some of the gains of one fund vs the other and are not as diversified as one might hope but they are slightly damped volatility and look at how 'lucky' the L funds were by having some I fund this year

[Bubba]

"My question to everyone is, has anyone tried doing something like that? You look at a strategy that's created a year prior and then used that strategy for the following year (assuming the 5 prior years were good)?"

If you do this with 100 different strategies, and find say 1/2 dozen that got that next year 'right', all you've done is found 6 strategies that not only fit the old data but also happen to 'fit' that additional 'new' old data too. OK, let's say you found strategies that worked on a 5 year data set from 7 years ago, and a few of those that also worked on that next year - how many of those would you need to have one that worked on that most recent of past 7 years too, and would that one actually have any chance of predicting the future year?

Hopefully, you see where this is going - You are selectively choosing strategies that happen to 'fit' old data that is badly skewed by non-seasonal effects; new regulations, banks collapse, companies bringing out new pat'd tech, changes in mgt, big lawsuits, 911, bad publicity etc that is hiding any 'seasonal' patterns, and Not discovering hidden seasonality or predicting future data which will be skewed by different non-seasonal effects.

fwiw - shifting money from one fund to another is probly going to be less unfortunate when markets overall are increasing in value, even if not optimum,... enjoy!

But what I like is the diversification strategy of the L funds. To some extent, they are tied into success of the economy at a pretty large scale. I suppose if that economy collapses, money will be lost but you'll probably be more concerned about finding toilet paper and food anyway. I expect mathematicians have poured through information, statistics, knowledge to arrive at the allocation percentages. They give up some of the gains of one fund vs the other and are not as diversified as one might hope but they are slightly damped volatility and look at how 'lucky' the L funds were by having some I fund this year

Paul Merriman has suggested to go to 50% S and 50% L (based on year you retire). I'm not so sure. In later periods of a business cycle, the C Fund tends to outperform, which has been the case for a while. In any case, it's not a bad set it and forget it style. Thanks!

Oh and I do agree with what you said. I've tried the 5 year rule, but then you get like 100s (if not 1000s) of options. Do they actually all work? Not really. Good luck to all!

[Wind]

Paul Merriman has suggested to go to 50% S and 50% L (based on year you retire). I'm not so sure. In later periods of a business cycle, the C Fund tends to outperform, which has been the case for a while. In any case, it's not a bad set it and forget it style. Thanks!

The 50% S, 50% L would of course be tied more heavily into performance of S stocks. S (and/or C if that was the choice) has good long-term results, and often has good short term results too, but C, S, and I have volatility and sometimes loses substantial percentage of value, and may take years to recover those losses. Those losses may be so substantial, that someone drawing income from their holdings (presently or in near future) may not recover the losses in their lifetime and have to reduce their annual income/withdrawal to avoid depleting their account. I believe 'that' is the premise of L funds increasing the percentage allocated to G over time, i.e., a continuous re-balancing of risk vs greed in view of the statistically-allowable tolerance for recovering from loss.

So that covers 'seasonal investing' and 'set it/forget it'.

There is of course other strategies such as 'market timing', ie buy low/sell high, or sell high/buy low. But either way - the profitability of market timing requires two correct decisions, i.e., when to buy and when to sell. To do that requires luck and/or knowledge and if it was easy - everyone would be doing it and not getting milked by those with greater knowledge and buy/sell flexibility.

* While the TSP participant might choose any 'set it/forget it' approach including L funds for example, the L funds are not internally managed as 'set it/forget it' - changes in allocation over time will happen and even though their past and present allocations are known and disclosed, I believe their future allocations will be continuously re-evaluated and adjusted according to the existing conditions by knowledgeable managers.

[Aitrus]

Good points all around. I've always been of the opinion that buy-and-hold is a solid approach to investment. Seasonal is just another option that can be used. That said, I have two big gripes with the L Funds.

The first is that they are too conservative for the military members and civil servants that utilize them. This is because they mirror allocations that are common practice in the private sector. However, whomever decided to use those allocations for the L Funds neglected to factor in a critical aspect: pensions.

The L Funds would be a decent allocation setup for someone without a pension, and who needs to have some portfolio amount invested into a stable asset in order to smooth out the volatility. But for someone who is contributing towards a pension, the pension itself acts as the smoothing out action of that person's net worth because the money paid to the pension is money that would otherwise have gone into the investment portfolio.

Since there's less money going into the portfolio, the allocation needs to make up the difference of a smaller account by being more aggressive to make the most of the money available. The L Funds allocate far too much to G and F once you take into account that every single TSP retiree will be getting a pension of some kind. By allocating to much to G and F in such a redundant manner, there's lots of lost opportunity for any investor who uses it.

Right now in 2025, the L 2050 allocates 18% towards the G and F Funds. At 25 years out from retirement? That's too conservative - especially if someone is just starting their civil service career and wants to retire with a 25 year retirement in 2050 (a 20+ year second career retirement is common for ex-military who transition to civil service). Fortunately, the L 2075 Fund changes this. The G and F Fund allocation is about 1%, and doesn't change until 2047. By 2065 - 10 years out from retirement - the allocation sits at around 30%. Too high for my taste, but still reasonable.

But that brings me to my second gripe: the overemphasis on the I Fund. The I Fund is a notorious underperformer. Sure, it has it's moments (like this year), but when was the last time in the last 20 years that the I Fund was the Belle of the Ball to such an extent that it was worth allocating more than about 10% to it? Answer: 2004 - 2006, and 2017. Aside from those years, investing in the I Fund was a losing game compared to either the C or S Funds. And in those years, the C and S Funds still did reasonably well (except for 2005, when the C Fund was just "meh" but the S Fund still did well).

The L 2075 fund allocates about 35% to the I Fund. That amount won't drop until about 2047. Then it gradually drops down to about 20% in 2070. That's a LOT of years of underperformance for 20-35% of someone's retirement money.

In contrast, the L 2075 has a piddling 13% (12.85%, actually), and staring in 2047 begins to drop down to a tiny 8% by 2070. This is despite the fact that the S Fund has outperformed the I Fund in 13 of the last 18 years (in the last 20 years, two years had them within 1% of each other, so effectively a tie after a full year of market action). At worst, the S Fund and I Fund should be an even allocation in the L Funds (preferably an edge would still be given to the S Fund), but it's not even close to being that. It's lopsided in a way that hurts the investor.

Bubba mentions Merriman's recommendation of 50% S, 50% Lifecycle. I view this as a reasonable compromise to both of the problems I have with the L Funds. It reduces the amount allocated to G, F, and I by half and puts money where there's a distinct lack of allocation: the S Fund.

I usually suggest a 50/50 or 60/40 split of C/S for new TSP members who just want a "set it and forget it" approach in TSP who have a time horizon past 10 years. If they want something a bit more conservative I'll suggest a 45/45/10 or 60/40/10 split into C/S/F. I think I'll add the Merriman approach as well.

[Scarfinger]

I don't have the time or discipline to follow the seasonal/daily strategies. I have deferred to an L-fund 20 years past my expected retirement date in 2030 (in the L2050 currently). So it will basically be what the 2045 is today: 77 equities / 23 G&F when I retire.

2 pensions and SS will cover my bills, house paid off. So I plan to keep the TSP fairly aggressive.