November 12, 2025 – Hello, readers! Welcome to another installment of the Safe Withdrawal Rate Series. Please see this landing page for an introduction to the Series and a summary of all the other parts so far. After a long hiatus from writing due to my busy travel schedule during the summer and lots of other commitments, I’ve found my groove again and put together something that has been on my mind for many years: Is there an asset allocation strategy that could have improved historical safe withdrawal rates? Specifically, could we devise an asset allocation strategy that shifts weights between different asset classes in a way to improve investment results? Of course, that’s easier said than done, but there are some interesting ideas out there. One such approach is to tactically shift asset class weights based on asset return momentum. Some people also refer to this flavor as “Trend-Following.” If you want to sound really techy and fancy, you’d also call this “Tactical Asset Allocation” (TAA), “Managed Futures,” or “Commodity Trading Advisers” (CTA) strategies; however, these three terms often encompass many other dynamic asset allocation strategies, not just momentum.
In any case, maybe a momentum strategy can help us avoid some of the worst historical asset market disasters if we could sell equities early enough during a bear market. How much Sequence Risk could we eliminate? By how much can we raise our safe withdrawal rate if we could have reliably avoided some of the worst historical asset market disasters? Let’s take a look..
Before trying Momentum, what else could hedge Sequence Risk?
Before we even jump into momentum, is there anything else worth pursuing? If you’ve followed my Safe Withdrawal Rate Series, you’ll remember that I’ve taken a very pessimistic view on the efficacy of all the proposed asset allocation tricks to improve safe withdrawal results:
- Dividend focus? Doesn’t work. See Part 29, Part 30, and Part 31 of the series.
- Small-Cap Value? Doesn’t work. See Part 62 from earlier this year and also “Small-Cap Value Stocks: Diversification or Di-WORSE-fication?” from last year.
- Small-cap + Micro-cap + International stocks, as proposed by Bill Bengen recently? I’m highly skeptical, as I’ve detailed in this short guest post on ChooseFI. It’s also part of ChooseFI Ep. 563.
- Risk Parity? I haven’t written about it, but I plan to have a designated post on why that’s also a bad idea.
What most of these strategies have in common is that they primarily pursue a fixed asset allocation with slightly different weights than the traditional 60/40 or 75/25, plus a few additional asset classes. Well, there is no free lunch: Only total returns matter, and a higher dividend yield lowers your expected price return. Stock-picking styles like Small-Cap, Value, and Small-Cap Value are no longer a reliable source of alpha, now that everyone is aware of them. Risk Parity is such complete charlatanry that I’m not even going into the weeds here. Stay tuned for a future post.
So, consistently holding the existing asset classes available to the average retail investor is not that fruitful in expanding the investment horizon. Then, one new route we should explore is the “smart” way of shifting between the different asset classes. One such approach is momentum.
Why would asset market momentum work?
In a nutshell, asset market momentum looks more promising than the previously proposed solutions from the mutual fund salespeople because…
- Asset returns depend on economic regimes, such as recessions versus expansions, inflationary versus deflationary regimes, tight versus loose monetary policy, as well as commodity cycles of abundance versus scarcity, which can take time to return to normal.
- Economic regimes persist for extended periods. For example, the probability of being in a recession next month depends on the current economic state: you’re more likely to be in a recession next month if we’re in a downturn this month. The Federal Reserve often goes through a whole sequence of rate hikes or cuts. There are production and harvest cycles in many commodities. And so on.
Let me justify both points really quickly with some empirical evidence. First, average asset returns vary significantly across different economic regimes, specifically between NBER expansions and recessions, as shown in the table below.
- Equities tend to perform well during expansions, offering high returns and low risk. However, they suffer during recessions, experiencing negative returns and almost twice the volatility compared to an expansion.
- Intermediate U.S. Treasury bonds do very well during recessions. But they still have decent returns even during expansions. In fact, even during expansions, they have a tiny edge above cash/T-bill returns (3.87% vs. 3.29%).
- A comprehensive index, such as the GSCI index, which encompasses all major commodities, follows a similar pattern to equities, yielding significant returns during expansions. But poor returns and high risk during recessions.
- The Gold spot price performs well across all economic regimes, but it has a particularly attractive return profile during recessions. Not as good as Treasury Bonds, though. Also, notice that Gold is a member of the GSCI Commodity Index, so it’s captured in the Commodities return series. But gold certainly has very different return characteristics from the others!
- Cash, such as 3M T-bills or related assets like money market accounts, performs well under all regimes. Even slightly better during recessions, which would make it an ideal risk-off/safe-haven asset during economic turmoil. Gold indeed performs somewhat better on average during a recession (7.27% vs. 4.08%), but at the cost of significantly higher volatility.
I also like to include a few more fun facts. Most asset class returns correlate not just with the current month’s economic regime. The correlations are even stronger if we offset the economic regime by a sufficient number of months. For example, equity returns correlate best with the economic status four months in the future, as shown in the table below. That’s perfectly intuitive: the equity market leads the economic cycle. Commodities by about 1-2 months, gold by one month. Also noteworthy is that the short-term return, closely tied to the Federal Reserve policy rate, leads the cycle by the longest, approximately 6 months.
Next, let’s look at some evidence of U.S. economic momentum. If we examine the NBER expansions versus contractions over time, we find that over the last 100 years, we have been in a recession about 16.8% of the time; however, the recessions have been clustered. Conditional on being in an expansion in one month, we had a 98.4% chance of continuing the expansion next month. Likewise, if we were in a recession in one month, we’d stay in a recession with 92%. The “escape probabilities,” i.e., the chance of changing the economic regime, are 8% from a recession, implying a roughly 13-month average recession length, and only 1.6% from an expansion, implying a roughly 62-month average expansion length. And just FYI, since 1970, the recessions have been slightly shorter and the expansions have been significantly longer, more than 80 months.
But isn’t the market efficient?
One objection I can already predict: if there is this predictable and reliable market return pattern, why isn’t this all priced in immediately? My response: Nobody knows in real-time whether we’re in a recession or an expansion. It’s also true that around the turning points, the market moves very rapidly, and prices in a significant portion of the expected outperformance of the new expansion or the underperformance of the impending recession. But certainly not all. For example, after the stock market trough in early 2009, the S&P 500 rallied by over 50% (from February 2009 to February 2010). However, it still left enough runway for the massive rally, which lasted until February 2020. Therefore, the positive news of a new and developing expansion and bull market will filter in gradually over time, which is one of the main reasons for asset return momentum.
Designing a Momentum Strategy
There isn’t one single way to design a momentum strategy. I will propose one that makes sense to me. It will not be the most effective momentum strategy available. However, keep in mind that alternative strategies with numerous bells and whistles are likely to suffer from in-sample bias; that is, someone may go to great lengths to maximize performance, making the backtest appear good, but at the risk of overfitting and potentially creating worse results out of sample.
Momentum assets
For the time being, I propose a momentum asset allocation strategy that utilizes four major asset classes: Equities (S&P 500 Total Return Index), U.S. Intermediate-Term Treasury Bonds (10-year Treasury Benchmark Total Return Index), Gold Spot Price, and 3-month T-bills. I will simulate returns since January 1871 through June 2025. Also, note that I use the returns from my SWR Google Sheet. See Part 28 of the series for the link to the sheet and for a guide. Notice that I use only monthly return data. In recent decades, we certainly have all series in daily frequency, and we’d be able to design a much more intricate momentum signal. But early in the sample, I only have monthly data, so I stick with the monthly frequency.
Momentum assumptions
There is no shorting allowed. Clearly, we could improve the results by allowing short positions of those asset classes that find themselves in a negative momentum spiral. Professional investors, like CTAs/Commodity Trading Advisors, will create the desired long and short positions through futures. However, the average retail investor will be hard-pressed to effectively and affordably short the major asset classes. Yes, there are some short-exposure ETFs, but their expense ratios and other costs appear excessive; more on that later.
I target a base asset allocation of 70% Equities, 20% Bonds, and 10% Gold in case all three asset classes display positive momentum. Any remaining portfolio weight not allocated to the three major asset classes is in the Cash bucket.
I calculate the momentum at three different horizons: 8, 9, and 10 months. Why those horizons? That’s precisely what experienced momentum traders tend to target. Any shorter horizon will create too many false alarms, and any longer averages will cause the momentum signal to miss the boat around the turning points. I also use multiple horizons to increase robustness and potentially spread out trades around the turning points.
I also calculated the momentum signal in two different ways. Notice that we can compare the most recent index level to the rolling average of index levels. Or, we can compare today’s index vs. the index N months ago. There’s also one additional twist: for the equity momentum signal, I compare the average of the latest two monthly index levels with the rolling n-month average. This is a bit of inside baseball, but equity returns are very slightly negatively correlated between two adjacent months (i.e., negative serial correlation). To smooth out that short-term volatility, I take the average index level over the last two months. Think of that as the monthly frequency equivalent of a 50-day over 200-day rolling average crossover signal.
Next, there is an additional design choice: Do I take the raw indices or the excess return over the cash return series? Tough choice! So, I calculate both. On the one hand, many momentum traders simply look at the raw index series. But on the other hand, I can also justify that if Cash is a fallback asset class, then we should use the excess returns over cash as our target series.
Summary so far: for each asset class, I generate 12=3x2x2 momentum signals, which is the product of three momentum horizons (8,9,10 months), two momentum formals (rolling average crossover versus the T-N to T asset return), and two index versions (total return versus total return in excess of the cash return series).
How to translate the momentum signal into asset class weights
I calculate each of the 36 momentum signals, 12 for each asset class, and assign 1.0 for a positive momentum raw signal and 0 for a negative raw signal. Then average over each asset class to get a raw momentum signal between 0 and 1 for equities, bonds, and gold, respectively.
Another crucial assumption is how the negative momentum weights are redistributed to the other asset classes. The most straightforward approach would be to transfer all unused equity, bond, and gold weight directly into the cash bucket. But that would be pretty inefficient. Why shift into the cash bucket if there might be other, more productive and profitable assets with positive momentum? So, I propose the following sequential reshuffling of the unused asset class weights: The weight not used by gold is added to the equity base weight. The part not used by equities is added to the bond base weight. Finally, the portion not used by bonds goes to the cash bucket. Here’s an example: Imagine that gold has a momentum signal of 0.25, equities have 0.75, and bonds have 0.50. Here would be the asset allocation weight calculation:
- Gold gets a weight of 2.5%, which is 0.25 times the base weight of 10%.
- Equities get a new, adjusted base weight of 70% plus the leftover from the gold allocation (7.5%), so 77.5%. Since the momentum signal is 0.75, equities get 58.125%. The remaining 19.375% carries over to bonds.
- Bonds receive a new base weight of 20% plus the 19.375% from equities, for a total of 39.375%. However, the momentum signal is only 0.50, so only 19.6875% of the allocation goes to bonds.
- The remaining 19.6875% shifts to cash.
Other details
I assume the following annual expense ratios for the different asset classes:
- Equities 0.03% (e.g., iShares IVV or Vanguard VOO)
- Bonds 0.15% (e.g., iShares IEF)
- Gold 0.09% (e.g., iShares IAUM)
- Cash 0.09% (e.g., Shares SGOV)
Furthermore, I assume that each transaction incurs a 0.03% drag on performance for each asset class due to commissions, bid-ask spreads, and other related costs. For example, shifting 40% of the portfolio from equities to bonds will trigger a fee of 2 × 0.4 × 0.03% = 0.024%, i.e., the t-cost hits both the bond and equity sides, hence the 2x. I was concerned that turnover costs would undermine the returns in the simulations, but I was pleasantly surprised: the drag from turnover was only about 0.09% per annum. Well, that’s assuming today’s financial technology and low or even zero-cost commissions at Fidelity and Vanguard. An actual retiree in 1929 would have paid more in transaction costs back then! Therefore, as always, we should acknowledge that these simulations involve a thought experiment that utilizes today’s financial technology, while assuming that some historical return sequences may repeat in the future. Under no circumstances would I claim that a 1929 retiree would have gotten exposure to the US stock market with just a 0.03% annual expense ratio and 0.03% transaction costs back then!
Momentum Signal Time Series
Let’s plot the asset class weights during a few interesting time periods. Let’s start with the period from 1925 to 1945. The strategy successfully shifted out of equities in late 1929, with only a few false alarms in between. For the most part, we were heavily invested in Treasury bonds, which performed very well during this deflationary period. Then, they returned to equities in 1933 to capitalize on the subsequent recovery. The reason Momentum did so well was that the drawdown was long and deep enough that even with a bit of a delay, you still would have stayed out of the equity market for much of the economic and financial malaise. Also, notice that gold wouldn’t have been a very prominent part of the portfolio during this period.
Next, let’s examine the period from 1965 to 1985. There were several periods when you shifted out of equities and straight into cash because bonds were also suffering during this inflationary period. Gold played a more prominent role than in the 1930s. Momentum was a useful tool during this period!
Finally, let’s examine the period from 2000 to 2025. We started with the backend of the 1990s equity bull market but quickly shifted out of stocks and into Treasury bonds during the dot-com bust. Treasuries performed well because you captured the long path down in yields, which generated excellent bond returns from the duration effect. In 2003, you returned to equities and rode that positive momentum until the onset of the Global Financial Crisis (GFC). Momentum also navigated the Global Financial Crisis very well, returning to Treasury bonds until 2009. The impressive equity bull market that followed 2009 was clearly interrupted here and there with some false alarms; i.e., we briefly dropped equities only to jump back in again after a few months. We also did quite well during the pandemic (only dropped equities very briefly) and the 2022 bear market, where we nimbly shifted into all cash because all risky assets hurt back then. Sweet!
Gold had a relatively consistent allocation of 10%, with only a few exceptions, which proved to be the right decision. This was because gold was one of the best-performing assets since 2000, outperforming the S&P 500.
Momentum Strategy Simulation Results
The momentum strategy results look pretty promising. Here’s the cumulative performance, adjusted for US CPI inflation. I plot three return series:
- A portfolio with static weights of 70% equities, 20% bonds, 10% gold, and no cash allocation, which are the base weights targeted in the Momentum Strategy if all three signals are positive.
- A portfolio with static weights equal to the overall average weights of the momentum strategy, which was roughly 52% equities, 27% bonds, 3% gold, and 18% Cash.
- The momentum strategy with the dynamic tactical asset weights.
I also plot the excess performance of the Momentum strategy over the two static strategies with fixed weights. Notice that the vertical scale is in logs to make the growth rates over time easier to see. The Momentum results look really impressive: It left the other two strategies in the dust! Granted, most of the outperformance relative to the 70/20/10/0 portfolio occurred during the period from 1871 to about 1940. However, the path of the momentum strategy was also much smoother post-World War II. Also, the higher average weight in equities helped the 70/20/10/0 portfolio. The asset allocation alpha has remained relatively consistent in the post-World War II period, as indicated by the purple line.
In the following chart, I plot the drawdowns of CPI-adjusted returns of the three asset allocations. The 70/20/10/0 strategy typically has the deepest drawdowns, while the momentum strategy usually has very shallow drawdowns. Well, not all the time! One fly in the Momentum strategy ointment: it had worse drawdowns during the 1940s. It’s understandable because asset returns during wartime may swing too rapidly for a momentum strategy to capture correctly. So, the WW2 performance is a bit of a downer (though, surprisingly, Momentum did well during WW1, which is a bit of a surprise). This will also have some unpleasant consequences for the SWR analysis later. The momentum strategy worked remarkably well over the last 30+ years, only barely exceeding 10% drawdowns, while the static allocation strategies suffered significant drawdowns of 30% or more during the dot-com and Global Financial Crisis. Notably, momentum performed well during the most recent downturn in 2022, with only a 12% drawdown, while the static allocations experienced drawdowns of 20% or more. This is impressive, considering that the 2022 mini bear market was very short-lived, which is not exactly where an 8- to 10-month momentum strategy is guaranteed to work well.
In the table below, I present the detailed return stats. There are two panels: the top displays the nominal returns, and the bottom shows the real, CPI-adjusted returns. Momentum outperformed all individual asset classes and the strategic asset allocations, including a 75/25 portfolio, which is my preferred retirement asset allocation at present. Momentum generates average returns that are about in line with equities, but with vastly smaller volatility and drawdowns.
Here is the same table for the last 30 years (January 1995 – September 2025). Notably, since 1995, the Momentum strategy has achieved an impressive 0.95 Sharpe Ratio using nominal returns. Qualitatively, we obtain the same results: Momentum performs well, within 100 basis points of annual equity returns, but with almost 50% less risk. Also, the drawdowns were very manageable. Note that when computing the “strategic” or fixed asset allocation, I use the average weights from this 30-year period, which came out to be 57% equities, 25% bonds, 6% gold, and 11% cash (totals don’t add up to 100% due to rounding).
Testing the Momentum Strategy in the SWR Google Sheet
How do I add the momentum signal returns to my Google Sheet? That’s amazingly easy: In my sheet, I have space for two “Custom Series” in the Asset Return tab. I currently store the momentum series in the Custom Series 1 column, which grabs the cumulative returns of my simulated series in column K in the “Asset Returns” tab.
I simulate four different asset allocations:
- My usual baseline is 75% equities and 25% Intermediate Bonds.
- The base weights form the momentum strategy, i.e., 70% equities, 20% bonds, 10% gold, and thus 0% cash.
- The Momentum strategy. I assume that 100% of the portfolio follows this approach.
- Half of the portfolio follows a momentum strategy, while the remaining 50% remains invested in equities. I use this as a proxy for investors who may have a constraint on how much of their assets they can tactically shift around. Think of an investor who prefers to maintain a floor of 50% of their portfolio in equities in a taxable account but can allocate the remaining assets across tax-advantaged accounts without incurring significant capital gains taxes.
Furthermore, I study two different retirement horizons: one with the standard 360 months horizon (e.g., a traditional retiree) and one with 600 months, i.e., the typical early retiree. Let’s assume our retirees are OK with asset depletion, i.e., a final portfolio value of $0.
In the main tab, I would enter the 100% Momentum strategy as shown in the screenshot below. Note that I set the expense ratio to 0.00% because all the momentum returns are already net of expense ratios and trading costs.
What effect does momentum have on the fail-safe withdrawal rates? As usual, I report the failsafe withdrawal rates by decade, so we can get a sense of how the different asset allocation strategies performed during the historical worst-case scenarios.
Let’s start with the 360-month retirement; see the chart and table below. Just as a refresher, even the baseline 75/25 portfolio (or any other Stock/Bond or Stock/Bond/Cash combination) had a safe withdrawal rate of slightly less than 4%; here it’s 3.82%. Not 4% as often falsely claimed. So, the 4% Rule is not 100% but only about 99% safe. See Part 1 of my series. However, the good news is that 4% Rule failures are rare. Only in the most outrageous financial crashes (1901-1904, 1929-1933, 1968-1982) would the 4% Rule have failed.
The Momentum Strategy performed remarkably well during the 1900s and 1920s, achieving a SWR of over 5%. Even in the 1960s, you reached 4.79%. As mentioned above, there is one fly in the ointment, namely the 1930s and 1940s. If you had retired right before the 1937 market peak, your SWR would have been only 4.07%. That’s the overall worst-case retirement cohort for the Momentum Strategy. The reason is that between 1938 and 1945, the Momentum strategy completely messed up the equity timing, as some of the large drops lasted only one or two months. That’s too short to capture with Momentum. Oh, well, you can’t win all the time!
Notably, the 50% Stocks + 50% Momentum mix performed well. You didn’t quite reach the 5%+ in the 1900s and 1920s, but you also avoided some of the equity market whipsaw in the 1930s and 1940s. If we examine the overall worst-case scenario for the 50/50 mix strategy, it remains the August 1929 cohort, but with a significantly improved SWR of 4.24% compared to 3.84% for the naive 75/25 portfolio. That’s more than a 10% increase in the retirement budget. Nothing to sneeze at!
Next, let’s move on to the 50-year horizon. Notice that I include the decades only up to the 1970s to capture the 1972 scenario. But the 1980s and 1990s don’t have a full 50 years of return data. As always, extending the horizon lowers my withdrawal rates. The overall worst-case scenario for the 75/25 portfolio was now in the 1900s (3.35%), closely followed by the 1920s (3.39%), 1960s (3.43%), and the 1910s (3.56%). In each of these decades, a Momentum strategy would have vastly improved your results and afforded you a SWR significantly above 4%. Unfortunately, we have that unpleasant 1937 cohort again, which was whipsawed during the 1938-1945 equity volatility. Aside from that one little mishap, the Momentum strategy trounced the 75/25 and the 70/20/10. Just like in the 30-year case, mixing the Momentum strategy half-half with an all-equity portfolio would have done very well in the 1930s and 1940s, though you would have slightly lagged behind the exceptionally strong performance in the other decades. So, the inflexibility due to the taxable account could have been a blessing in disguise: The half-half strategy had the overall best failsafe withdrawal rate: 3.86% relative to the 3.35% in the 75/25 portfolio. That’s a 15% improvement in the annual retirement budget.
What about Momentum ETFs?
Implementing and regularly monitoring this momentum strategy, while not exactly rocket science, may still be too much for some investors. So, are there any ETFs that can perform the same task? Note that an ETF would also have a significant tax advantage: you’d keep the realized capital gains within the ETF wrapper and only owe capital gains taxes when you sell the ETF itself. Though the ETF would still regularly pay out taxable dividends, of course. Moreover, ETFs can likely do more advanced momentum strategies, for example, shorting some indexes or commodities through futures. But will a Momentum ETF do better? I have my doubts! Several ETF offerings pursue so-called managed futures strategies, i.e., run a momentum strategy implemented through futures contracts:
- iMGP DBi Managed Futures Strategy ETF (DBMF) –
- WisdomTree Managed Futures Strategy ETF (WTMF) –
- First Trust Managed Future Strategy ETF (FMF) –
- KraneShares Mount Lucas Mgd Futs Idx Stgy ETF (KMLM) –
The last one has only a very short history, but the other three at least go back to the pre-pandemic era. Let me calculate the return stats since at least the DBMF ETF, which started in 2019. Additionally, to address the brief history of the KMLM, I have filled in its early return data for 2019 and 2020 using the DBMF returns. Here are the return stats, all in CPI-adjusted real returns. As standalone strategies, most, if not all, of the Managed Futures funds were subpar. The WTMF, FMF, and KMLM returns were just barely above inflation. Even the DBMF produced only 3.35% annualized real returns, which is a bit thin in retirement. I should also point out that I’m not cherry-picking the time period because the WTMF and FMF have returns going back further; however, the results were even worse before 2019.
Here’s a chart with cumulative real returns, i.e., adjusted for U.S. CPI inflation. One quick comment: I hope you recognize why I took the liberty of backfilling the KMLM returns with the DBMF returns in 2019-2020; KMLM and DBMF seem to have very similar trend-following models. They are highly correlated during the period when both have data, albeit the KMLM appears to have a consistent drag relative to the DBMF, so the KMLM returns might have looked even worse if they had started earlier.
In any case, DBMF and KMLM indeed performed admirably well until the Fall of 2022; they must have quickly identified the 2022 bear market and shorted most asset classes with negative momentum, hence the strong positive performance in the spring of 2022. Of course, the price you pay when shorting risky assets is that around the turning point, you may face some serious pain. All the impressive gains from shorting assets during the 2022 bear market evaporated again within a few months when the market turned around more quickly than the trend-following models could reverse their positions from negative to positive. The Lord Giveth, The Lord Taketh! Neither DBMF nor KMLM is even close to its prior high. WTMF and FMF also look pathetic. In contrast, the ERN Momentum strategy just chugs along from one high to the next. All this, despite (or perhaps because of) being a long-only strategy, with no shorting allowed.
And, finally, the drawdowns. KMLM is a complete clown show. DBMF also had close to 20% drawdowns. That’s less than the equity market in 2022, but not much. FMF had only modest drawdowns, but it’s also the ETF with the lowest average returns.
So, my verdict here: Stay away from most of the Managed Futures ETFs. They only enrich the financial services industry. If you like this momentum approach, you’d likely have to implement it yourself.
If you absolutely want a momentum strategy, the best option currently available may be the DBMF, but it has not yielded particularly impressive returns in absolute terms. It only outperformed three absolutely terrible competitors. That said, DBMF had a pretty neat equity correlation and beta during this period (-0.19 and -0.12, respectively). DBMF is not intended to be a complete replacement of the entire portfolio, as we likely want net-long equity exposure in the long run. So, perhaps a portfolio with some consistent equity exposure, around 70-80% plus DMBF as a diversifier, would have been a good approximation to my momentum model. I will research this further.
Conclusion
There you have it. A momentum strategy, while not a perfect hedge for Sequence Risk, could be the key to a slightly safer retirement. Even a relatively naive and mundane equity/bond/gold momentum strategy outperformed a static asset allocation, not just in plain buy-and-hold return stats over the last 150+ years, but especially in safe withdrawal rate simulations for most of the historical retirement cohorts. The reason is that you achieve average returns almost as high as an equity portfolio, but with significantly lower volatility and far lower drawdowns. That’s the secret sauce you need to mitigate Sequence of Returns Risk.
What I found most noteworthy is that the momentum signal has consistently helped you over the last 150 years. Sometimes, you test a strategy and it works well during the first half, but then fizzles during the second half. Not so for momentum. It has been remarkably consistent across all major asset market cycles, particularly in the equity market volatility of the last thirty years: the Dot-Com bubble, the Global Financial Crisis, the Pandemic, and even the relatively short and shallow, garden-variety bear market in 2022. If anything, Momentum became more reliable recently. For full disclosure, the major risk would be a repeat of the jagged equity performance of the 1938-1945 period. However, if we have confidence that future recessions and bear markets are more in line with those of the last thirty years, you might find this momentum approach appealing.
Would I switch my portfolio to this momentum strategy now? I will certainly monitor the momentum signals in the future. I wouldn’t liquidate any of my Fidelity equity mutual funds in taxable accounts, but I may shift around some funds in tax-advantaged retirement accounts. If there is demand, I may even update the momentum signals regularly and post them either on my website or in the Google SWR spreadsheet. Please let me know if people are interested in that kind of information.
Update (11/14/2025): It appears there is demand. I have posted a CSV file with the most recent signal statistics on my Google Drive here. I also added the same info in a new tab in my Google SWR Sheet. I will try to run my Python code (almost) every day.
Also, one alpha source doesn’t rule out another. I still employ my options trading strategy, which adds about 4-5% annualized returns in my taxable account (about 1.6% spread over all accounts). Maybe combining the two would be the ultimate Sequence of Returns Risk killer!?
OK, we’re at 5,000 words now. If you read this far, kudos to you! Happy retirement, everyone!
Please leave your comments and suggestions below! Additionally, be sure to explore the other parts of the series; see here for a guide to the different parts so far!
Title Picture Credit: WordPress AI + ERN edits
Technical Appendix:
A few short technical notes that didn’t make it into the post
- Of course, I played around with many other momentum construction rules. For example, instead of the hierarchical rule gold -> equities -> bonds -> cash, simply allocate all unused weight to cash. And other rules. All methods I tried yielded very similar results, with Sharpe Ratios differing by no more than 0.03 from the preferred method I eventually used. I don’t include all those additional simulation results in the post because we’re already exceeding 5,000 words.
- I also simulated returns for a scenario in which we delay the momentum signal by one full month. That’s to address the concern that you would have traded at the market close every month, but waiting until you know the market close to plug that into your momentum signal, it would be too late to trade at that point. I found that even if I trade this momentum strategy with a 1-month delay, you would still have done okay. You face a deterioration in the Sharp by about 0.07 to 0.09. The drawdowns were still about the same. In practice, of course, you wouldn’t wait a full month. You might trade on the first trading day of the next month or even measure the momentum signals a few minutes before market close and get your trades executed that same day. But having worked in tactical asset allocation many years ago, I learned that it’s always a good exercise to see how quickly your signals deteriorate. This signal seems to be very robust to delays.
That confirms my suspicion that Frank Vasquez, Michael Covel and Meb Faber are charlatansz
I’m with you. I stopped listening to Vasquez’s podcast after he basically claimed the other day that he’s the only smart DIY investor and everyone else is clueless. Ego bigger than the planet, that guy.
Yeah, he used to be funny. But his favorite soundbite…
… sometimes applies to his own rants. Ironic.
Frank doesn’t make cash on his predictions, though. So I think lumping him is a bit overstated.
Good point. He’s a true believer, I give him that much!
Haha! Thanks for weighing in. I will not comment on individual
charl…., uhm, I mean personal finance contributors.ahaha I saw what you did there and I applaud. I guess we’re not part of his “top drawer” audience – gladly !
You’re very perceptive! 🙂
Frankie boy doesn’t hold Corporate Bonds…but held TLT (-31%) all through 2022. So glad I was in Cash equivalents.
Yeah, that one hurt. Worse drawdown than stocks.
Thanks Karsten and please post the Momentum Signals on your website.
In the SWR toolbox please! 🙂
Yes, please. SWR toolbox would be great. Love the typo at the very end: “see how quickly your sgianls deteriorate.” Is that german for “I’ve just typed 5000 words and I’m tired?” Thanks, Karsten, for all your detailed analysis and calling out the snake oil salesmen.
Haha, that was last night at 10:30 PM. Thanks for the catch, I made the correction. I guess even my spell-checker was tired by that time.
Trying my best.
Thanks! I will try to put this together in the SWR sheet. Stay tuned!
Absolutely – well worth the wait Karsten, thank you. I’m slightly taken aback by how effective the strategy has been. And I’m not sure you were asking a real question about whether we want to see the signals. Of course we do!
Thanks! Posted on the Google Drive: https://drive.google.com/file/d/1UuQkgrLpnSbSuKpZmRZ761TYrhbmnky9/view?usp=sharing
Tx for the post. Very interesting reading. I am always on the look out to sweat the SWR. Is it not possible for you to create a momentum fund based on your rules that we could buy into. Aside: please do post the momentum signals in case I am brave enough to DIY this with more research and knowledge.
Setting up and running an ETF is quite expensive. Probably at least $250k per year. But then again, if we could collect $1b in AUM quickly this would certainly work.
I’m grateful for the note at the end:
> All methods I tried yielded very similar results, with Sharpe Ratios differing by no more than 0.03 from the preferred method I eventually used.
I was about to criticize the seemingly cherry-picked algorithm of overflowing weights, as the gold falling through to stocks and so on sounds very arbitrary. The natural way to do it would be e.g. using cash as the only fallback. Glad to hear this would work fairly similar!
Well, I sorted asst classes by vol and then trickled down into the lower-vol asset classes. Other methods are available and I can study this more nd see if I can come up with anything better. But at some point this becomes an overfitting exercise, I fear.
It’s worth noting that most “momentum” ETFs focus on selecting individual stocks based on recent price momentum, rather than applying a tactical asset-allocation approach like the one you’ve outlined here. While I’m strictly a passive investor, this framework for monitoring momentum across asset classes seems quite logical and may have merit for managing sequence risk. Another interesting angle is the Merton share formula for optimal allocation, discussed in The Missing Billionaires by James White and Victor Haghani -a great read if you haven’t seen it.
Yeah, I like Victor’s work. But note that his angle is more on the valuation side, which often points in the opposite direction of Momentum. I would like to see a way to combine the two signals and get the best of Momentum and Valuation. Future research!
Josh beat me to it… Victor’s work is intriguing. An analysis of Merton’s share with regard to SWR would be very interesting
(and probably more fruitful than RP). Please put it on the never ending list of future topics!
I also support posting/publishing momentum signals.
Appreciate your work and analysis. Another great post.
Thanks! It’s on my to-do list!
Another beautiful Article! Congrats Karsten!
Thanks, for the feedback, Karl!
Just wanted to personally thank you too!
I actually asked you about this in another article from earlier this year, as these “managed futures” strategies seemed too good to be true…and it seems they may be a little true. I had a mix of KMLM and DBMF, also added some CAOS recently and DBMF actually performed really well during the minor contraction we just went through, KMLM not so much.
You are one of the very few people I trust to analyze and draw conclusions in a very objective, unconflicted and reliable way. This to me is a strong signal to keep holding maybe 10% of my AA in these, especially as I feel increasingly worried about holding gold or long term bonds as prices of the former is shooting up (gold bubble?), and excessive government indebtedness/ inflation risk (assuming central bank independence is maintained…) endanger the latter going zig next time the market goes zag.
“In crises all correlation go to 1” as they say, but I hope these strategies will do a little bit less so than the aforementioned traditional countercyclical buffers. A lot of it is new and unproven, but I feel like taking a chance there may be wiser than assuming things will stay the same.
Thanks for the kind words!
Yes, DBMF did well this year. Crossing my fingers that it keeps up well going forward.
Nice!! I’ll just wait for the ERN ETF to be launched then. Keep it cheap Karsten, 0.10% tops
If we could collect $1b in assets, that’s all we need! 🙂
It sounds like on the second shout-out for this etf you are starting to believe that you have it in you to rally up $1 bn. Let me know if I can help
Thanks! If I ever set up that ETF, I will let everyone know!
Great article! Please let us know about the momentum signals. Would you consider sharing the excel sheet/code so that we can play around with the strategy?
Here’s the sheet: https://docs.google.com/spreadsheets/d/1AOpx8QhqvnRNbSdEzQ_z6BVD-FXuz8pAC2Zve_a7LQU/edit?usp=sharing
It’s view only. To edit you’d need to make a copy.
Karsten – Thank you for another intriguing and impressive analysis! I considered the TAA strategies over a decade ago, but your approach performs so much better than rotating into cash.
I’m trying to reproduce your results with the sheet you shared, but I’m having some trouble. (My Google Sheets chops are nothing like yours!)
What are the log-returns? That isn’t mentioned in your post. And I can’t find any reference to the N-month return comparisons in the sheet.
I’m also trying to source the returns data independently (SP500TR, 10Y BM and Gold) and I can’t reproduce the SP500TR. My month/month return ratios don’t agree with your SPX-TR ratios, not even for the last few years. They are “close”, but different. You must be doing some non-trivial stitching together of multiple data sources from 1871 to present. Would you mind please explaining?
log returns = log(X(t)/X(t-1) where log is the natural logarithm, i.e., with base e=2.71828…
% returns = X(t)/X(t-1)-1
If your return calculations are different due log versus % returns, then this is the reason.
My total return index values are clearly consistent with the SP500TR. For example:
Yahoo Finance: 9/30/2025: 14826.80, 8/29/2025: 14304.68, Return: 3.6499943%
My data: 81,084,838.71 and 78,229,466.94, respectively. 3.6499952% (levels different due to different starting point scaling)
The difference is immaterial and due to rounding errors
The cross-over momentum rule uses just the average of rolling averages of X(t), never Mont/Month returns
The point-to-point return rule uses the average of rolling log returns, which has the same sign as the X(t)/X(t-N)-1 formula.
Thanks again. As I suspected, my spreadsheet skills were the culprit. SPX-TR and 10Y BM match now!
I’m still having trouble matching the historical data series you are using for Cash and Gold (especially gold), but as you pointed out, the small differences don’t really matter. This is true even when the momentum signals differ slightly. The most significant performance criteria (for me as an early retiree) is the consistently large reduction in Worst DD.
That said, could you share the source of the Cash and Gold SWR asset returns? 🙂
For Cash I’ve tried https://fred.stlouisfed.org/graph/fredgraph.csv?id=TB3MS and https://www.spglobal.com/spdji/en/indices/fixed-income/sp-us-ultra-short-treasury-bill-bond-index/#overview. TB3MS fits better, but neither is great.
For Gold, I’ve tried using GOOGLEFINANCE() using the close prices for GLD and IAU, since gold has no income, but neither fits well.
Cash returns:
https://www.federalreserve.gov/datadownload/
I take the 3M rate H15/H15/RIFSGFSM03_N.M 3-month Treasury bill secondary market rate discount basis
TR(t) = TR(t-1)*(1+Yield(t-1)/12)
Gold: London AM fixing
Hi Milo, did you manage to replicate the results in the end? I’m struggling with my python script and could use some help.
Ben
@Ben – Yes, but my replication was limited to obtaining the same raw data and updating ERN’s spreadsheet. The spreadsheet does all the calculations, including excess over cash.
All I needed was data sources for SPX-TR, 10Y BM, Cash and Gold, for which I am using, respectively:
GOOGLEFINANCE(INDEXSP:SP500TR)
https://www.spglobal.com/spdji/en/indices/fixed-income/sp-us-treasury-bond-current-10-year-index/#overview
https://fred.stlouisfed.org/series/TB3MS
https://www.kitco.com/price/fixes/kitco-fix
For the first three, I was able to download the historical series and did a correlation analysis a few years back, and got to within two or three nines agreement, which seemed sufficient. I couldn’t find the historical series for the Gold London AM Fixing, so I’m taking that on faith.
Another thought provoking article! Would love for you to regularly publish the momentum signals and perhaps a bit more of a primer on how to implement them … the section above explains the theory of it but i’m not sure i fully grasp how to execute the nuts and bolts of it.
OK, thanks! I will see what it takes to regularly publish the momentum numbers.
I’m still quite taken with how delightfully Markovian the entire architecture of this is… Part of me remains entirely convinced that with access to enough data, there’s probably an Economics Nobel somewhere in proving that the market is irrational in ways that are fundamentally Markov chain model-able [yet capricious enough that this isn’t a golden ticket to the advent of the next conceptual hedge fund explosion].
Just having a poke around and manually optimizing static allocations I keep finding marginal improvements (even if they are a bit gold heavy and international light for my innate preferences). The other concept would be trying to identify other potential alpha signals on volatility that could justify parametrically shifting the momentum horizons (vice tail-chasing asset class weighting), since that would let you preserve a long-only approach but potentially mitigate that 1937 example case without creating a new performance low. Just being able to poke around, you could probably crowdsource some of this optimization.
I would definitely be interested in having the tools here be part of a different ‘sandbox’ version of your spreadsheet, because introducing some decision-based SWR and asset allocation responsive modelling to run unattended with some numerical solvers would be tremendously interesting. I think there is some ‘there’ there, without appealing to the financial product charlatanization of the industry.
Thanks for the feedback. Gold has been awesome recently. But it’s not the norm. And it cannot continue forever either. So, finding a consistent way to enter and exit would be crucial before I throw money at an asset class that doesn’t pay earnings and dividends.
Yes, we use your SWR toolbox and would really appreciate you adding the momentum indicators there. Thanks for your many labors, Big ERN!
Same. Great work. Really appreciate it. Would love to see it in the SWR toolbox.
Stay tuned!
Will look into that! Thanks for the feedback!
Karsten, the managed futures ETFs such as DBMF and KMLM shine as a portfolio element, not as standalone 100% allocations. Your ERN momentum strategy has a 0.81 equity correlation, while DBMF and KMLM both have significantly negative correlations. These are clearly entirely different exposures. The portfolio context matters.
Could you add a section where you compare a baseline equity portfolio or 60/40 with portfolios including a 20% allocation to ERN momentum vs. DBMF and KMLM? Only that is the right comparison.
This. Managed futures should only be seen as part of a portfolio. Like other tools, such as the high-beta-crash “first-responder” ETF BTAL, these are deep professional-level picks for a portfolio. Have to understand their function to size them correctly, and align expectations with likely outcomes. 😉
If you look at a chart of 2022 and imagine you retired right before equities tanked, that would have briefly sucked alot. But MF worked well, in a year where almost nothing worked. You can check tickers DBMFSIM or KMLMSIM on testfol dot io to see effects around 2002 (Dotcom bust), 2008 (GFC), or 2016. I hear you Karsten, red tiny sample size alert, sure, sure. So far MFers carried the portfolio forward when markets were tanking. Ludicrous CAGR is not the point of MF, raising Sharpe by lowering variance is.
If you compare an already damn good 40:30:30 SPYSIM IEFSIM GLDSIM portfolio on above site vs. 32:24:24:20 i.e. scaled to 80% and 20% KMLMSIM added, CAGR will be the same, but max drawdown will diminish from 21.5% to below 15%. Nice increase in Sharpe too. And if you check out the underwater graph, how is that not appealing for any retiree?
Now there are MF and then there are other MF. KMLM has a nice and long history (index predates the ETF), but is a bit of a drag in good times. My own picks are DBMF and CTA, because less drag, because very uncorrelated to each other and other portfolio parts, because who runs them (Andrew Beer and Altis Partners) and because of DBMF’s intrinsic diversification through its replication methodology.
HTH
I didn’t see the link to testfol.io for “the above site”, putting it in for completeness.
Yes, thanks, that’s helpful. testfol.io is a great tool.
Good points. My only concern: who could have known in 2019 that DBMF is the great MF fund and the others will be dogs?
With KMLMSIM on testfolio which is available back to 1992 you can simply eyeball the chart to see how it likely behaves. Years of drag in times of equity booms (KMLM holds no equities by design), but great outperformance in major crises. Which is where your portfolio needs a helping hand the most, right.
DBMF (DBMFSIM extended ticker on testfolio) uses a different methodology and normally just drifts up. Like the hedge funds it replicates by looking into the rearview mirror, but importantly without their fees. Also whatever trend or alpha those funds come up with, DBMF will earn it, too. When there is a crisis like in 2022, chances of a desirable chart camel hump are good.
Another hint to select MFs is AUM. Can’t decide what MF to mix in? Sort by AUM on etfdb or wherever. Let the market decide, not the worst choice.
Will that all work good in the future… big question. To reduce impact of single manager risk, some further 1/n split seems prudent. “Diversify your diversifiers” as the FinTwit people say. So in my own FIRE portfolio I mix the rational choice DMBF with the bipolar and autistic kid among managed futures, CTA. Same in the long-short aisle, BTAL with its known mechanics is put alongside ORR, run by a former poker player and bonsai afficionado David Orr. Normal 60/40 investors are advised to keep strong painkillers within reach before looking at latter one’s holdings. 😉
No affiliation to anything mentioned. HTH
Interesting. Visually, the KMLM and DBMF looked very similar during the period when both had return data. But I guess they do things differently.
Good luck with your exotic ETFs. ORR looks really awesome!
Noted. I changed that section. I wanted to write that, but forgot. It’s exactly why I included the correlation and beta stats. DBMF might work as a diversifier for an equity-heavy portfolio. It would have worked better than bonds (due to the drubbing in 2021-2024), but this was clearly a very unique period. During normal bear markets, Treasuries would also have very strong negative correlations with the stock market.
could you comment on TIPS as an asset class to use in your example or as a comparison to how they perform to the other asset classes you described (equities, intermediate bonds, gold commodities, cash)
I don’t have returns going back far enough for TIPS. They only came about in about 1997.
FWIW, TIPS didn’t provide much diversification during the GFC. Treasuries did much better. TIPS also got whacked in 2022-2024 because real rates rallied. You had the inflation protection, sure, but TIPS lost due to the duration effect. Not as bat Treasury bonds, but still pretty bad.
thank you – quite helpful!
Great article. I have more of a set-it-and-forget-it FIRE mentality (glide paths are the most tweaking I would like to do) , so trend following is not for me. I missed your post on ChooseFI about Bengen’s new 5.5% withdrawal rate, so I am glad you linked to it – I was expecting you to post about it on this blog! Very much looking forward to your upcoming risk parity post. Thanks for all your work!
Thanks! Yes, I might comment more on the Bengen book. Haven’t written it up yet. But I should also have a more detailed post here on my own site.
Thanks! Would love to read more about momentum, and get into the weeds on various methods of calculation.
Really looking forward to your upcoming Risk Parity post; I’ve attended a few events where it was championed, but still haven’t found any compelling evidence which suggests it’s beats 75/25 if one accounts for historical biases, which are unlikely to continue or reemerge. I’m just some random dude, so when I raise these concerns I’m dismissed; it would be nice to be able to point to your detailed analysis.
Yeah, the historic biases are the crux. RP worked well during a special time span. But it’s suboptimal most of the time. Stay tuned!
Thank you, Karsten.
I’d love to see the momentum signals updated.
I also appreciate your awareness of the risk of overfitting:
“alternative strategies with numerous bells and whistles are likely to suffer from in-sample bias; that is, someone may go to great lengths to maximize performance, making the backtest appear good, but at the risk of overfitting and potentially creating worse results out of sample”
Looking forward to hearing your thoughts on Risk Parity. I get lots of questions on this from people I’m working with and I expect it will help to have your analysis to add to my own.
Thank you!
Aubrey
Thanks! I will find a way to put the momentum signals together. Stay tuned!
RP is a suboptimal asset allocation strategy: it relies only on the variance matrix, not on expected returns. You should do better with modern portfolio theory. But there are many other reasons, too. Stay tuned for a 5000+ word post on that one, haha!
I might have misunderstood this (not unusual for me 🙂 ), but in order to implement this approach in real life, it would not be enough to update the SWR toolbox spreadsheet with the data, we would need to calculate the current momentum stats each month and then rebalance our portfolio monthly?
If I was following this approach and we entered a significant and long lasting bear market, it would result in me reducing equity allocations each month, reaching potentially a zero equity allocation after 8-10 months?
Once we then entered a bull market again I would miss the start of it, but would then start to rotate back in to equities over the following 8-10 months or so?
I would be selling equities at short term losses but (hopefully) medium to long term gains?
I am in the UK so possibly transaction charges could be worse using this approach depending on the investment platform.
Yes, this involves monthly trading.
Around turning points you might move much quicker than shifting down over 10 months. It’s possible you go from 70% equities to 0% in one ssingle month if the signal is string enough.
Taxes are a major headache. I’d do this only in tax-advantaged accounts!
It would be interesting to see the result from adding a vol target of 20% to the strategy.
20% vol target is huge. Even 100% equities would get you there.
And sometimes, it’s best to just go down to 0% vol with all cash.
But then again, if you short the negative momentum assets you will reach higher vol targets again. But I would likely target ~10% in that case, at least in retirement.
Take a look at Allocatesmartly.com. Excellent site that tracks and backtests 95 Tactical Asset Allocation strategies and also contains SWRs and PWRs for all of these strategies. Loads of great research on this website. AllocateSmartly will provide trading allocations and signals based on the TAA models you chose to employ so all of the math is done for you! I’ve been subscribed for 5 years and was pointed to this website by Todd Tressider at FinancialMentor.com. I’ve been 100% invested in TAA and have been very happy with the results.
Yup, I spoke to Todd about his approach. Thanks for sharing the link here. I recommend it.
Terry, I’m curious what allocations you currently have (Dec 2025) based on your chosen set of AllocateSmartly TAA strategies, and how similar or different it is than what Karsten presented in the above (SWR Part 63) Momentum article (which Karsten has set in the MomentumOutput.csv, as of Sept-Dec 2025, to 70% (primarily US) equities, 20% US bonds, 10% gold)?
I’ve been aware of Todd/FinancialMentor and AllocateSmartly for the past year and receive Todd’s newsletter, but I’ve only found one example of someone who posted their allocations who is part of FinancialMentor/AllocateSmartly and uses at least one of Todd’s strategies (Optimum 3). Their allocations (as of Oct 2025) were 37% in gold, ~27% intl equities (almost all EM/SmallCap), 31% US equities and 5% cash (and no bonds at all). This is quite different than Karsten’s momentum allocation recommendations, so it’s difficult to grasp how to evaluate the relative merits of each approach.
Wonderful post! I am partial to using Hybrid Asset Allocation as the system is designed to be more robust than other trend following setups (fewer parameters). Gary Antonacci has also done extensive research on momentum and has shown that it boosts withdrawal rates. His website is a goldmine of info.
KMLM tracks the Mount Lucas MLM index which started around 1988. Commodities trend following tends to have very large drawdowns as seen in recent performance, and best used as a small component in a portfolio to reduce asset correlations. I found historical data here if you’d like to do extended backtests: https://www.reddit.com/r/LETFs/comments/yuu864/kfa_mlm_index_kmlm_data_going_back_to_1988_charts/
Thanks for the link. I’m curious how that index performed recently. Between 2009 and 2020 the index was flat. Which is a huge opportunity cost vis-a-vis other asset classes. Of course, it rallied again, but if you lost your shift again in late 2022, then this approach is not really that appetizing.
Another excellent post! I wonder what’s the frequency of buying/selling & its impact of taxes vs buy/hold annual balancing. Of course some folks may be able to do part of the allocation in a tax-free or deferred account, but others may not have that option.
The portfolio would have had a turnover of around 280%. So, in an average year, you’d sell 140% and buy 140% of other asset buckets. Quite a lot of trading!
Likewise, I would love to see the momentum signals!
I am curious what the momentum signals indicated this year during the April Tariff tantrum? Did it have you selling for 1-2 months or holding? (Sorry if this is linked somewhere).
Historic back tests: https://docs.google.com/spreadsheets/d/1AOpx8QhqvnRNbSdEzQ_z6BVD-FXuz8pAC2Zve_a7LQU/edit?usp=sharing
(view only, save your own copy)
It did pretty well in April.
Have you ever tried to automate your options trading strategy using Options Alpha? https://app.optionalpha.com/home
I’m considering multiple “options” (pun intended). Among them QuantConnect. Eventually, all trading will need to shift over to trading robots, I’m sure.
I can help you convert the SWR toolbox into a web app that is easier to use…what about that? I bet people would use it A LOT!
https://saferetirementspending.com/
Haha, you beat me to that reply!
Wow where was that before? I dont think people heard about it.
That spreadsheet UI is awful but this webapp is fantastic. Please spread the word of it….
It’s been around for a few months! I added links everywhere I could. 🙂
Someone beat you to that: https://saferetirementspending.com/
Thanks for the informative article! I would like to see the momentum signals in the SWR Tollbox.
Stay tuned!
Thank you for taking on SWR with momentum. It is one I have hoped you would get to one day.
A digression, but since it was mentioned a couple of posts up…if you decide to automate your options strategy, what gets you to whatever technical solution you go with would be some great reading. I think I’m on year 4 of doing it with NQ options and I am indebted to you good sir!
Thanks! I will share that if we ever automate it. Glad the NQ options work for you. I find the SPX ones are already enough. Glad to hear that the vol premium extends to the NQ as well.
Been waiting for this one. I think you should have a conversation with Todd Tresidder about TAA strategies. There are easier ways and more options to implement them, but it does take some work up front.
I also like Darrius Dale’s work on this topic.
The conclusion I’ve come to is I don’t want to leave my portfolio to the whims of the market, nor any one strategy. So I have deployed a couple of Todd’s TAA strategies, I use Darrius’s KISS framework, I still have some index funds I’ve owned for years, and I rotate tbills.
Diversity across strategies helps me stay invested, invest in things I might not, stay dispassionate and systematic, and sleep.
Thanks for all your work too BigERN!
Glad we’re on the same page! Yes, I’ve talked to Todd about this, and the post is partially inspired by that chat when I saw him at the Portland FinCon in September.
1. How would this interact with your discussion of equity glidepaths? These also were designed to mitigate the sequence of return risk, would you combine both strategies somehow or is this a replacement? Maybe it would be worth including one such glidepath in your comparison in this article?
2. In this article you mentioned that your preferred allocation at present is a 75/25 portfolio, but in the glidepath articles you found the best safe withdrawal rates were with portofolios that started with a bunch of bonds but then scaled to 100% equity. So why did you not end up following this?
1: One could (should?) do both. For example, one could set the momentum equity target lower initially, but then walk it up later in retirement.
2: As long as the equity market is still doing well, I’m fine with 75/25. I could always walk the equity share up if we have a deep enough bear market. So, think of that as a fixed AA, but you only trigger the start of the GP once the market is down.
I wrote a paper recently that used KMLM and BTAL alongside equities to make drawdowns shorter and shallower — which should positively impact the SWR: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5584190
Very interesting. Thanks for sharing. This period was special: bonds were terrible diversifiers. But going forward, I’m sure they will do well again. But it’s still useful to have more options rather than less.
I’m suffering from analysis paralysis. Not on the SWR (this series has been great for that) but for asset allocation – which is why this thread is so interesting for me.
Looking at the paper’s OMFL/MOAT/BTAL/KMLM using testfol.io, I only see back to 2017-11-10, which results in a Sharpe ration of 0.70, ie much less than 1.15 for the longer period in the paper. Although it still looks better than 60/40 equity/bonds, not suffering the early 2022 drop (but also not getting the 2025 rise).
I had been looking at a SPY/BTAL/QMNIX mix. At 50/25/25 it performs similar to the paper’s portfolio.
I guess my question is how do you evaluate hedging/diversification strategies? I’m wary of an active approach, but using a static allocation of BTAL/etc, is just delegating the active decisions.
I’m also wary of using something based on options to save me when things go bad. I don’t understand the risks enough. Ern’s strategy doesn’t have this risk.
Results varying based on the time period means I don’t trust my conclusions. Ern’s analysis going back over all the problematic periods is a more convincing analysis
Outstanding write up and thought given to an important strategy! Thanks for sharing the spreadsheet and I’ll cast another vote for sharing the momentum signals.
I’m curious to test different bond maturities. In my own momentum backtests against the S&P 500 I found a split of IEF and TLT smoothed returns in recent decades. I’ve seen people recommend Small Cap, Emerging and even REITs as alternative assets during downturns but they’ve always performed poorly in my tests compared to gold, bonds and T-bills.
I think an ETF is possible. There’s gotta be 9 figures in your comment section alone.
I haven’t looked at the 30y yet. The 10y and 30y are highly correlated, not just the returns but also the momentum on/off signals.
Small-Cap would be a nice addition to a momentum strategy. I would probably design this as momentum in the SMB Fama French factor. And then decide if the equity portion goes to VOO/IVV or some Small-cap ETF instead.
Total agreement that the ETF could work… with the one caveat that since the idea is not inherently crazy, some larger players would be able to quickly build an even lower fee version of it thanks to their economies of scale on this, and offer some additional flavors in the mix as well (ala moving to VOO/IVV/SPY/FXAIX, then leveraging existing Small-cap fund they already have generated) as Karsten alluded to.
I’m sure doing a slightly more fuzzy matrix version of these to account for signal latency and precision could improve that slightly, but the fact that this works and works well at the Google Sheets level speaks volumes.
Yeah, it’s an uphill battle to set up your own ETF!
Hi Karsten!
Thank you very much for your article! It is again amazing work! I have come across a similar strategy in a German startup called “Democratic Alpha” https://www.democratic-alpha.com. Not surprisingly they come to the same conclusion that a 50% addition of momentum funds will yield equity-like returns at lower risk which eventually leads to higher save withdrawal rates. Their approach is however way more expensive as they are using hedge funds to get things done.
They use the SG trend index which is compiled by Societé generale as a proxy for their theoretical deductions. https://wholesale.banking.societegenerale.com/fileadmin/indices_feeds/ti_screen/index.html
Have you ever tried to compare your momentum approach to a benchmark like this?
They have also added a long vola fund (Assenagon Alpha Volatility). What is your view on these funds?
Best regards,
Jan-Oliver
Danke/Thanks!
These momentum signals are too pay much in fees to hedge funds. Of course, the funds have the advantage that they take the emotions out of the equations, i.e., a retail investor might not have the stomach to make large shifts as sometimes required by these signals.
Thanks for the SG link. Obviously much more advanced strategy/index. But not very profitable this year, it seems. But it’s hard to compare the two approaches.
I don’t have an opinion on the long vola fund. Usually, short-vol is the way to, because of the vol premium. But you’d need a good risk model. See my work on options.
Hi Karsten, glad to see another article!
I have to say that I’m really surprised you found appreciably better results with this momentum technique. I suppose I would have expected any more “active” technique like this would be self defeating, with more traders piling on until the benefits vanish. Any thoughts on why this isn’t the case?
Seems very logical to do this only for tax sheltered accounts, though I know some companies have frequent trading restrictions like Vanguard. But since you’re looking at trading no more than once a month (I think?), that wouldn’t be a problem, right?
I was also gonna ask how this technique works with a glidepath, but it looks like someone else already asked about that and I see your answer above.
As always it’s great to read your very technically focused articles (even if the financial language you use is very different from the engineering language I’m more accustomed to, I still really appreciate it and learn a lot).
I have been doing something similar to this for a while and I will say that your best bet is an IRA if you have one. I tried doing something like this in my 401K, but it was a hassle. The Vanguard funds in my 401 only allow a trade every two months, there is no gold, and the only bonds available are much shorter term. I finally decided to simply buy and hold in that account and consider it a hedge when my IRA is going to cash.
Yeah, the trading restrictions can be challendging. Maybe use a self-directed 401k. Or do a rollover to Fidelity.
It would be interesting to explore if a similar strategy could be available with typical 401k index funds. There is usually an S&P500, a bonds one, an a cash option. Perhaps gold could be replaced with an equivalent of VXUS which is typically available? It correlates with S&P500 more than gold does but it’s the only other diversifier that is typically there.
Big Ern did an article back in 2018 that was a more basic version of this. He did 100% stocks to 100% cash or 100% bonds (10 year). You should be able to emulate the stocks->cash as long as long as trading restrictions don’t limit you from trading once a month. Most 401K bond funds are shorter term than 10 years and probably land somewhere between his cash and bonds.
I don’t think VXUS will act anything like gold. You are better off simply leaving gold out of it. I believe the final results only had 3% gold anyways, so probably not a big impact.
To further expand on the VXUS comments, Erm had a recent article “100% stocks for the long run?” where he questioned the value of international as a diversifier. I would also argue that adding a diversifier is more of a buy and hold concept. Moving potentially your entire portfolio into cash/bonds based on a momentum signal is such a massive diversification that will dwarf the impact of a diversifier. What is much more important to diversify is the momentum signal itself (as Ern does with his 12 different combinations), both because it can spread out the moves to cash and it helps to combat overfitting.
Nice explanation! Thanks!
Correct. VXUS is not like gold.
One could model a US vs. non-US momentum signal and then allocate accordingly, though.
In the typical retirement plan, you might simply have to skip the gold option and just work with equities, bonds, and cash.
Thanks, Corwin!
Trading restrictions shouldn’t be an issue because you trade only up to once a month.
One could certainly combine this with a GP, yes. Please see the answer there.
Thanks, Karsten! Very interesting way of monitoring the pulse of the market to time the market. It is also interesting that when people do periodically asset allocation rebalancing to maintain consistent asset allocation such as 70/30, they are moving money out of equities to bonds when market is up, which seems to be the opposite of what the momentum approach would do. I’d like to hear your comments on this. Appreciate!
Yup! Momentum is often the opposite of valuation investing. Doing the value-style will hamper your asset returns during trending markets because you constantly sell the winners and buy the losers. But you must rebalance at least occasionally, especially around the turning points. That’s when valuation shines and momentum sucks.
Love it. Thanks for adding the signals. I use these guys and they send monthly commentary ( if you sign up) on markets.
https://logical-invest.com
How did their momentum strategy work out recently?
Take a Look at the web site. Most subscribers use the top 3 strategies portfolio. If you open the links above, you can view the ytd performance as well as historical of all the strategies.
Thank you for the detailed analysis on this investing approach. I’ll be following the results with interest as the market enters a more volatile “AI” period versus my current 60/40 portfolio allocation.
Thanks. Good luck!
I had been anticipating this post for a long time. Really cool you put the numbers in the spreadsheet. I do have some questions regarding your methodology, especially as compared to your momentum post in 2018:
How much value did you get out of using momentum on gold and bonds? I have never had much luck using moving average on non equities asset classes.
The 100% stock -> bond momentum you ran in 2018 had a little more risk than this (although less than 75/25) but almost 2% better return. Would that have raised the SRW more than the one you used here? It seems like return is a little more important than risk for SRW as the best static portfolio is roughly 75/25 as opposed to the max sharp ratio 40/60.
If starting at 100% equities generates a better SRW, is there a reason to start with 70/20/10 that lowers risk, return and SRW? Is the idea that you are hedging against the effectiveness of momentum?
Well, the bond momentum is essential because if you go out of equities you need to know if you want bonds or cash.
Gold was also quite reliable. So, all four assets contributed to the overall return. Equities were the main engine, but don’t want to miss bonds and gold either.
Not sure what you mean by “If starting at 100% equities generates a better SRW” ? 100% equities will not raise your SWR. You need some bond/cash allocation.
I meant starting at 100% equities instead of 70/20/10 and going to “cash” from there like what you did in the 2018 article. Mostly I am just trying to wrap my head around starting with gold and going to equities. Normally you would think to go the other way. Does gold -> stocks have a better return than stocks -> gold?
I am also curious about the 10 percent in bonds -> cash. That should reduce both the risk and return on the overall portfolio, but I am wondering if that is necessary? Given how much the volatility and drawdowns are being reduced with stocks -> bonds -> cash, I am wondering if there is a need for lowering the volatility even further at the cost of some return?
As for momentum on bonds, I am running momentum on IEF -> SHY (8-10 months) and the return is lower than just sitting in IEF. It was a big help in 2022 but was otherwise a drag on the rest of the sample. But IEF only has a little over 20 years of history and you are using a much bigger sample. Maybe there were a lot more 2022 scenarios prior to IEF existence?
I guess after reading the comments on the 2018 article I was anticipating that you might start with 100% stocks and then use a diversified portfolio for cash including bonds, gold and cash to account for differing economic environments. I don’t have any way to emulate your cascading momentum with my limited tools, so wondering how big a difference it makes.
In my 2018 article I simple simulated equity momentum only with two versions: the unused capital goes to either bonds or cash. In 2018, I never simulated any SWRs.
In this version, I simulate an entire portfolio. It’s not comparable to simple exercised like mine in 2018 or your with IEF vs. SHY.
I sorted the assets from most volatile to least Gold, Stocks, Bonds. I’m more comfortable shifting the gold weight into equity (it’s only 10%) than shifting 70% of equity into gold, especially during volatile regimes. I like to keep a lid on volatility, not just increase returns.
So my understand is you are using a baseline of 75/25 (stocks/bonds) because its the most effective passive method to maximize SWR. The bonds hurt overall return, especially during expansion, but they make up for it by reducing volatility and drawdowns, helping to reduce sequence of returns risk. But the stock percentage needs to be high (significantly higher than optimal sharp ratio) to generate enough return to last the entire retirement horizon.
In the momentum portfolio you have 30% in bonds and gold presumably to play the same role as bonds in the 75/25 portfolio. That will similarly reduce the return during expansions, but its ability to reduce the volatility and drawdown during recessions will be significantly muted. That is because the momentum signal will move the stocks to cash during much of the recession. The bonds/gold will be helpful in the first month or two of the market drawdown before the momentum signal triggers the move out of stocks, but that should be a much smaller impact than when you are 75% stocks throughout the recession.
Because of this, I am guessing the optimal initial stock percentage from a SWR perspective is probably close or equal to 100%. Even if you start at 100% stocks, it will actually only be around 75% long term due the momentum signal with the rest being in some combination of bonds, cash and gold. My though is if you wanted to reduce overall portfolio volatility further, you would probably be better off holding some bonds (and potentially gold) in the static taxable account and get the most bang for your buck in the momentum account.
100% stocks is optimal if stocks are trending up. But occasionally, you want less weight, of course.
Great article! I had read some research papers from SSRN that also showed how momentum can improve SWR. Glad to see it validated independently by you. But one thing, KMLM actually has data going back to 1988 (the MLM index, which is what KMLM tracks). I would love to see you do an analysis with longer data series for trend following managed futures. One thing: KMLM did use 1 yr duration holdings for its cash assets, which did make it have higher tracking error during 2022. KMLM currently matches the MLM index’s short duration cash holding now though. I actually like the higher volatility of KMLM (their volatility target is 15% over longer durations, while DBMF is only 8%). It’s more capital efficient so I won’t deallocate as much from equities.
I could cheek out the MLM data backrest. Though, we should use that with a grain of salt: it’s likely optimized to look best in the backtest.
But good point about the higher risk target. But in their realized risk was closer together: 11.32 in DBMF, 13.52 KMLM.
Great work, Karsten. What’s the easiest way for a civilian to implement it? How about this (in tax-deferred accounts)? When momentum signals bull, sell public debt and buy public equity; and when momentum signals bear, sell public debt and buy public equity?
You should go into the personal finance business (I for one would back you), although that would mean coming out of retirement. 😊
All four asset classes are easy to implement with the ETFs I mentioned. Tax efficiency is the greater challenge: likely best implemented within a tax-advantaged account.
Of course, you could use the momentum signals on the micro level as you described, i.e., determine what you sell. But the major impact comes from shifting large swaths of the portfolio to do the market timing, not from the small changes like you described. 🙂
Super interesting article and very intuitive – reminds me of Markov Models to a large degree, which makes sense. I do have a couple questions, though.
1) Tax wise: Considering someone who might not have a tax-advantaged account (and who could theoretically get hit with a 30-40% tax bill on profits – while not being able to deduct past of future losses as well), would this sort of momentum strategy also make sense?
The 30-40%+ upside in SWR would suggest so, but the benefit could be minimal, correct?
2) Do you have any other way to use these signals to greater effect for these individuals? For example:
a) Using them to help speed up/down a rising equity glidepath, or
b) Looking for large signal fluctuations/shifts that hint at a potential sell-off to rotate out of harms way (i.e. adjust weights way less frequently using some form of threshold)?
What are your thoughts on this? 🙂
Yeah, without a tax-advantaged account, the juice may not be worth the squeeze, due to the tax drag.
One option I envision would be to wrap this into an ETF.
One could use the signals as guidance for GP and/or insight into what assets to liquidate during retirement. But the active weights generated through that would be small, so not much of an impact on your overall returns.
That’s precisely what I was thinking; use it to guide where contributions go on a rising GP. Could be a mitigator for SORR if you’re, say, aggressively shifting from a conservative positions (mostly bonds), into equities.
You’d use this on an active GP to not only guide when, but also into what assets to buy each month/whatever.
Do you think under this particular scenario – and especially considering current valuations – this would make any meaningful SORR mitigation?
Yes, that would be a nice SORR mitigation.
Nice work, but I want to mention KMLM and DBMF has data way before 2019, as the strategy is old, KMLM has data that can be backfilled to 1992, and DBMF backfilled to 2000
I think you can download the data from their website or from testfol.io
Thanks. The problem is that I don’t trust back-filled data from financial institutions because they designed these indexes to make strategy look optimal in history but it’s often overfitted.
Thanks for replying, I’m not so sure about DBMF, but I think for KMLM it’s not overfitting, cuz the index is first invented back in 1988, then some one else decided to make an ETF out of it so there’s not really any overfitting.
I agree the performance of KMLM is horrendous recently, but this is generally true for all trend following ETF, not just KMLM, so it’s probably just alpha decay.
But given that the sharp drop in performance of trend following ETF after 2008, and the revive of trend following in 2022. It seems the culprit of poor performance is more likely to be macro economics, I think AQR also had some discussion on this.
Thanks for the info and historical background!
Agree: the macro environment will determine how well this one is doing. So, I would prefer to use my own momentum signal that worked more reliably through all macro environments.
Great work Karsten, thanks.
You keep talking about SWR, but, are you still recommending 100% stocks in the accumulation phase?
For accumulation, yes. You may shift out a little bit before retirement, depending on risk tolerance.
https://earlyretirementnow.com/2021/03/02/pre-retirement-glidepaths-swr-series-part-43/
I actually use momentum during accumulation, but I do it a little differently. Ern had an article a while ago on how to beat the market. One of the options was to bet on beta and another was to bet against beta. I am toggling between the two with momentum. During a bull market I am betting on beta (growth/tech) and when momentum signals to sell (bear market) I am betting against beta (low volatility/aristocrat). Its a lot easier to stomach during accumulation because the whipsaws are not nearly as painful (you are always in the market). The focus is on return and the volatility is a little higher than the market itself. But as I am nearing retirement I am starting to shift into something similar to the momentum in this article as the focus will be avoiding sequence risk.
Excellent, thought-provoking stuff, as always, Karsten. I always appreciate digging into your SWR and Options posts.
One clarification. In reality, my taxable accounts have legacy holdings in bonds that I don’t want to sell for tax reasons. In your simulations the non-momentum sleeve is 100% equities, but in practice can that sleeve just be whatever I already hold in taxable, as long as the momentum sleeve is properly allocated?
Or does the static sleeve need to be literally 100% equities for the 50/50 strategy to work as intended?
Happy holidays!
I think this would be easy to enter in ERN’s spreadsheet. I might try that myself: since >50% of my portfolio is in taxable accounts, I want to look into something like 50% equities, 10% bonds, 40% momentum.
Yep. For my specific situation, I found that 35% stocks and 65% momentum worked best (significantly better than 100% momentum or 75/25 stocks/bonds which were similar). I could not get any value from bonds, but that is probably because momentum is taking on that role.
If you have a taxable account then you would plug in the stock and bonds percentage of your overall portfolio from that account, and play with the numbers for your tax advantaged account. My guess is if your taxable account is sizable you will be best off doing 100% momentum in your tax advantaged account. I did find that if my taxable account was half or more of my total portfolio, I was better off having some bonds in the taxable account. In fact if my taxable account was 60% of my portfolio your suggestion of 50% stocks, 10% bonds and 40% momentum would be optimal for me.
I would also suggest that if your taxable account is over half of your total portfolio, you might be better off starting your momentum at 90 or 100% stocks instead of 70/20/10. It would be kind of like leveraging your momentum signal to get the most out of it for your limited sized tax advantaged account.
What if you have no tax-advantaged accounts to begin with (some countries have no such thing to begin with)? What did you find was optimal? 75-25 stocks-bonds?! Something else?
Using the spreadsheet, 75/25 stocks/bonds is roughly the best. You can improve on that with a glidepath (Ern has multiple articles on glidepaths), but that gets tricky if you are 100% taxable. You could maybe do a tax efficient glidepath. That would look something like:
During accumulation go 100% stocks.
Roughly 5-10 years before you retire, start putting your savings into bonds instead of stocks. The goal is to get at least 20% of your portfolio in bonds by retirement. This might be difficult to do depending on your savings rate. It also may be unnecessary if market is struggling in the years before retirement.
At retirement withdrawal from stocks (after dividends) if the market is near its peak, otherwise withdrawal from bonds.
Exactly!
Doesn’t have to be. Also the split doesn’t have to be 50/50.
Merry Christmas!
Karsten, I’m trying to replicate your results with my own python script but I’m struggling. I think I don’t calculate properly the “excess return over the cash return series” you talk about.
Would you share the formulas you use?
ExcessReturn = (1+Return)/(1+CashReturn)-1
If you use ExcessReturn = Return – CashReturn you will get different results.
This is a very interesting (and for me, very timely) post.
The back of my mind is saying: isn’t this market timing, and I thought nobody could do that???
I suspect the answer is that this won’t actually beat the market, and that is ok since it is really just trying to be a better tradeoff than the usual 60/40 split.
I like that you already raised the obvious objection. However, your response concerns determining whether we are in an expansion or recession – which isn’t the proposed approach. So doesn’t the objection still apply to using index momentum?
Yes, this is market timing. Yes, this actually beats the static portfolio. Who said that nobody could do that?
Hello Karsten,
What about ILS ETF (Cat bonds) as a diversifier/smoother for SWR ?
Best.
I don’t have enough return history to comment on that ETF. But a bond fund with 1.58% seems mighty unattractive.
Thanks for the work. I understand your results are based on monthly rebalancing. Could you say how frequency would affect the results, semi- monthly, weekly, daily.
Of course, when the momentum signal points to a weight change you must rebalance. But how about there is no change in the recommended weights? Should you rebalance every month?
My experience: Rebalancing slightly less frequently and thus letting momentum run more is helpful. We should thus view monthly rebalancing as a lower bound for your returns. I’ve shown that in Part 39 of my SWR series: https://earlyretirementnow.com/2020/08/05/rebalance-swr-series-part-39/
But don’t expect major alpha from a different rebalancing frequency. These are tiny differences!
Thank you for your time to reply, your first point that rebalancing importance is more a function of the deviation between your currrent Asset Class Weight and the Asset Class Target Weight makes sense. Therefore would setting a rebalance trigger on a % deviation be a more interesting (larger alpha) comparison? Why I ask is when I look at the historical monthly target weights you generated there are some very large jumps, take the 2/28/25 – 6/30/25 trade/tariff bump with Equity Weight falling from 70% to 11.67% back to 70% all within this short period. I can see if I rebalanced only on 2/28 and 6/30 I would have aliased the entire event and the allocation would be a constant 70%. What I lack a feel for if more frequent rebalances adds any false movements (noise), I don’t think so because your methodology of using 8,9,10 month momentum signals establishes the signal smoothing not the calculation frequency, but I’m not positive.
Regarding the second point, I read your Part 39 and it’s conclusion but I don’t grasp that rebalancing to stable target weights (either fixed or glidepath) with a relatively smaller deviation acts the same as rebalancing to much larger fluctuating target weights in this momentum approach.
Again I appreciate your insightful work, before your analysis I was sure my VTI provided meaningful diversification over VOO, you showed not true! You’ve also applied mathamatical rigour to debunk alot of strategies I didn’t believe in like Dividend Only. Now I’m trying to reconcile that selling equities in a market sell-off can significantly increase my CAGR, reduce my risk, tremendously increase my Sharpe ratio, and dramatically improve my worst DD. Definately not what I expected possible!
The large tactical changes are the meat of the TAA strategy. So, you can ignore them but at your own risk. You might miss the escape point in one of the large market crashes.
Stronger models for TAA at Allocatesmartly particularly Financial Mentor’s (and Meb’s). SWR data there too.
Thx for the pointer.
Meb is at Cambria, right? To see if these models are just hindsight magic, I looked at how well actual predictions worked: GMOM and TRTY. Both do much worse than 60/40 when I compare using testfol.io
Are there any examples of momentum funds that have been clearly successful?
If not, why? No one has tried (seems hard to believe), or everyone goes too complicated? Or the fees kill the returns?
I was referring to Meb’s DIY models on Allocatesmartly such as “Meb Faber’s Global Tactical Asset Allocation – Agg. 3”. I generally wouldn’t use Momentum funds. Maybe BLNDX which it 2/3 trend and 1/3 buy and hold.
Yes. I don’t claim that my model is the strongest. But caution with optimizing the backtest too much. It might be overfitted.