Of course, relying on programmatic models is very risky, it’s just a starting point which requires careful adjustments and judgment calls1. What motivated me to dig deeper is the fact that part of the valuation compression seems to be about the narrative, like CEO controversy2, not about fundamentals.

Reverse-DCF of Lululemon: what does the price imply?

I am a big fan of the expectations investing framework: instead of projecting 10 years of cash flows, which is impossible by definition, we are just looking at a 10-year path which justifies the current price. The degree of plausibility of this path tells a lot about whether the company is fairly priced.

Here is one of the scenarios3 which fits the $125 price and 14.15B enterprise value:

You can see the full results and customize the output on the calculator page.

So what do we see here?

To justify the price the company needs to demonstrate a very moderate growth, gradually fading to the terminal 2.5%. The model allows both ROIC and EBIT margin to deteriorate quite significantly: from 28% to 9.5% and from 19% to 11.5%. Moreover, that’s under not-so-friendly other conditions: WACC 9.5% and minimum reinvestment 15% of NOPAT. If we ease WACC to 9% and reinvestment to 10%, the implied year-10 EBIT margin falls to 10%.

That doesn’t look like being priced for perfection. If all of them: low growth, sinking margins and falling ROIC are already baked in the price, there indeed might be an investment opportunity. It’s still too early to say that there is this opportunity. We need to assess the probabilities of different scenarios first.

Base case

Story: the business continues roughly on the current trajectory. Management doesn’t do anything revolutionary, but keeps financial discipline and executes the current strategy: more products in different categories. It means growing revenue but deteriorating margin (a wide selection of products competing with different brands naturally have lower margin than one iconic product). Growth in China and rest of the world continue to offset the decline in North America for several years.

Model assumptions: WACC 9%, terminal margin 14%, 5 years of growth on the last year pace, minimum 12% reinvestment.

Result: $18.09B enterprise value, $160 share price.

Calculator link.

Bull case

Story: the business partially mitigates the brand’s struggle and tariffs effects. Several moderately successful products in the new categories without too much CAPEX for new locations. North American margins are flat and improving in China. In 10 years margin still deteriorates, but only to the point of observed historical minimum.

Model assumptions: WACC 9%, terminal margin 16.4%, several years of high single-digit growth, minimum 10% reinvestment.

Result: $21.47B enterprise value, $191 share price.

Calculator link.

Bear case

Story: the business can’t sustain growth, pricing power declines. Despite high CAPEX spending, growth is only on terminal rate. China growth slows, so the offsetting effect on NA sales mostly vanishes.

Model assumptions: WACC 9.5%, terminal growth only starting now, 15% reinvestment floor, margin falls to 10%.

Result: $12.26B enterprise value, $108 share price.

Calculator link

Probabilities

That’s the hardest and the most interesting part.

First, about the bear case. It implies that several things happen simultaneously:

• New products don’t bring growth.

• China growth slows significantly.

• All mitigation efforts on tariffs fail.

It doesn’t seem to be the case. Worldwide the brand shows growth after relatively weak December and January:

The new products demonstrate growing demand, while the iconic leggings decline:

That might indicate that the current strategy is actually correct: the demand for leggings would decline no matter what the company would do. It’s quite hard to milk one successful product indefinitely. So expanding into different product categories might be actually a solid, while less cool choice.

Anyway, I see no indication that all three things are here. At least growth is happening (though Q1 is likely to be weak). So the probability of the bear case looks relatively low. I assign 20%.

What we saw above is more consistent with the base case. I think 45% probability is appropriate here.

The bull case requires either tariff mitigation and flat North America sales, or stronger growth in China, or successful defense of margins. So there are several catalysts, each can give the market more optimism; they don’t have to happen simultaneously. That’s why I think bull case is more probable than bear. I assign 35% of probability.

So what are we getting after weighing the probabilities and the prices:

The potential upside satisfies my personal margin of safety requirement. So I bought at $125 yesterday (12.05.2026).

Please note that it’s neither a buy recommendation nor a forecast. I am just sharing my own investment process. LULU fits my investment horizon and risk appetite. Quite possible it doesn’t suit yours. Do your own research.

It is not a generic disclaimer. Investing in LULU at this stage means bearing significant risks.

What can go wrong?

• First, a meaningful turnaround is unlikely till the end of 2027. The new CEO starts in September; she simply has no time to make significant changes in 2026. The activist fight will continue to spoil the narrative, there is no clear deadline to end it. The further repricing catalysts in the future, the higher the risk that they will not materialize at all.

• While the bear case scenario is relatively mild (just $108 vs $125) it doesn’t mean at all that price can’t go any lower. $108 is derived from a smooth 10-year path. If margin drops to 15% next year, the market will immediately react with a severe sell-off.

• I might understate the risks of margin decline. If mentioned above drop to 15% margin materializes not as a one-time hit, but a new reality (for example in case of quicker decline of the leggings popularity), then path to 10% margin becomes much shorter and the real bear case would be around $80. I didn’t model it in the DCF because now I don’t see concrete indications for this scenario for now.

• In case of constantly deteriorated narrative (“apparel brand for millennials, which lost its relevance”) the market can just suppress multiples and even if the math is correct, the shareholders will wait ~10 years while FCF gives the appropriate return.

• I already mentioned that Q1 is likely still weak. There is no strong indicator that margins are holding. Judging by Google Trends, February and March were quite soft. If the next quarterly report shows EPS miss and weak margins, the price can easily drop below $100. If it happens, I intend to add to my position, provided the fundamentals still imply today’s long-term picture.

Conclusion

I see LULU as temporarily undervalued because of the combination of:

• Negative publicity and CEO controversy.

• Natural deterioration of the iconic product’s popularity.

• Transition to multi-category strategy which is generally a good decision, but hasn’t yet yielded solid results.

• Tariffs hit.

I don’t see them as company-destroying problems. I believe that lululemon athletica becomes more mature, diversified, less “cool” company, which can still create a lot of value for long-term shareholders. I also like their buyback program4. Not only because it mechanically supports price, but because it forces management to be more disciplined in capital allocation. Another healthy sign of a maturing company.

In short: LULU is not an “obvious buy”. It’s a messy, out of favor stock. That’s why there is a possibility of asymmetric returns: they come with risk.

Subscribe now

This publication is for informational and educational purposes only. It is not investment advice, tax advice, or a recommendation to buy or sell any security. I am not a licensed financial advisor. Investing involves risks, including the possible loss of capital. Always do your own research or consult a professional before making financial decisions.

I took 78 tickers1 and compared:

• Total inflation-adjusted returns, assuming dividends are reinvested and accounting for taxes (20% by default) and transaction costs (0.2% by default).

• Risk metrics: max drawdown, drawdown duration, Sortino ratio, and Calmar ratio2.

• Rolling returns with 252- and 756-trading-day windows to neutralize the importance of the starting point in the time period.

I am focusing on the timeframe 2019-07-01 to 2026-04-013, which leaves me with 32 USD-, 11 EUR-, and 5 GBP-denominated ETFs

You can see and tweak the comprehensive results on this page (warning: a lot of numbers).

Performance of different strategies

• Buy and hold is precarious: timing is critical, safe quiet compounding is questionable. Only 2 of 32 USD tickers have more than 50% of 3-year rolling returns positive after adjusting for inflation. Only five tickers have a best 3-year rolling total return higher than their best 1-year return4.

• Broad cap-weighted indexes performed relatively poorly. Among USD funds, ranked by CAGR, only the 9th spot is claimed by a standard “buy the market” strategy. Moreover, this ETF has negative average 3-year rolling returns.

• Dividend strategies also don’t outperform. Ranked by CAGR, they generally sit in the middle or even in the end of the list. Setting the tax rate to 0 doesn’t help much. On the other hand, if we consider a shorter period (starting 2021-11-30 to cover the last three market regimes, starting with inflation shock), all the dividend ETFs climb the list.

• Strategies that exclude or underweight China worked relatively well. That’s no surprise given the weak performance of China equities from 2008 to 2025.

• Value strategies also demonstrate stronger than average performance. For example, iShares Edge MSCI EM Value Factor (5MVL) clearly dominates the EUR-nominated selection for this period with 8.19% inflation-adjusted CAGR:

  

  If we use a 1-year period to find more EUR tickers with data, two of the top four funds with more than a 30% return are ex-China and two are value-oriented. Among top 10 USD tickers there are several with a value tilt5.

Notable funds

Not a buy recommendation, just the ETF that caught my eye during the research.

Freedom 100 Emerging Markets ETF (FRDM)

This fund has the highest CAGR (9.16%, adjusted for inflation) among the USD tickers; most of the other metrics are also notably better than average. For example, it has the lowest drawdown duration (213 days, while half of the tickers have more than 1000 days).

The ETF has a unique methodology. It uses weighting based on personal and economic freedom metrics via quantitative composite country scores (a combination of 87 personal and economic freedom variables). This approach resulted to underweighting China and India in comparison with the cap-weighted EM indexes. We already know that ex-China ETFs performed better. But is this a structural advantage or just bad timing for Chinese equities?

Let’s run the comparison of these ETFs again on the timeframe when Chinese equities actually performed well. From 1 Jan 2025 to 1 March 2026 Shanghai composite index is up 27%, which is impressive. But over the same period FRDM delivered 96% return, other ex-China funds occupy the second and the third places by CAGR.

So even when China grows quickly, other EM demonstrated even quicker growth.

State Street SPDR S&P Emerging Markets Dividend ETF (EDIV)

• The highest % of positive 3-year rolling return windows (81%, while the second-best Freedom 100 Emerging Markets ETF, has 53%).

• 7.35% CAGR since December 2021 (i.e. during the three most recent market regimes). Only FRDM has higher returns; but EDIV is superior in terms of average and median 3-year rolling returns.

EDIV is based on Emerging Markets Dividend Opportunities Index, so it is not just a high-yield ETF (more info below).

Cambria Emerging Shareholder Yield ETF (EYLD)

Top-3 by CAGR among USD tickers. Focuses on return to shareholders via dividends, buybacks, and debt reduction. This fund also has a value tilt, screening the investable universe by price to cash flow, price to book and enterprise value to EBITDA.

Why dividend strategies perform so differently?

As I mentioned, dividend funds have had mixed results. There are dividend ETFs among both good and bad performers. Let’s dig a bit deeper and find out why.

Take a look at the comparison of the two underlying indexes:

While both of them have “dividend” in the name, their methodologies are different. Select Dividend focuses on the yield itself. Dividend Opportunities applies much more sophisticated screens, for example:

• Dividend Payout Ratio. The ratio of a stock’s Dividend-Per-Share (DPS) to its Earnings-Per-Share (EPS) must be less than 100%.

• Earnings Growth. Stocks must have a positive, three-year earnings growth.

• Momentum. Stocks that rank in the bottom 10% by momentum scores among all constituents are removed.

In the cheap money era this complexity might seem redundant. Select Dividends performed on par or even better than Dividend Opportunities in the first half of the chart and around 2021. But in the recent years Dividend Opportunities delivered much stronger results.

It’s just impossible not to quote Buffet here: “You don’t find out who’s been swimming naked until the tide goes out”, indeed.

Conclusions

Let’s quickly summarize the key findings:

• A long-term uptrend isn’t guarantied; just holding EM in the portfolio without rebalancing might lead to long underperformance.

• Value/quality strategies outperformed broad market.

• High yield might be a sign of disciplined business and a component of a screening for the quality strategy. But high yield alone is often a dividend trap.

Routine reminder: it’s all about past results. The future might be different.

Subscribe now

This publication is for informational and educational purposes only. It is not investment advice, tax advice, or a recommendation to buy or sell any security. I am not a licensed financial advisor. Investing involves risks, including the possible loss of capital. Always do your own research or consult a professional before making financial decisions.

Why MELI's valuation is so tricky

Mercado Libre is a hybrid: ecommerce plus fintech. It is a synergistic ecosystem, which is awesome for business, but a nightmare for analysis.

One of the most interesting and rarely discussed parts of MELI’s financials is float: money that Mercado Libre does not own, but holds temporarily. The two main sources are user balances and the lag between processing a payment and transferring it to a merchant.

This money is an important source of liquidity, helping MELI to finance its aggressive growth. More float means less need to borrow. MELI cannot use these funds as they please, though. Banks are bound by law to keep substantial reserves in safe investments, like government bonds. MELI shows a lot of cash restricted for this purpose:

The growth of the float from less than 9B in 2024 to more than 16.5 in 2025 is very impressive, it’s more than 80% year to year. But the amount of restricted cash is growing much faster: from 2B to 9.9B. Almost a 5x increase1!

Still, the company has around 16.5B - 9.9B = 6.6B excess funds to fuel growth. Growth, in turn, boosts float: the more customers you have, the more deposits and transactions you get. That’s one of the reasons why compressing margins for faster growth makes total sense: MELI can partly offset the acquisition cost via excess low-cost liquidity.

If central banks increase restricted cash requirements, MELI’s ability to utilize float will diminish. This is a big risk for the company and we must bake it into our thesis.

Reverse DCF model assumptions

How should we model float in a reverse DCF? I decided to approximate it as an increase in return on invested capital. Float funds assets that would otherwise require investor capital, so subtracting it from IC reveals the return on what investors actually had to contribute.

We have 11.43B of invested capital and 2.25B of NOPAT2, so our return on invested capital is around 20%. If we treat 6.6B of float as offsetting part of invested capital, ROIC jumps to 2.25 / (11.43 − 6.6) = 47%. Wild. That’s the power of float.

It is an oversimplification, of course. Float is not is permanent and not fully available. But it does the job, i.e. translates float into one of the knobs in the DCF we can use to estimate the impact of regulatory pressure.

I estimate WACC as 12.4%3.

Base case: what expectations are already priced in at $1750

Here is one of the scenarios to reconcile the current enterprise value around 93B and future cash flows:

1. Strong growth for five years, then it fades.

2. Margin gradually expands to 15%.

3. ROIC still benefits from float, but on a smaller scale and the effect fades too.

4. With terminal growth 2.5% and terminal ROIC 14.4%, terminal value dominates.

Note that free cash flow remains fairly weak until the later years.

Bull case

Let’s assume:

• Better growth.

• Float continues to help on the same scale for a couple of years and the effect fades slower.

• Margin also grows to 15%.

• Slightly cheaper capital (12% WACC).

I used 55% ROIC in Year 1 to make FCF close to the adjusted FCF MELI reported for 2025.

In this scenario we are getting 126B enterprise value which means 35% upside and stock price $2356.

Bear case

Assume:

• Capital is more expensive (WACC 14%).

• Regulators will quickly break the float uplift by restricting more cash.

• Slower growth trajectory (the risk is real even if the execution remains excellent: FX reality in LATAM is brutal).

• But margin grows to 15% in this case too.

Here we got 59B enterprise value with corresponding downside 37% and stock price $1106.

Probabilities and buy zone logic

Let’s summarize.

• Base case: MELI continues compounding, growth gradually fades, but in 10 years it will be a huge business with 265B revenue.

• Bull case: MELI keeps its growth on the current level at least for 5 years and simultaneously expands margin; float continues to help for a long time.

• Bear case: still strong execution, but macroeconomic and regulatory headwinds cause much weaker growth and return on capital.

What are the probabilities of these scenarios? As usual, that’s the trickiest part where everyone disagrees.

My best guess:

• 20% bear case. I do not see major company-specific risks right now; the volatile environment is the main danger.

• 45% base case. The market is right with annoying frequency; I usually give 50% to the base case if there are no real holes in its rationale. In the base model described above, the future already looks somewhat harsh, even 2026 is already materially worse than 2025. So 5% goes to the bull case.

• 35% (the rest).

Now we can weight the probabilities and their stock prices using weighted scenario calculator:

So, the weighted average price is just slightly higher than the current price around $1750. Margin of safety is quite thin, so I still pass. My final take:

• 1750 (5% implied upside): interesting business, weak edge.

• 1600 (15%): worth attention.

• 1500 (22%): expected value starts compensating for uncertainty.

• below 1450 (26%): probably compelling if fundamentals are the same.

Note that these numbers are NOT advice, a price target, or a forecast. They are the result of modeling with the assumptions stated above. I made them as part of my own investment process for my own portfolio.

I sincerely invite everyone to poke into the weak spots of the models! I would be happy to get feedback and improve my understanding.

Subscribe now

Disclosure: on the publication date I have no positions in MELI, neither long nor short.

This publication is for informational and educational purposes only. It is not investment advice, tax advice, or a recommendation to buy or sell any security. I am not a licensed financial advisor. Investing involves risks, including the possible loss of capital. Always do your own research or consult a professional before making financial decisions.

1

Mostly due to new regulations that came into force in Brazil:

3

First, we need to get cost of equity = risk-free rate + beta × mature market risk premium + country premium. I use Damodaran’s estimate for Brazil risk premium here.

Cost of equity = 3.97 + 1.48 × 4.23 + 3.24 = 13.47.

Then I use my own calculator and to add the impact of debt. I estimated cost of debt as 5% because recently MELI issued bonds with 4.9% yield and it was heavily oversubscribed.

A strong bull market forces investors to think that dividends are a sign of a company living in the past. Who needs dividends, if hyperscalers can grow 100% per year? (Until they do not).

It would be nice to have an asset which performs well right now and also has a good chance to deliver results amid less optimistic sentiment. There are two reasons why I think EU dividend ETFs could be such an asset.

High-dividend EU ETFs held up pretty well during the last market downturns

In September I tested how different assets behaved when the market sharply declined:

One of the three asset classes that demonstrated both decent stability and returns, were high-dividend EU ETFs. They experienced a drop, but much less than the world market.

High-dividend EU ETFs outperformed broad EU ETFs

Look at the results for the last 5 years2.

• Invesco MSCI Europe UCITS ETF: +76.01%.

• Amundi MSCI Europe High Dividend Factor UCITS ETF EUR (C): +89.43%.

The same is true for the last 3 years (47% vs 52%) and 1 year (16% vs 19%) despite the fact the dividend ETFs have higher management fees.

Yes, past performance does not guarantee future results, everyone has heard that and that is true. On the other hand, I think there is a fundamental reason here.

Europe now struggles to grow on par with the US, there are few high-growth companies which can boost the whole index. But there are well-established EU companies which steadily print cash and distribute it. These companies often operate on the global scale and are quite successful. Their regular dividend payouts are a sign of financial health.

Moreover, MSCI Europe High Dividend is created by filtering broad index by additional criteria, such as last year price performance, return on equity, earnings variability and debt to equity. So the final product has a tilt toward several investing factors, not only to dividend yield.

Note for the US-based readers

Above I compared EU-domiciled products. But the overall implications are the same. ETF is just a proxy for an investment strategy.

Let’s take a quick look at two products:

• Vanguard FTSE Europe ETF (VGK).

• First Trust Stoxx European Select Dividend (FDD).

We can’t compare their price performance head-to-head, because a significant portion of their returns are dividends paid every year, unlike accumulating ETFs I mentioned earlier. I was surprised that the sites from the top of Google do not show total returns, just separate prices histories and dividend yield. So I crafted a simple Python script which gets data from Yahoo Finance and calculates the period performance:

The dividend ETF also wins here, but in 2023-2024 it underperformed, so the picture is nuanced.

The underperformance might be partially explained by total expense ratio difference. FDD total expense ratio fee is high, it’s 0,56%, while VGK is dirt cheap, with TER of 0.09%. FDD’s year drag is almost 50 basis points higher.

Last year the dividend fund outperformed so strongly that it was enough to overcome both TER drag and accumulated underperformance. This impressive recent growth reinforces the thesis that the current regime favors EU dividend ETF.

Conclusions

All of this does not mean you need to buy an EU dividend ETF. I certainly do not recommend doing it without your own research. I personally buy it because I want EU exposure to balance my specific stock picks in US.

I also do not claim that dividend ETF are inherently better for this purpose than broad EU indexes. Fever stocks in the fund — less diversification. Generally higher TER is also a significant disadvantage.

On the other hand, true diversification is the exposure to different risk factors, not just thousands of stocks in your basket.

In short:

Dividend EU ETFs are worth considering on their performance and factor concentration. Not cheered because of the naive bias “I get real money from my investments!”. But also not ignored because of the equally naive bias “dividends are anti-growth”.

This publication is for informational and educational purposes only. It is not investment advice, tax advice, or a recommendation to buy or sell any security. I am not a licensed financial advisor. Investing involves risks, including the possible loss of capital. Always do your own research or consult a professional before making financial decisions.

1

I started to buy in September, so far the asset is growing steadily:

Overall gains are 14% in EUR; I invest in the accumulating variant.

On the Hidden Costs of Passive Investing

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5327757

Conventional wisdom: index funds that have low tracking error (deviation between benchmark and real performance) are doing a good job.

The paper: zero tracking error means the fund left money on the table.

Market indices such as the S&P 500, FTSE 100, and MSCI benchmarks are dynamic: to maintain accurate representation of their target market segments, securities are periodically added and removed. A large literature documents abnormal return patterns around such index reconstitution events.

One of the reasons for such abnormal patterns is that large public indices announce reconstitutions several days prior to the implementation date, while an estimated ∼ 56% of passive ETFs place much of the trading positions at the closing price of the reconstitution day. These passive investors wait until the exact moment that the index change occurs so that they do not incur tracking error with respect to the benchmark. But, the rest of the market can participate in anticipatory trading of the name that has been announced.

For passive managers, delayed execution of reconstitution trades —especially at the closing auction of the reconstitution day—can result in substantial increase of the implementation costs. Executing trades gradually or in advance (even fractionally) may reduce these costs without materially affecting tracking error which is often the reason for the delayed trading.

So: the whole market knows that passive funds will buy a certain asset on a certain date. Traders buy it in advance and then benefit from rising demand, which means passive ETFs buy the asset when it is expensive, lowering their returns.

The paper mentions this as an opportunity for traders, but I do not think it is feasible for a private investor. This opportunity almost certainly is eaten by high frequency traders with sophisticated algorithms and ultrafast connections to the exchanges. The more practical conclusion: we need to keep this mechanism in mind when we compare ETFs and not assume that lower tracking error is automatically better.

To put it more cynically:

Market indexes are not a perfectly optimized investment mechanism. They have weaknesses. Traders exploit them. Fund managers can turn a blind eye to this so they look perfect and safe with zero tracking error. And the guy who pays both of them is a customer who invests in ETFs.

Negative Cross-Impact from Index Investing

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5449979

Conventional wisdom: index funds increase stock correlation because investors buy and sell assets in bundles.

The paper: there are economic mechanisms which lead to negative comovements; the overall effect of indexing on the correlation may be indefinite.

The negative cross-impact is generated by two economic channels. The first channel is the information-driven substitution effect for assets’ private signals. As the single-asset investors are more informed than index investors, they will trade more aggressively when they observe a higher private signal. Given a fixed supply for the asset, this “crowds out” the demand from the index investors. Thus, index investors will submit a lower demand for the index fund, decreasing the prices of other assets. This channel echos the insights from Haddad et al. (2025), which show that investors with better information have higher demand elasticity.

The second channel is the supply-driven complementarity effect for assets’ supplies. In particular, a higher supply shock of an asset will be cleared by higher demands from both single-asset and index investors. Thus, the index investors will increase their demands for all assets via the index fund, elevating the prices of other assets. Both channels generate negative cross-impacts between asset prices.

The text is pretty dense, but the intuition is simple. Let’s imagine a food court with two types of customers: Specialists and Indexers. Specialists can buy burgers and fries separately; Indexers can buy only the combo meal.

Information-driven substitution effect:

Specialists learn that burgers today are super tasty. Specialists buy only burgers. The price of burgers skyrockets. The price of the combo also jumps, so Indexers go eat sushi sets instead of the expensive combo. Nobody buys fries anymore. The burger joint is forced to make a discount on fries just to break even. The result: burger price up, fries price down.

Supply-driven complementarity effect:

The burger joint has a surplus of meat today and announces a discount on burgers. Combo price is also down, so Indexers go buy it, increasing demand for fries which are scarce. The result: burger price down due to the initial discount, fries price up.

In both cases Indexers are amplifying the natural price competition between assets.

The described mechanics are interesting, but the real blind spot I detected after reading this article is:

A higher ex-ante expected return (e.g., from a higher supply level) benefits the index investors. This is in line with the fast growth of index investing in the bull market after the global financial crisis.

The authors use complex math to explain why index investing becomes more efficient and popular during a bull market. Again, the intuition is easy to grasp: in a rising tide you are well-off if you are staying in the average boat; the risk of jumping between boats in search of the best one brings relatively less reward.

So the narrative “index investing is inherently superior, only a Buffett-like genius can beat the average returns in the long run” might be true in current conditions, but if the market becomes much less bullish, the situation can change dramatically.

We cannot yet estimate how significant the change would be (prediction is futile!) but being aware of this scenario helps us prepare.

International Investing: Diversification and Beyond

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5767162

Conventional wisdom: adding international indices to the S&P500 brings diversification.

The paper: by buying international indices you are mostly buying exposure to global factors, indices are highly correlated, so diversification benefit is minimal.

Naive international diversification over our sample period yields Sharpe ratios between 0.48 and 0.56. Our approach to exploiting country-specific risk premia yields Sharpe ratios from 1.44 to 2.32.

Sadly, these impressive Sharpe ratios are attributed to a strategy that requires short positions, which makes it inherently more complicated and costly. But the general idea is certainly applicable to long-only portfolios too.

The real diversification comes from tilting the portfolio towards country-specific factors. That sounds simple and logical — when said aloud.

I personally thought that just buying a combination of index ETFs from different regions offers enough diversification. The paper proves this wrong. On the other hand, constructing a portfolio with the right tilts is a significantly more challenging task than just buying an all-world ETF.

That’s the latest catch from the research nets — let’s see what the next haul brings. Stay tuned!

Related article:

This publication is for informational and educational purposes only. It is not investment advice, tax advice, or a recommendation to buy or sell any security. I am not a licensed financial advisor. Investing involves risks, including the possible loss of capital. Always do your own research or consult a professional before making financial decisions.

We’ve seen plenty of tech giant recoveries in recent years. Maybe META is also going to rebound soon, giving us a tempting buying opportunity?

Let’s find out.

I gathered the necessary financials1 from 10-Q and 10-K filings, used them to calculate basic performance metrics2, and then constructed 3 models of the company’s future performance.

The idea is simple: to justify a certain price, the company must earn a certain amount of cash in the future. More precisely, the present value of the future cash flows must match the current enterprise value. It is the cornerstone of expectations investing.

Base scenario: “what the market is already paying for”

Let’s model a scenario when the present value of the future cash flows translates precisely into the current price of $630. The corresponding enterprise value is $1.58T3

The standard practice is to assign some growth for 5–10 years and then imagine that the company will will steadily but slowly print cash like a perpetuity (think the eternal savings account).

Here is the modeled path for the next 10 years:

We see a typical path of a maturing company: growth and margin gradually decline, returns on investments decline too, but so does the reinvestment amount. As a result, the company enjoys high free cash flow.

As you see, in that case META would generate ~600B of present value in the next 10 years. Around a trillion is assigned to cash flows from the perpetuity (terminal value):

The assumptions I used for the terminal value are pretty favorable: 16% ROIC, 3.5% growth. They might seem modest by the current META standards, but remember, we are talking about the eternal growth, the equivalent of a never-ending savings account.

The market believes that META will be a successful business, able to deliver superior results even after 10 years of decent growth.

Bull case: “everything goes right”

Let’s be generous and assume:

• Much higher growth: 19.3% during the next 5 years.

• Higher terminal margin: 35% instead of 33% in the previous model.

• Reinvestment demand fades quicker.

• Lower capital cost (in the previous model I used 9%, now I am setting 8%).

• Still generous terminal growth 3,5% and terminal ROIC 16%.

• Lower effective tax rate: 14% instead of 15%4.

Here’s our 10-year path:

Free cash flow explodes to 200+ billion (four times year one!).

Enterprise value in this scenario is 2.56T, which means 2.56/1.58 = 1.62 or 62% upside.

The corresponding price is 1.62 × $630 = $1020.

Bear case: “ordinary, still good, but not special”

Now we are modeling a harsher scenario. Nothing particularly nasty, just a few tweaks:

• 9% cost of capital.

• 2.5% terminal growth.

• 15% terminal ROIC.

• 16% effective tax rate.

• 28% target margin.

• The model demands reinvestment to be at least 10% of NOPAT (hard floor).

• Growth fades on the same trajectory as in the base scenario, but to 2.5%, not 3.5%, because it is our new terminal growth rate.

Here is the 10-year path:

Still tons of cash. Still, it is a very good business, still a huge terminal value:

Enterprise value is 1.25T, which means 20.6% downside and translates into a $500 stock price.

Weighting the cases to get the price target

• Current market price implies that META will grow at a double-digit rate for 6 years, gradually maturing. Margins go down from exceptional 40+ to moderate 30+. A long future as a successful, stable money-printing machine after 10 years.

• Bull case implies growth to $700B in revenue and keeping high margins — 35% at the 10th year and steady returns after 10 years at the same juicy (for a mature business) rate. We still need to compress margins somewhat because the guidance indicates higher CAPEX.

• Bear case implies slower growth trajectory, worse margins, and 2.5% terminal rate of return. Still, the company grows to to $500B+ in annual revenue in ten years.

In terms of business operations:

• Bull case: META finds additional revenue streams, like WhatsApp monetization or hardware. It is hard to imagine $700B in revenue just by squeezing ads more efficiently. Efficiency has its limits. Significant expansion of the user base is also problematic. META is already a dominant player. It is hard to climb higher when you are already on the summit.

• Bear case: moderate overall economic headwinds, failing to reliably monetize new directions, expensive CAPEX, still very powerful ad business.

• Base case: as usual, something in between. Meta has some success in new departments, some failures, runs an exceptional ad business, which gradually matures.

How probable is each of this scenarios? It’s the trickiest part where most disagreements live. My assessment is:

• Base case: 60% (market scenario looks boring and moderate, minimal assumptions).

• Bull case: 20% (distinct possibility, but we need a lot of things stacked in our favor)

• Bear case: 20% (the same, a few things must become worse).

I am probably biased against META because haven’t heard strong enough narrative about how and why it would make the bull case a reality. But clearly I could miss something important. So let’s make it 50/30/20, giving the bull case more probability.

Our weighted price will be, then:

0.5×630+0.3×1020+0.2×500 = 721.

That’s 14.5% upside.

Note that this numbers are NOT price target or a forecast. It is the result of modeling with stated above assumptions.

Conclusion

Let’s summarize:

At ~$630 the market is already paying for:

1. High-double-digit growth for a few years.

2. Margins drifting down from 40%+ to low 30s (which is plausible in light of growing capital intensity).

3. A still-good 16% return on capital in perpetuity.

My bull case needs several things to go right (including success of new parts of the business) to get to ~$1,020. My bear case needs harsher economic conditions plus growth to fade a bit faster and margins to land at 28% to get to ~$500.

Probability-weighted, I get ~$720 — about 14–15% above spot. Not exactly the margin of safety I dream of.

That does not mean META cannot be a good investment. It is still possible. But from the models I run, I do not see a mispricing and screaming buy opportunity.

If the price keeps falling while fundamentals stay roughly the same, the three scenarios will start indicating 25–30% upside instead of ~15%. That’s when it turns from “fine company, fully priced” into “buy with a margin of safety.” Still, I prefer my margin of safety to be higher.

That’s my analysis. I made it in my own investment process for my own portfolio from a long-term investing perspective. It is in no way an investment advice; always do your own research.

I sincerely invite everyone to poke into the weak spots of the models! I would be happy to get feedback and improve my understanding.

Disclosure: on the publication date I have no positions in META, nor long, nor short.

This publication is for informational and educational purposes only. It is not investment advice, tax advice, or a recommendation to buy or sell any security. I am not a licensed financial advisor. Investing involves risks, including the possible loss of capital. Always do your own research or consult a professional before making financial decisions.

A spin-off means creating a new company by separating a division from an existing company. Shareholders of the parent company become shareholders of both companies.

There are plenty of reasons why the “child” may be undervalued. For example, big institutional investors may not want to bother with a new, unknown small-cap and sell the stock. A proponent of the approach can name a lot of other reasons — including plainly irrelevant today, like insufficient data coverage. What matters, though, is whether the facts still support the idea now, a quarter of a century after the publication. Let’s take a look.

The portfolio: 16 spin-offs from 2023

• “GEHC”, GE HealthCare

• “STHO”, Star Holdings (iStar legacy)

• “CR”, Crane Company

• “KVUE”, Kenvue (J&J)

• “KNF”, Knife River

• “FTRE”, Fortrea

• “PHIN”, PHINIA

• “BATRA”, Atlanta Braves Holdings (split-off)

• “VSTS”, Vestis

• “VLTO”, Veralto

• “KLG”, WK Kellogg

• “LAC”, Lithium Americas (NA) post-split, NYSE

• “NATL”, NCR Atleos

• “NLOP”, Net Lease Office Properties

• “MURA”, Mural Oncology

• “WS”, Worthington Steel

I took only spin-offs available without friction, no over-the-counter stocks.

Then I used a Python script that calculates the portfolio’s performance, assuming we invest $10,000 in each stock and hold. I tried several options: buying on the first date available and with a 10-, 20-, 30-, or 60-day delay. The delay is an important part: if we believe the mispricing comes from a sell-off by institutions and index funds, we need to give them time to do so before entering.

The result: drastic underperformance compared to the S&P500

Here are the results for our 60-day-delay strategy vs. the S&P 500 (assuming we buy the index on the same date we buy a spin-off):

• Total invested: $160,000.00

• Current value: $195,999.94 (S&P 500: $233,537.43)

• Total return ($): $35,999.94 ($73,537.43)

• Total return (%): 22.50 % (45.96 %)

• IRR (annualized): 10.78 % (20.99 %)

• TWR (cumulative): 32.88 % (63.14 %)

• TWR (annualized): 11.77 % (21.12 %)

Actually, no matter how many days we wait, the picture is the same: roughly similar performance in 2023–2024, then drastic underperformance in 2025.

Examples:

• 0 days delay — total return is $44,264.73

• 10 days — $40,835.32

• 20 days — $39,730.43

• 30 days — $45,216.75

You can find this script here (works in browser) and adjust settings to see the variations of the strategy.

Shall we still monitor spin-offs?

This backtest does not prove that spin-offs cannot be a source of alpha. After all, it is only a one-year cohort. On the other hand, I picked 2023 for a reason. It is recent enough to capture modern trends but old enough to give our portfolio time to show itself (which did not happen). So we have evidence as strong as one case can deliver.

Mechanically buying every spin-off, we are likely to get a market-level performance (adjusted for industry and size).

Sometimes simple strategies work (usually when they are in perfect alignment with the current market regime). I recently wrote about one such case:

As you saw, spin-offs are a different story. That does not mean we should ignore them. In the 2023 selection, we got 3 stocks that delivered a return of more than 100% in two years. Not bad.

Performance Per Ticker (spin-offs 2023)

6.08KB ∙ XLSX file

Download

Download

I think spin-off announcements can serve as a source of investing ideas. Not a strategy for guaranteed outperformance, but a hunting ground where we can find promising stocks for thorough research.

Footnotes for nerds

• When I checked the material before publication, I found that KVUE and BATRA were actually split-offs, not spin-offs. I decided to leave them as they are, this error actually suited the naive strategy: buys are triggered by every press release without deep understanding. Excluding them did not change the overall picture. For 60-day delay the total performance is $37,779.74. Colab script.

• Data source — Yahoo Finance; returns calculated using adjusted close.

• To simplify the process, I calculated performance until 2025-09-25 (the last day KLG was traded), no matter what delay period was chosen.

• Sharpe ratio (annualized, risk free rate 3.00%, 30-day delay): 0.48.

• Transaction costs and taxes were not taken into account. But they would make the picture worse for the spin-off portfolio, not for the S&P 500. 16 separate new stocks vs. one of the cheapest ETF in the world, that’s not even close.

• Strictly speaking, the S&P500 is not a valid benchmark here. To find an academically suitable benchmark, we have to take into account industry tilt, size, and so on. It is very likely that the main reason of this portfolio underperformance is the current regime, unfavorable for small caps and value stocks. But from the practical point if strategy’s alpha is not big enough to beat S&P500, why bother yourself at all?

This publication is for informational and educational purposes only. It is not investment advice, tax advice, or a recommendation to buy or sell any security. I am not a licensed financial advisor. Investing involves risks, including the possible loss of capital. Always do your own research or consult a professional before making financial decisions.

Business websites bombard us with info about new products, competitive dynamics, P/E ratios, analyst forecasts. Twitter adds to the confusion with ardent posts about new high-conviction bets. How to make sense of this noise?

I found the answer in “Expectations Investing: Reading Stock Prices for Better Returns”, written by Michael J. Mauboussin and Alfred Rappaport.

Quick review

• The book provides an elegant and convincing framework for active investing. Even if you eventually decide that expectations investing is not your style, it’s still very much worth reading just for the sake of sharpening overall market understanding.

• The updated edition (2021) is quite contemporary. It describes, among other things, a useful approach to evaluating high-growth, unprofitable companies.

• The language is simple and clear enough; the necessary terminology is explained or understandable from context. However, the authors are academically precise and careful. It is generally a good thing, but they do not share many heuristics that would be helpful for internalizing the material. This is definitely a book that requires an intensive thought process on the reader’s side.

That is why in the rest of the post I will do a quick (and very much oversimplified) summary of the core idea to help you grasp the intuition. I am writing the article I would like to read myself before starting the book.

The core idea: stock prices are based on expected future cash flows

The authors argue that market prices stocks just like any other asset. Prices are always based on the present value of future returns. It is easy to see when you think about bonds: the cash flows are predictable. Working with stocks means much more uncertainty, but the basic principle is the same. The way to deal with uncertainty is to make a rigorous reality check of market expectations baked into price:

Expectations investing is based on two simple ideas: First, you can read stock prices and estimate the expectations that they imply. Second, you will earn superior returns only if you correctly anticipate revisions in those price-implied expectations.

How to estimate expectations that stock prices imply

The core concept here is shareholder value. To get shareholder value, we need to:

1. Get free cash flow (the money actually available to pay out to shareholders in a given year).

2. Calculate corporate value: present value of this future free cash flow. Just like the present value of getting $100 a year later (assuming the risk-free rate is 5%) =100/(1+0.05) = $95.2; the corporate value = free cash flow/rate of return expected by the market.

3. We’ve already done the heavy lifting: to turn corporate value into shareholder value, we simply need to add nonoperating assets and subtract liabilities.

Back to estimating expectations. We need to construct a model: how the company can reach the level of shareholder value per share that equals the current price?

Here is the example from the book:

Looks intimidating, but the logic is simple: we reconcile the company’s performance with the actual stock price. The authors stress that we do not need to forecast the exact future cash flows: we need to focus on drawing a plausible way the company will generate enough value to justify the current price. And then we need to attack this plausible way as a devil’s advocate!

If all the plausible scenarios of reconciliation collapse — we might have spotted a mispricing and an opportunity.

Still too abstract? No problem, here is an example. Please note, the calculations and assumptions are rough and dirty — just an illustration of the logic.

Google’s share price at the end of April 2025: $150 (after another dip). From the February 10-K we can get:

• Net operating profit after taxes: 94B.

• Weighted-average diluted shares: 12.447B.

So the market cap is 12.447B×150 = 1.87T. Let’s draw a plausible way to justify the current price. How do we reach this amount of shareholder value?

Well, one of the ways is to imagine Google as a stable and slow-growing company — a tech equivalent of Procter & Gamble. All we need is 3% growth with a market-average cost of capital is 8%. Assuming that, we can get the present value of the company by discounting the cash flows by 5% (8% - 3%).

Let’s calculate the next year’s cash flow: 94B×1.03 = 96.82B.

And now the discounting itself:

Present value = 96.82B/0.05 = 1.94T.

Wow. 1.94T > 1.87T. The assumed slow growth is more than enough to justify the current price.

Note that we can use different scenarios to explain the pricing. We can build a model with a gradual declining in growth, for example. We can take into account current cash reserves and obligations. What matters is the gap between these models and business reality. In the case of Google, all roads lead to Rome: every scenario implies near-stagnation.

How to anticipate revisions in price-implied expectations (PIE)

Here is the 3-step process the authors recommend for identifying revisions.

Step 1: Estimate High and Low Values for the Sales Trigger and Calculate the Shareholder Values That Result

We begin with the sales trigger because revisions in sales are likely to produce the most significant changes in shareholder value. Starting with sales also allows you to quickly determine whether you should spend time on the other two value triggers, operating costs and investments. This substantially reduces your analytical effort because you focus on what matters.

To estimate the impact of the sales trigger on shareholder value, you first need to estimate a range of sales growth rates, including high and low scenarios. Use the benchmark data (historical performance and PIE) and analytical tools (the expectations infrastructure and competitive strategy analysis) to create these forecasts.

In case of Google sales were indeed the turbo trigger. The price-implied expectations were in a stark contrast with Google’s actual performance in previous years (double-digit growth). The market priced in AI disruption of search and disregarded growing and wide-moat sectors of Google’s business like Cloud and YouTube. On top of that, Google has fast-developing AI services and enough resources to sustain them.

Any forecast based on historical performance shows that shareholder value is significantly higher than the market cap. So if we believe that fears of AI disruption are overestimated, Google at $150 per share is significantly underpriced.

All the company needed to do was to keep the current pace of growth and show that the main revenue stream — paid search clicks — is not disrupted just yet. The market quickly recognized that:

From $150 to $250 in six months. Of course, it sounds simple only in hindsight. But the example shows how it works: revisions in expectations happen when the market sees new evidence about the company’s performance and usually sales are the most convincing factor.

Step 2: Select the Turbo Trigger

You can figure out whether costs or investments qualify as the turbo trigger by determining how far they must vary from their PIE estimates to have a greater impact on shareholder value than does the sales trigger. Assume that you estimate the PIE for a stock that currently trades at $20. The high and low estimates of the sales growth rate, filtered through the expectations infrastructure, yield values of $30 and $10, respectively. Use the expectations infrastructure to calculate the high and low operating profit margins, solely as the result of cost efficiencies, required to create a comparable impact on shareholder value. Look at the result and consider the likelihood that the margin will be that variable.

Sounds a bit complicated, but actually it’s a simple step. We already have forecast prices for positive and negative scenarios. Now we need to just put them into the model, make the candidate trigger unknown (X), keep other values intact, and solve the equation to find which level of X leads to the given price. We need to do it four times: low and high prices for both costs and for investments.

After that, we can see whether such changes are likely (for example, if the positive scenario for or a consumer-staples company requires 40% reducing in costs, it fails the laugh test).

Step 3: Refine the High and Low Estimates of the Turbo Trigger and Calculate the Shareholder Values That Result

You should refine your initial estimates of turbo trigger variability before you decide to buy, sell, or hold. Specifically, drill down one more level to the leading indicators of value. Leading indicators are measurable, current accomplishments that significantly affect the turbo trigger and hence shareholder value. Examples include customer retention rates, time to market for new products, number of on-time new-store openings, quality improvements, and average cycle time from order date to shipping date. Two or three key leading indicators typically account for a substantial percentage of the variability in the turbo trigger.

Just common sense: do not trust the dry model as is. Cross-check its plausibility with additional data directly tied to business performance.

Important side notes

Up above I sketched a skeleton of the process of the expectations investing. Now I will add a couple of quotes that help strengthen the intuition. They show how different some things look through the lens of expectations investing.

Belief: Earnings per Share (EPS) Dictate Value

Reality: Earnings Tell Us Little About Value

(...)

The shortcomings of earnings include the following:

•  Earnings exclude a charge for the cost of capital.

•  Earnings exclude the incremental investments in working capital and fixed capital needed to support a company’s growth.

•  Companies can compute earnings using alternative, equally acceptable accounting methods.

Not all revenue sources are created equal.

Let’s continue poking into Google’s statements to find an example. For Q1 2025 they reported EPS of $2.81 and noted that $0.62 (22%!) came from a gain on equity securities. That’s great, but Google is not a hedge fund. What matters for shareholders is how well Google earns money in its own business and how successfully it innovates to sustain and grow it.

Moreover, the report also states: “$8.0 billion unrealized gain on our non-marketable equity securities related to our investment in a private company”. You see now how messy EPS is? It brings together everything — including paper gains.

We’ve developed a golden rule of share buybacks that you can use as a universal yardstick for evaluating the economic attractiveness of buyback programs: A company should repurchase its shares only when its stock is trading below its expected value and no better investment opportunities are available.

Pretty straightforward. I caught myself thinking “buybacks are good, they usually drive prices up”. But buybacks are not a free lunch. Just like with dividends, spending money here means no investing there.

I can keep quoting — the book is really full of insights. But I think the best you can do at this point is switch to the book itself:

• Buy Expectations Investing on Amazon (affiliate link).

• Extra chapters that were not included in the final version of the book

This publication is for informational and educational purposes only. It is not investment advice, tax advice, or a recommendation to buy or sell any security. I am not a licensed financial advisor. Investing involves risks, including the possible loss of capital. Always do your own research or consult a professional before making financial decisions.

Not all stocks were winners, 9 of them brought losses. I was going to analyze them more thoroughly to detect patterns that led to bad performance.

I started enthusiastically but soon realized that there was no single clear pattern. Each stock is unique, and analyzing each carefully means a huge amount of time and effort. It might be interesting academically, but I don’t want to do a PhD — I want a big, fat, thriving portfolio.

So I decided to test a shortcut: what if we ask AI to identify the best picks? Can it help to weed out losers and improve performance?

Here is my prompt:

Using the official statements only (earnings calls, filings etc.) available on December 1, 2024 (hard rule: no use of documents or statistics that are more recent), make a ranking from best to worst stocks over 1-year horizon. Explain ranking based on data, quotes from official documents and so on.

(+25 tickers copy-pasted from the previous article)

I used ChatGPT-5 (Thinking).

Here is the start of ranking:

ACGL — Arch Capital Group
Thesis: peak-quality insurer with exceptional underwriting & investment tailwinds; still compounding.

PGR — Progressive
Thesis: personal lines pricing power still flowing through earned premiums; combined ratios sub-90s in 2H’24 months.

BBVA
Thesis: record profitability, double-digit NII/fees growth, high-teens ROTE; 3Q print clean.

The results are mixed. On one hand the best-ranked ACGL is actually detrimental, as well as PGR. But! BBVA is actually a top performer with YTD around 90%.

What about TOP-10 of the ranking?

The results are much more consistent. Among the 10 ChatGPT picks, 6 stocks are in the real TOP-10.

Then I ran a backtest: TOP-10 ChatGPT picks vs S&P 500, starting December 2, 2024 and ending October 10, 2025:

The difference is huge!

Top-10 ChatGPT picks:

• Final Value: $135,015.23

• Total Return: +35.02%

• CAGR: 43.23%

• Volatility: 21.47%

• Sharpe Ratio: 1.79

• Max Drawdown: -15.38%

SPY (Benchmark)

• Final Value: $112,212.11

• Total Return: +12.21%

• CAGR: 14.78%

• Volatility: 20.67%

• Sharpe Ratio: 0.77

• Max Drawdown: -18.76%

In my previous backtest the period started on 29.10.2024. Here I wanted to be sure that all 25 companies submitted their Q3 reports before the date in the prompt. To protect experimental integrity I had to start the backtest after the date in the prompt.

What does this mean for a practitioner investor, anyway?

Well, one case proves very little. Honestly, I did the backtest just out of habit, and I take the results with a grain a sack of salt.

There is no way to be sure that the ranking was not biased towards real winners because of GPT’s internal knowledge. Yes, I strictly stated in the prompt that we should rely only on the info available on December 1, 2024, and reasoning seems to comply. But it all can be just a rationalization of sorts.

On the other hand, AI seems to be quite capable of spotting risks, assessing guidance confidence and spotting lacking information. I think it is best to use AI as the second stage of screening: between quick filtering by fundamentals and painstaking manual analysis.

There is a danger that AI with a risk-oriented prompt will be too conservative and make some false negatives (i.e. flagging a potential superstar as high risk). But I can live with that. There are plenty of opportunities to invest and only so much time for research. Weeding out bad deals is more important.

I already started to use a more sophisticated prompt in my research. I will fine-tune it further and describe my experience in future posts. Stay tuned!

This publication is for informational and educational purposes only. It is not investment advice, tax advice, or a recommendation to buy or sell any security. I am not a licensed financial advisor. Investing involves risks, including the possible loss of capital. Always do your own research or consult a professional before making financial decisions.

• Bought a wrong stock.

• Bought the right stock but way too early.

• Sold too early.

• Held too long.

• (You name it.)

Did I lose money? No, I usually gained something, but almost every time I felt I could get from much more. An example: buy at 100, sell at 120, than see how the stock soars to 150 in one month. Agrhh, frustrating.

Part of this feeling is just FOMO, and it comes with the territory. But the main reason was that I just pounced like a hawk on every promising opportunity. I had no thought-out approach. Luckily, after a dozen such cases I finally came up a system to play the long game.

What do I do with stocks that have gone up

It all comes down to the goal of active investing: to gain more money than a passive index fund. The system must allow only the sales that reliably lead to outperformance, it should be hardwired. After I realized that, the rest was simple.

Here is my current process.

1. I notice that one of my stock picks is surging.

2. I calculate how many shares of a world stock market ETF I could have bought back then (on the date when I bought the stock).

3. I calculate how much money I need to buy the same number of world ETF shares now.

4. I calculate how many shares of my active bet I need to sell to get this sum (accounting for taxes and transaction costs) — and how many will remain after selling.

5. If I can retain at least 10% of my initial position, I sell the number of shares calculated in the previous step and immediately use the gains to buy the world ETF.

So, if my stock pick turns profitable, I use this profit as fuel for a time machine: the result is exactly the same as if I had bought the passive index and got some additional stock shares for free.

Of course they are not free. The price I paid is the risk I took. There is always an ugly possibility that the stock you bought goes down, no matter how awesome it looks.

How does it work in practice? Here are my recent trades with Nebius:

On 12 September I bought 100 shares per 75 euro. On 29 September: the price was 95 euro.

Let’s check the Vanguard FTSE All-World UCITS ETF performance for this period. The price went from 138.39 to 139.65. So, 7500 euro would buy me 7500/138.39 = 54.2 shares. I always round up to be on the safer side, so I need to buy 55 shares, which means I need 55*139.65 = 7680.75 euro. Selling 90 shares of Nebius yields 90*95 = 8550. Realized taxable profit was 1800 (8550 - 75*90). Even if transaction costs and taxes eat 500 euro, I can still buy even more ETF shares than planned. Which I did.

Now I can hold my remaining shares of Nebius indefinitely without any worries. I got them “free”, remember? No FOMO when the stock soars. No panic if it sinks. I’ve already caught up with the index, so anything left in the stock is just extra. As long as it’s worth something, I have beaten the passive-only strategy.

Initially I thought about this principle as only of a cheat sheet “what to do when your bet does well”. But actually it forces me to think systematically about all the stages of investing:

• I intend to hold the part of my successful active bets indefinitely → it only makes sense to buy high-quality businesses, able to deliver results in the future, not chasing hype.

• I will sell more than half of the position in the near future → my bet should be big enough so the remaining part can be noticeable in my portfolio → I need to make fewer bets → I go only for the best opportunities I strongly believe in.

• 10% actually is a simplification. I started with this arbitrary threshold, but quickly realized that I need individual targets for each investment.

Nebius trade was an exception: I bought it without proper research, just on the assumption that market evaluates a strong and unexpected signal not momentarily, and the momentum will continue. That’s why I kept the position small and sold at 10% threshold.

Is this the best system that works for everyone? No. It works for me now, but there is no way to be sure that there will always be enough good opportunities for stock picks. Or that selling the majority of the shares and switching to the world index is better than just staying in promising stocks forever.

And most importantly: it is not an investment strategy, it is a self-management strategy. A way to avoid doing dumb things when the market is moody, sleep better, and stay on course.

This publication is for informational and educational purposes only. It is not investment advice, tax advice, or a recommendation to buy or sell any security. I am not a licensed financial advisor. Investing involves risks, including the possible loss of capital. Always do your own research or consult a professional before making financial decisions.

Let’s run an honest backtest and find out.

Test portfolio: 25 stocks with rapid earnings growth, low P/E and PEG

I wanted to test a real-life scenario:

1. A lazy investor visits the Yahoo Finance Undervalued Growth Stocks page.

2. Invests $100,000 in 25 companies shown there, $4,000 for each. No additional research — just blindly buying what the screener shows.

3. No rebalancing; buy and hold.

If this test outperforms the S&P 500, it suggests that the applied filters are sufficient to give valuable insight.

I used Wayback Machine to travel back in time and collect tickers to test:

I hoped web.archive.org had a longer period of history for this page, so I could run multiple tests from different start dates. Sadly, the earliest date is 29 October 2024, so we can run only one near-year-long backtest.

I collected the tickers and ran a Python script that gathers the data about each stock and calculates total performance from 29.10.2024 until 06.10.2025.

I immediately ran into a issue because JWN (Nordstrom) was delisted and became a private company. Shareholders were cashed out.

I adjusted the code to emulate the real-life situation and avoid survivorship bias. The money from the forced sale was evenly distributed among the remaining 24 companies.

And here are the results:

The test portfolio beats the S&P 500 in returns, volatility, Sharpe ratio, and maximum drawdown

Portfolio

• Final value: $123,274.47

• Total return: 23.27%

• CAGR: 25.02%

• Volatility: 19.37%

• Sharpe ratio: 1.26

• Max drawdown: -14.86%

SPY

• Final value: $116,846.00

• Total return: 16.85%

• CAGR: 18.08%

• Volatility: 20.01%

• Sharpe ratio: 0.94

• Max drawdown: -18.76%

The test portfolio is better by just about every standard metric! I certainly didn’t expect this. My intuition was that returns may be on pair or slightly higher, but so would the volatility. I was thinking along the popular narrative: “the market is stagnant, the growth is fueled by AI-megacaps”.

But here we have strong outperformance from quite a mixed basket (among the TOP-5 winners are finance, retail, education, gold mining).

Here are the files with return data for each stock:

Individual Results (with Jwn Redistribution)

6.77KB ∙ XLSX file

Download

Download

Individual Results 24 Only, No Jwn Impact

6.6KB ∙ XLSX file

Download

Download

As you can see, the results are quite promising not only in terms of portfolio performance, but also in the selection of the companies. We got:

• 3 champions (more than 90% return)

• 6 big winners (more than 35%, twice the S&P 500)

• 3 winners (outperformed the S&P 500)

• 4 stocks with positive returns

• 9 stocks with negative results, among them only one big loser (minus 37%).

Not bad for a naive screener!

Limitations

• I did not account for taxes and transaction costs; however, the difference would be insignificant compared to the magnitude of outperformance.

• The relatively short period of time (just over 11 months) means the backtest tells very little about how the portfolio performs across regimes.

• This could easily be luck. One portfolio, one time period isn’t enough to separate skill from random chance. Testing 20+ different starting dates would be needed to validate whether this screener actually has predictive power.

Conclusions

So far, results are anecdotal but have far exceeded my expectations.

Value combined with growth, even if identified by simple metrics, delivered strong results in both absolute and risk-adjusted terms — at least in the last year’s regime. So screeners focusing on earnings growth, PE and PEG is a promising starting point for stock-picking research.

I am going to analyze the stocks with negative returns from this portfolio in one of my next posts. I want to create heuristics that help avoid buying future losers. Stay tuned!

Update — the following post:

This publication is for informational and educational purposes only. It is not investment advice, tax advice, or a recommendation to buy or sell any security. I am not a licensed financial advisor. Investing involves risks, including the possible loss of capital. Always do your own research or consult a professional before making financial decisions.

I decided to really dig into the mechanics behind gold prices. This is the first post about my ongoing research.

I started with a simple matter: seasonal trends. However, in the first 5 minutes of googling, I realized that it’s not that simple. For example, a reputable study says September is the strongest month. Another claims September is the worst. Okay, it seems I need to skip reading and start gathering data.

Gold futures monthly performance (2000–2025)

I downloaded prices for GC=F from Yahoo Finance from 31 Aug 2000 to 30 Sep 2025 and calculated monthly returns.

Controversial September looks just boringly average. January and August seem to be the strongest, and the only dent is in June. Let’s see whether the picture changes if we switch to median returns.

That’s interesting. September went from mildly positive to negative; January, August, and June show the same behavior as with average returns, but more pronounced.

Let’s look at how often each month yields positive returns.

Again, June is the worst; August and January are strong; April and December are on the same level; September is neither here nor there.

Now we’ve got the general picture of how gold behaved over the last 25 years. Time to ask the killer question:

Are these results statistically significant or just noise?

25 observations per month are a modest sample, prone to noise. But taking more years makes the results less relevant: the world is changing quickly, and so is asset seasonality. We need both statistical significance and relevance.

That is why I used the bootstrapping technique:

1. Shuffle all monthly returns randomly among months to create an “alternative history.”

2. Repeat this 50,000 times.

3. For each shuffle, measure the difference between the chosen month and all other months.

4. Count how often the shuffled difference is as large (or larger) than the real one.

That fraction is the p-value — the probability the observed edge is just random noise.

• If p < 5%, the effect is usually considered statistically significant.

• If it’s closer to 20% or higher, it’s almost certainly just noise.

It is quite a standard approach, so there are ready-made Python functions for this. As I expected, most of the results did not survive the test. Here are the two most significant:

• January p-value: 4.7%

• June p-value: 10.4%

In plain words, January’s superiority seems quite real:

June’s weakness is much less certain, but also probable:

Infamous September shows one of the highest p-values — 78.2%, meaning that there is no point in looking at any trends here (at least without additional conditions like regime splits). September sits comfortably in the middle of the distributions, looking like a typical month in the random simulations:

A note for statistics nerds: these are exploratory, unadjusted p-values from a small sample; treat them as suggestive, not definitive.

Sanity check: performance of SPDR Gold Shares ETF (GLD)

To check the results, I decided to repeat the analysis using another gold-related product.

The picture is quite similar: strong January, weak May and June. This increases the credibility of the initial findings.

GLD has a shorter history (Yahoo Finance has data since the end of 2004), so I expected less-pronounced statistical significance. I was half-right:

• June’s result lost even a hint of statistical rigor, with a p-value of 0.26 (a 26% probability that the relative weakness is just bad luck).

• But January reinforced its edge, with a p-value of 0.02 (in only 2% simulations I got the same or better results).

I would be glad if someone verified my results. Feel free to use these files with prices and returns:

Gold Futures Monthly Returns Detailed

21.2KB ∙ XLSX file

Download

Download

Gld Monthly Returns Detailed

15.4KB ∙ XLSX file

Download

Download

Conclusions

• The overall patterns are similar across gold futures and the gold ETF.

• All the months except January and June have results that are too inconsistent; there is no reliable pattern.

• June’s results have some significance. Is it enough to influence investment decisions? Everybody should decide for themselves. For me, it is enough to make a mental note not to buy gold in May and June, but not enough to try shorting gold in these months.

• January tends to outperform other months. That doesn’t mean we can win every time by buying gold in December and selling it in February. Note that 34% of all Januarys have negative results. Still, a bullish bias in January exists. Again, for me personally, it is sufficient reason to try to exploit this tendency.

This publication is for informational and educational purposes only. It is not investment advice, tax advice, or a recommendation to buy or sell any security. I am not a licensed financial advisor. Investing involves risks, including the possible loss of capital. Always do your own research or consult a professional before making financial decisions.

The vacation was fantastic, but attending the clubs boosted my dopamine tenfold. Wow, that is it, thought I. I am really there, I am successful. We are discussing serious things, listening to presentations from startup founders looking for investment. Everything is like what real adults, real rich people do. Wow.

Investing in startups seemed like something you inevitably do after you have some money. Like after kindergarten you go to school, after school you go to college.

A couple of years later, the new toy energy faded and I started to compare the real pros and cons of funding startups with other types of investments.

Risk/reward is not everything

On the surface, participating in a startup round seems quite balanced. Yes, there is a high risk of losing all your money, but in case of success, you get an enormous reward. But risk/reward is not the whole story.

Startup investments are also long-horizon and illiquid: you cannot withdraw your money at will.

Are there high-risk/high-reward investments, but also with liquidity? Of course! Just look at the stock market with margin. Or leveraged index ETFs. Or even a leveraged structured product. I do not recommend them, mind you. These are just examples. My point is that in the open market we have plenty of other opportunities.

Let’s briefly compare pre-IPO startup investments and the stock market with margin:

1. Risk/return: both carry high risk and (potentially) high reward.

2. Time horizon: startups are long-term only — public markets allow any horizon.

3. Liquidity: low or non-existent for startups — high for the stock market.

4. Process: bespoke agreements, legal support required — standardized contracts under state regulation.

If leveraged bets can seem somewhat reasonable by comparison, that alone speaks volumes.

On top of that, investing in startups is also possible via the open market, without serious problems with liquidity! We are talking about more mature startups, and consequently lower returns, to be sure. But the actual difference in annual rate of return might not be so dramatic.

It is possible to hit a superstar company as a VC or angel and get a 20x return -- over 7-10 years. But it is also possible to hit another superstar company on the open market and get a 4x return in one year. Take Nebius as an example: the price jumped from 20 to 80 EUR in 6 months (April -- September 2025).

Can pre-IPO startup investments be expedient? Of course! But only if you actually know what you are doing and aren't just buying a flashy presentation with hyped trends. For example:

• You have deep industry knowledge and your analysis confirms a high probability of success.

• You want access to particular market insights or technology because it is useful for your own company.

Being involved in startups is rather entrepreneurship than investing. If your goal is returns only, you need to consider it a full-time job, not a portfolio diversifier. You really have to be involved in the industry in order to get access to juicy deals, increase your bargaining power, and know the real track record of founders. Why? Because you compete with people who do have all those advantages.

Another bit of wisdom I learned the hard way: if an investment opportunity comes bundled with dopamine, it is the opportunity I should skip.

This publication is for informational and educational purposes only. It is not investment advice, tax advice, or a recommendation to buy or sell any security. I am not a licensed financial advisor. Investing involves risks, including the possible loss of capital. Always do your own research or consult a professional before making financial decisions.

To compensate for the small number of events, I searched only for the assets that demonstrated near-cash stability. So I got an answer to the question: “Which assets have been defensible enough since June 2024?“.

Today I will try to answer another question:

Which assets in the recent past were more resilient than the market and yielded respectable returns over the 1-year period?

Study design & dataset

I used the same dataset. But this time I didn’t just filter the list of ETFs — I ranked them by performance across all four tests (three crashes + one-year return).

Here’s how the ranking works:

1. Crash ratio: for each crash, compare the asset’s return with the benchmark’s return.

2. Hard rule: if there isn’t at least a slight outperformance in all three crashes, the asset is screened out.

3. Score: for the survivors, combine the three crash ratios and the one-year return into a single score.

The logic behind it is super simple:

• Each crash is a stress test — did the asset hold up better than the market?

• The one-year period is the payoff simulation — how much the asset rewards for holding it?

Note that we can create an infinite number of different rankings with a similar methodology. All we need to do is to change the weights of different factors.

The results I discuss in this post are designed to ensure that during each event the asset must perform notably better than the market, just to be in the ranking. But the most weight in this ranking is pulled by 1-year return. All the math logic you can find in this SlideShare presentation. You may also be interested in the Python code; using Google Colab it is easy to change weights and create your own ranking.

Findings

Here is the file with the ranking:

Ranked Scores Per Ticker

123KB ∙ XLSX file

Download

Download

Of course, I am not interested in the ranking per se. It is not a way to find the best ETF and blindly buy it. The purpose of this file is to find strategies that are worth considering, not particular assets.

The fact that the top of the ranking occupies the hyper-risky Leverage Shares -3X Short Coinbase Coin ETP USD alone should remind you that any ranking is only a crude reflection of reality. To make this crude reflection useful we always need to remember what it reflects and what doesn’t.

So the next step is to classify each eligible ETF by underlying asset and strategy. Here’s the classification (done by AI; then I manually did a sanity check and fixed a couple of mistakes):

Etfs From Ranking With Strategies

13.6KB ∙ XLSX file

Download

Download

As you can see, among 214 ETFs and ETPs, there are:

• Precious metals.

• Specific sectors (gold miners, infrastructure, utilities).

• Specific sectors + specific geo (telecommunications in the EU, climate in emerging markets).

• Country ETFs (China).

• Dividend income.

• Low-volatility equities.

• Bonds.

There are also examples of more exotic themes, like sustainable food, but I counted only strategies that occurred at least a few times.

Three observations

1. Equities and commodities dominate the ranking

Earlier I warned about blind use of rankings. On the other hand, we should not just ignore it. We just need to see the bigger picture.

More than half of the list is bond-related products. However, the top 50 is dominated by commodities and equities. It makes sense: bonds are usually more stable, but in a bull cycle equities outperform.

2. China ETFs shine — but after years of underperformance

I was a bit surprised to see 5 China ETFs in the top 15. Out of pure curiosity, I can someday explore why, but from a practical point, I do not think China ETFs can play a role of a moderately defensive asset.

A 3% return over 5 years is definitely not my dream. Not to mention that concentration risk associated with a single country defeats the very idea of a defensive strategy.

3. Equity strategies are significantly overlapping

The ranking includes sector (telecom, utilities), high dividend equity, and low volatility ETFs. A first instinct might be to combine them for diversification. But these different ETFs often hold similar stocks.

Here’s the breakdown of sectors for one of the dividend ETFs:

• Financials 28.20%

• Utilities 16.48%

• Industrials 12.66%

• Energy 11.58%

• Telecommunication 8.52%

And here’s the same for one of the low volatility ETFs:

• Financials 19.96%

• Consumer Staples 19.62%

• Utilities 15.34%

• Industrials 13.12%

• Telecommunication 11.77%

There is a difference, of course. But the similarity is obvious. Moreover, note that utilities and telecom are also here.

Conclusions

I think the best fit for my purpose (some defense, without sacrificing growth entirely) is one of three approaches:

1. precious metals or mines equity (very tempting, but how long the gold rally will continue is questionable);

2. low-volatility equity (actually, these ETFs are specifically designed for my goal);

3. dividend income equity ETFs (most of them distribute, but I go only accumulating; in my account it’s less tax drag, your mileage may vary).

I think I will end up picking 1+2 or 1+3. To choose, I am going to do a bit more research on low volatility vs. dividend income.

This publication is for informational and educational purposes only. It is not investment advice, tax advice, or a recommendation to buy or sell any security. I am not a licensed financial advisor. Investing involves risks, including the possible loss of capital. Always do your own research or consult a professional before making financial decisions.

Today's market is scary. Valuations are climbing to the level of the dot-com crash. Even optimists like Sam Altman admit we’re in an AI bubble.

So just buying more stock ETFs (even with a value tilt) doesn't look like a smart plan for developing my portfolio. I need to consider redirecting new buys toward safer assets. But what are the safer assets?

I decided to take a closer look not at the textbook definitions, but at data relevant here and now.

Study design & dataset

I wanted to find assets that remained stable across the last 3 market downfalls. At least partly, the recent corrections were also induced by fears of overvaluation, so they might reflect how the market would behave in a bigger crash.

I invest in EUR, so I picked Euronext as the biggest EU stock exchange. Vanguard FTSE All-World UCITS ETF (VWCE) serves as a proxy for the world stock market.

Using a Python script with Yahoo Finance data, I searched for the 3 worst 10-day periods of VWCE performance over the last 2 years. Initially, I wanted to go with a 1-year period, but there were only two significant declines over a 10-day window. I didn't want to reduce the window because I wanted to find relatively long corrections, not random bearish jolts.

I ended up with the following periods:

1. 2024-07-23 → 2024-08-06: Weak earnings

2. 2025-02-27 → 2025-03-13: U.S. tariff threats; earnings also weak

3. 2025-03-26 → 2025-04-02: Tariff anticipation and announcement

The overall period turned out to be just a couple of months wider than 1 year.

At least partly those declines were indeed caused by fears of overpricing — that is exactly the reason why today's market makes me nervous.

I parsed historical data about 1725 ETFs and ETPs trading on Euronext Milan to assess how they behaved during those 3 periods.

Then I cleaned the list a bit, deleting tickers without full data (mostly these were recently added ETFs). I ended up with 1603 ETFs and ETPs. By any standard, it is a diverse set of investment strategies.

It became immediately obvious that the set of strategies was diverse indeed. Maybe too much for my taste. Here are the Top 5 winners during the April crash and their strategies:

• WisdomTree S&P 500 Vix Short Term Ftr ETN USD — Leveraged index short

• Leverage Shares -3X Short Coinbase Coin ETP USD — Leveraged single equity short

• WisdomTree Copper 3X Daily Short — Leveraged commodity short

• WisdomTree Brent Crude Oil 3X Daily Shrt EUR ETF — Leveraged commodity short

• WisdomTree Euro Stoxx 50 3X Daily Sh Etfort EUR ETF — Leveraged index short

Their performance during the 10 days was up to 243%. Yeah, I found the assets that are great for precise market timing. Okay, I will save the list for the next life; maybe in my next reincarnation it will work.

If I want to find something more actionable, I clearly need to weed out the leveraged shorts and other super risky strategies. I ran the script again, but this time I also got the 1-year performance.

This time I was not sorting by returns, but filtering by those two criteria:

1. The maximum decrease across all 3 events is 1% (to keep cash-like stability).

2. The minimum 1-year performance is better than 3.25% (current estimation of EU 10-year Eurozone government bond yield — a proxy for risk-free rate).

Here is the file with the dataset and the results of the filtering:

3 Recent Market Declines Dataset

813KB ∙ XLSX file

Download

Download

Findings

There are 41 assets that match both criteria for all three events.

The overwhelming majority are ETFs based on bonds of all sorts: government, corporate, inflation-protected.

It is both normal (bonds are the classic hedge against growth uncertainty) and somewhat surprising. I expected to see at least a few commodity-based products and maybe even low-volatility equity or mixed ETFs. In reality there are only two "survivors" outside the bond universe:

• S3CO Leverage Shares -3X Short Coinbase Coin ETP USD

• XGDE Xtrackers Physical Gold ETC EUR Hdg

S3CO is a specimen of the very high-risk leveraged products I weeded out by comparing 1-year performance with the risk-free rate. Well, it would be a lucky pick back then, but I doubt we can rely on its performance in the future.

XGDE is more interesting. There are plenty of other ETCs on the market, but only this one satisfied both criteria. I think the reason is EUR hedging. For example GOLD (Amundi Physical Gold ETC USD) was down by 1.5% during the first event (2024‑07‑23 — 2024-08-06).

The extra stability didn’t come for free. TER (total expense ratio) is 0.28% for XGDE and only 0.12% for GOLD.

Conclusions:

1. Fixed-income products showed the best steadiness. That doesn't make them a universal hedge, of course. The findings show that they hold during these 3 events, no more, no less. In recent history there have been enough examples of strong correlation between stocks and bonds.

2. Commodities also have defensive potential, but they are much more nuanced. It’s worth noting that among the winners were shorted commodities, such as oil and copper. This means oil and copper prices plunged in unison with stocks.

In the next article I am going to continue the analysis and use the same data, but with a bit more complex metric of performance, not just a strict 1% threshold. Stay tuned!

UPD: Here is the next article I promised:

This publication is for informational and educational purposes only. It is not investment advice, tax advice, or a recommendation to buy or sell any security. I am not a licensed financial advisor. Investing involves risks, including the possible loss of capital. Always do your own research or consult a professional before making financial decisions.

Outperformed by AI: Time to Replace Your Analyst?

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5222427

Method: different LLMs and human analysts wrote stock SWOT analyses, then an expert assessed them.

Results:

1. LLMs often match or exceed human analysts in output.

2. But LLMs sometimes “miss the forest for the trees” on more strategic or qualitative aspects. Human expertise remains essential for nuanced strategic insights.

It’s no surprise that AI can process financial data. But it was refreshing, for a change, to read not a hype piece, but research with a methodology. Of course, it would be much better if we had truly quantitative data—for example, the performance of LLM-generated portfolios over the long term with regime shifts. But at least here, we have a thorough framework for evaluating the SWOTs.

The author also states that "advanced prompts produce significantly better SWOTs than basic prompts". It is interesting to watch how the general attitude towards prompt complexity is changing. In 2022 "prompt engineer" was a sexy profession of the future. A couple of years later AI enthusiasts claimed that LLMs could understand user needs without sophisticated prompts. Now the importance of advanced prompts is apparent again.

The article was written before the release of ChatGPT 5. From personal experience, compared with ChatGPT 4, the new model requires more precise and detailed instructions to force deeper answers.

Small, Value, or Small/Value?

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5199124

Method: compare the performance of two types of portfolios: a) core broad market + 1 satellite fund of small and value stocks; b) core broad market + 2 satellite funds, one small and one value.

Results: the one-satellite strategy has higher return, slightly higher volatility, and somewhat higher risk-adjusted return than all the two-satellite strategies considered.

This fits intuition about factor exposure. By picking funds separately, we capture small-growth and large-value stocks. A cap-weighted core already tilts toward large caps (including large-value), so adding a separate value fund increases overlap.

The authors also noted that the one-satellite strategy would make the portfolio a bit simpler; it would be slightly easier and cheaper to rebalance.

Rethinking the 60/40 Portfolio: Dynamic Hedging with Commodities

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4909458

Method: compare the performance of three portfolios: a) standard 60% stocks and 40% bonds; b) 60% stocks, 30% bonds, and 10% GSCI (a commodities index); c) 60% stocks, with 40% dynamically allocated between bonds and GSCI.

Results: the dynamic allocation showed the best results; classic 60/40 came second; 60/30/10 had the worst returns, volatility, and Sharpe ratio.

The outperformance is modest—around 0.5%. I am not convinced that the strategy as it is would perform well enough in the real world. We need to account for execution costs and taxes.

The most interesting part is the indicator used for the allocation. Let's unpack the allocation rule that gave a pulse to the worst-performing 3-asset portfolio:

When the current stock-bond correlation exceeds the correlation observed over the previous six months, we allocate 2% of the portfolio from bonds to the GSCI index at the end of each month. This process continues until the allocation to the GSCI index reaches a maximum of 20%. Conversely, when the correlation over the preceding six months begins to decrease, we incrementally reallocate the weight back to bonds.

The stock-bond correlation looks like a good candidate for detecting regimes. The authors show that the dynamic portfolio gains value when the correlation spikes:

So, this indicator:

• worked well in this backtest;

• very simple to calculate;

• rooted in metrics directly connected to assets already present in real portfolios.

Using stock-bond correlation leaves room for another indicator without overcomplicating the model. Adding other metrics, for example, for selecting commodities, could give the strategy a more distinct edge. Of course, it’s an untested idea; I might explore it properly in the future.

That’s the latest catch from the research nets—let’s see what the next haul brings. Stay tuned!

This publication is for informational and educational purposes only. It is not investment advice, tax advice, or a recommendation to buy or sell any security. I am not a licensed financial advisor. Investing involves risks, including the possible loss of capital. Always do your own research or consult a professional before making financial decisions.