All There Is to It

Modern Money Mechanics

Release ♓︎

The world is an insane place, but you shouldn’t let it get to you.

There is no salvation in becoming adapted to a world which is crazy.

— Henry Miller, The Colossus of Maroussi (1941)

It’s obvious that the world is changing. The old systems we depended on are deteriorating and will continue to do so.

It’s the end for some things, yes. But for every end, there is a new beginning.

Do you have the courage to keep searching for a better life—not just for yourself but also for those around you? This world doesn’t have to be a zero-sum game.

Is society healthy, that an individual should return to it? Has not society itself helped to make the individual unhealthy? […] Why should the individual adjust himself to an unhealthy society? If he is healthy, he will not be a part of it. Without first questioning the health of society, what is the good of helping misfits to conform to society?

— Jiddu Krishnamurti, Commentaries on Living, Series III (1960)

Before you judge others, consider yourself—your actions, your intentions.

Are you living just for yourself?

Have you offered a helping hand to that friend you see struggling?

The real hopeless victims of mental illness are to be found among those who appear to be most normal. Many of them are normal because they are so well adjusted to our mode of existence, because their human voice has been silenced so early in their lives that they do not even struggle or suffer or develop symptoms as the neurotic does. They are normal not in what may be called the absolute sense of the word; they are normal only in relation to a profoundly abnormal society. Their perfect adjustment to that abnormal society is a measure of their mental sickness. These millions of abnormally normal people, living without fuss in a society to which, if they were fully human beings, they ought not to be adjusted.

— Aldous Huxley, Brave New World Revisited (1958)

The world of what “could be,” “would be,” or “should be” does not exist. There is only the world as it is. Are you aligned with the world as it is?

Consider your mind, body, and soul. Do they need healing?

You’re not alone. There are people who want to help you.

I have a foreboding of an America in my children’s or grandchildren’s time — when the United States is a service and information economy; when nearly all the key manufacturing industries have slipped away to other countries; when awesome technological powers are in the hands of a very few, and no one representing the public interest can even grasp the issues; when the people have lost the ability to set their own agendas or knowledgeably question those in authority; when, clutching our crystals and nervously consulting our horoscopes, our critical faculties in decline, unable to distinguish between what feels good and what’s true, we slide, almost without noticing, back into superstition and darkness.

— Carl Sagan, The Demon-Haunted World: Science as a Candle in the Dark (1995)

Are you living your authentic life?

There is still time.

Anticipation ♒︎

There is a place for you in the future, but it will require sacrifice.

What happens to one of us happens to all of us.

Do you have the courage to cast out the meanness in your heart?

Do you have the courage to connect to the love inside you?

Oh Long Johnson

It has been almost an entire year since I last mentioned Johnson’s SU distribution. This is a real shame because it’s far and away my favorite distribution!

Johnson’s SU is a flexible, unimodal distribution that describes central tendency, variability, skewness, and kurtosis in data. SU stands for “unbounded system,” meaning it is defined over the entire real number line.

If you’re a total nerd like me, read Johnson’s 1949 paper on his systems.

If you are familiar with the normal distribution (“the bell curve”), then you won’t be surprised to know that it only describes central tendency (mean) and variability (standard deviation). Beyond these two parameters, we can also describe skewness, which means the data can favor lower or higher values, making it non-symmetrical. Additionally, kurtosis indicates how prone the data is to extreme values.

The Johnson SU distribution captures all of this variation in parameters, making it a robust yet flexible choice for organic datasets. To truly appreciate what makes Johnson’s SU so powerful, it helps to understand how its flexibility is achieved mathematically. Let’s dive into the nuts and bolts.

Nuts and Bolts

If we take a gander at the cumulative distribution function (CDF) for the Johnson SU, we can begin to understand how this flexibility is possible.

It might look scary at first, but I can break it down for you:

• 𝑥 is the variable we want to transform.

• sinh⁻¹ is the inverse hyperbolic sine (IHS) function.

• Φ (phi) is the CDF of the normal distribution.

• ξ (xi, pronounced “kzai”) shifts 𝑥 in linear space.

• λ (lambda) scales 𝑥 in linear space.

• δ (delta) shifts 𝑥 in IHS space.

• γ (gamma) scales 𝑥 in IHS space.

Each parameter contributes to shaping the distribution, allowing us to model skewness, kurtosis, and other characteristics of the data in ways that a standard normal distribution cannot.

While it’s tempting to assign a one-to-one correspondence between these four parameters (γ, δ, ξ, λ) and the four properties we want to normalize (skewness, kurtosis, centrality, variability), it’s not technically correct. All four parameters interact to control each property.

For our purposes, we can ignore the CDF of the normal distribution—it’s not relevant to our understanding of the transformation. Instead, we can simplify the description of the Johnson SU CDF into two linear transformations. The first happens in linear space, where ξ shifts 𝑥 and λ scales it. Then, we map from linear space into IHS space and transform again. In this second transformation, δ shifts and γ scales.

A critical feature of the Johnson SU is its use of the inverse hyperbolic sine (IHS) function, which enables it to handle data with a wide range of values, including negative and zero values, in a log-like manner.

Let’s take a look at its formula:

If you’ve worked with data from natural systems, you’ll know how important taking the logarithm (log) of your data can be. This is because exponential series are linearized after taking the logarithm. (I talked about this in an old article.)

What’s frustrating is that we can’t use the log function on negative numbers or zero! The IHS function “solves” this problem. By adding 𝑥 to the square root of the square of 𝑥 plus one, we ensure the range is always positive. Taking the log value of this result allows for a “log-like” mapping for negative, zero, and positive numbers.

See for yourself:

In case you’re wondering, IHS, sinh⁻¹, and asinh all refer to the same function, just in different notations. Notice that the IHS has a variable slope—it is steepest near zero and flattens out in either direction. The parameters of the Johnson SU distribution wiggle this curve to find the “perfect spot” from which to normalize a given distribution, much like how a French curve helps draw a wide variety of contours.

Once we understand the mechanics of the Johnson SU transformation, the next challenge is to fit its parameters to real-world data, but how does that happen?

My Computer is Optimistic

I’ve covered mathematical optimization before on the blog (link below).

To refresh ourselves, mathematical optimization attempts to minimize or maximize the output of a given function. I wish I could fully explain why we use certain functions to fit parameters for a given distribution, but that might take more space than we have now.

Put simply: we minimize the sum of the absolute differences between an actual and theoretical quantile function, then further refine the parameters using the Kolmogorov–Smirnov test.

One last previous article to check out would be the first in this series. Refresh yourself on why we’re trying so hard to accurately model returns from the S&P 500 Total Return index (SPXTR).

Without further ado, here are the results:

Without a doubt, Nassim Nicholas Taleb would have something to say about the obvious tail risk in markets that the above chart demonstrates. While the most likely result in a given year is a moderate return (our model has a mode of +16.7%), there is a wide variety of outcomes for market returns.

For instance, there’s an 18% chance that returns will be negative for the year. For years with negative returns, the average is -15.9%. For years with positive returns, the average is +18%. The market likes to move. There’s only a 13.6% chance that the return for the year will be between -5% and +5%.

Grains of Salt

Overall, the Johnson SU provides a nuanced and flexible way to model market returns, capturing their variability and tail risks. However, like any model, it is only as reliable as the data it’s based on.

I have to stress that while I believe the Johnson SU distribution adequately captures the nuance of market return distributions, the data is biased to its domain, which spans from 1989 to today.

In the end, the Johnson SU distribution underscores the importance of flexibility in statistical modeling. As markets evolve and datasets grow more complex, tools like these will remain invaluable for making sense of the unpredictable.

Futures

Dealers are smart, managers are dumb.

If we were to create a formula in TradingView to embody this hypothesis, it might look like the following:

We calculate the square root of the ratio between the product of manager longs and dealer shorts, compared to the product of manager shorts and dealer longs. The result is a ratio that indicates how bullish the “dumb money” is.

For example, this ratio is currently 7.2x, meaning there are 7.2 bullish “dumb” trades for every bearish “smart” trade.

This ratio is as high as it has ever been in its history, dating back to 2006.

With so much dumb money piled long, should we be buying the market here?

Options

The VIX reports the absolute 30-day volatility priced into options, while the SKEW indicates the direction in which these options are being priced. Typically, the VIX is low when the SKEW is high, so we can take their ratio for a combined interaction.

We observe that this ratio is near the top of its range, surpassed only by bullish options positioning at the end of 2017.

If the majority of the options are betting for the market to continue its upwards momentum, should we expect the same?

Credit Spreads

Corporate bond spreads are calculated by taking the difference between yields on corporate bonds and yields on the risk-free alternative (Treasuries).

Fortunately for us, Bank of America already calculates this, so all we need to do is refer to the proper series.

To find a level lower than what we observe today, we would have to go back to March 2005. Needless to say, this is a historically tight spread.

If corporate bonds cannot be expected to pay much over Treasuries currently, should we expect earnings payouts from corporations to be much better?

Labor Market

In addition to the unemployment rate, we also consider the civilian labor force participation rate. By taking the ratio between these two, we get a clearer picture of the labor market as a whole.

From 2019 to about 2023—excluding the COVID period—the labor market has been historically tight, meaning many people are active in the workforce. Such tight labor markets have only occurred in the early 2000s and the late 1960s.

If job losses continue and individuals need to fund their loss of income, can we expect them not to sell their wealth tied up in stocks?

November in Review

Without further ado, let’s see how markets changed in November.

The most notable positive movements were seen in the Canadian, Australian, and United States stock markets. For the U.S. markets, much of the bullish sentiment is attributed to “red sweep” of the election results favoring Donald Trump, who is perceived as a pro-business candidate.

On the other hand, the most significant negative movements occurred in the VIX, Hang Seng index, and gold. Volatility decreased following the election, while gold remained subdued due to reduced safe-haven demand.

This suggests evidence of a re-ignition in equity markets and a reduced likelihood of a severe recession. However, we know that such optimism rarely persists for long.

I remain skeptical, particularly about the current data on market positioning. It is important to remember that markets are historically expensive. Furthermore, the unresolved issues with corporate debt are likely to reach a critical point in 2025–2026.

Rich and Boring

Whenever you lie to yourself, it’s like you’re pushing your soul away.
— ASAP Ferg

Get Cracking

We’ll keep it brief today.

The monetary base (M0) is the total amount of money in the economy that is directly created and controlled by the central bank. It consists of two components: bank reserves and currency in circulation.

While the decline of physical coins and bills has been evident, you might be surprised to learn that there is still $2.35 trillion floating around in the economy! This series also serves as an interesting basis for normalizing the growth of the stock market.

Below is the S&P 500, adjusted by a scaling factor to approximate the size of the equity market, and expressed as a ratio to currency in circulation.

This is one of the most consistent ranges I have observed when normalizing the SPX. Deeply depressed market periods, such as those in 1982 and 2009, appear under a ratio of 6.75x, while extremely overvalued periods, like those in 1965, 2000, and 2024, occur above a ratio of 18.5x.

It’s probably nothing. 😉

Value Trap

When I was a child, I was fond of turning over rocks and playing with the bugs that crawled out. Now that I’m grown, I do the same thing with stocks, uncovering nasty-looking companies to see if they’re worth a closer look.

Then again, sometimes all you find are maggots.

Early is Wrong

Being right too early is indistinguishable from being wrong. If I know something’s going to happen in the market, but I’m five years too early, that’s the same thing as being wrong.

— Christopher Ailman

This idea hit home for me early in my investing journey, long before I started this blog. Back then, I was eager to dive into opportunities that seemed obvious—sometimes too obvious, and I learned the hard way.

One such moment came with my buying Netflix (NFLX) stock on April 20, 2022.

The day before the 20th, NFLX was tanking in after-hours trading, and I got excited. That night, I placed my limit orders, planning to “ladder-down” my long position by adding more shares as prices fell. When I woke up shortly after market open on the 20th, I was surprised to find that all my orders had filled—and I was now in the red, with an average share price of $240.

I had caught the falling knife.

My position stayed underwater for so long that, by October 2022, when it finally turned positive, I sold for a +14% gain. At the time, it felt like the right move. I was being risk-averse, and the price kept returning to the level where I’d sold.

Fast forward to today: NFLX is trading near $824, which would have been a +243% gain on the position. What caused me to trade this position so poorly? I entered too early, and the anxiety of unrealized losses clouded the more rational, patient parts of my mind. I forgot there was still a bullish case for holding the position.

I am learning to be more patient—not just to wait for a bottom to settle before entering, but also to let a position breathe and grow.

Wrong is Wrong

I wrote an article on this blog about Spirit Airlines (SAVE) on May 22 of this year. The stock was trading at $3.88 a share then. Today, it’s at $1.08, which would have been a -72% loss if you’d gone long based on my recommendation.

In my folly, I argued that if SAVE’s price-to-tangible-book (P/TB) ratio returned to 1.0, where it “should be” compared to the 0.42 it was trading at, a +117% gain was possible. At the time, this seemed like a reasonable assumption. To put it mildly, I’ve since learned a thing or two about balance sheet analysis.

Initially, I had some hope for the recommendation. I was feeling smart, and that alone should have made me suspicious. Just two trading days after I wrote the article, the stock jumped as much as +30% from my recommendation at $3.88. But this seeming victory was short-lived, and by late July, it was clear that the price was losing support near that level.

If there’s any silver lining, I did highlight the risks of such a drastic drawdown. Unfortunately, those warnings were overshadowed by my overly bullish sentiment.

Learning Lessons

How has my analysis changed since May? Let me show you.

This is what I call my “value dashboard,” an eclectic mix of technical and fundamental indicators. It includes linear regression, cumulative averages, drawdown retracements, turnover, income-statement ratios, and balance-sheet ratios. It might look busy, but these indicators give me a holistic view of a stock’s prospects.

So why has SAVE been so toxic? How did it lose another -72% from a drawdown that was already -96% deep? What did I overlook back in May?

To start, consider the debt-to-market capitalization (D/P) ratio. This ratio makes it easy to say, “For every dollar you invest in SAVE, how much debt are you effectively buying?” In SAVE’s case, the answer is $62.61. That’s right—for every dollar you invest in SAVE, it’s like you’re buying $62.61 of their debt. I have yet to find another ratio in the market as extreme as this.

SAVE’s debt load is alarmingly high—so high, in fact, that I should have flagged it much earlier. Unfortunately, I missed it, and for that, I owe my readers an apology.

Still, how can their P/TB ratio be so low, at a mere 0.15? While this might suggest an undervalued opportunity, the discount factor gives a clearer picture of their financial health. Let’s calculate it, as we did last week with Chinese equities.

Given SAVE’s enterprise value and the value of its tangible assets, its discount factor is about 7%. That means there’s only a 7% margin of error between the book value of their assets and their real liquidation value. It’s a tiny cushion.

Compare that to SAVE’s discount factor in mid-2020, which was closer to 28%, and you can see that the opportunity to go long was far more attractive then than it is now.

I made a mistake—a big one—but I’m trying to learn from it. All I can do now is more thoroughly vet the stocks I showcase on this blog.

I’ll continue learning, improving, and sharing those lessons here.

If you can’t respect that, your whole perspective is wack
Maybe you’ll love me when I fade to black

— Jay-Z, “December 4th”

Zhaocaijinbao

Riches and treasures come into the house.

Liquidation Value

Imagine if you had to sell all of your belongings in a hurry—say, within a week. To offload everything quickly, you’d likely need to sell most items at a steep discount to attract buyers. For items that don’t attract buyers, you’d simply have to discard them.

Under financial stress, a company might face a similar situation. The book value of its assets becomes less relevant, and those valuations must be discounted. However, the company still owes money in the form of liabilities.

All told, we can calculate liquidation value by taking total assets less intangible assets, applying a discount factor, and then subtracting liabilities from the discounted tangible assets. While this is not the complete formula (different asset types require different discount rates), it’s good enough for our purposes.

We can also flip the problem on its head and instead find the discount factor that would allow the discounted tangible assets to cover both liabilities and market capitalization. The larger the discount factor, the higher the margin of safety!

Net Current Asset Value

Another conservative valuation of a company’s balance sheet is net current asset value (NCAV). Simply put, NCAV is calculated by taking current assets and subtracting liabilities. Current assets are the most liquid ones, like cash, short-term investments, inventory, and receivables. By covering liabilities with only the most liquid assets, we get a clear picture of how the company would perform in a liquidation event.

I like to use ratio analysis on NCAV and liquidation value. All we have to do is divide the market cap by these measures, and if the ratio is under one, we’re looking good.

Dogpile

Enough math—let’s look at some companies!

Looking at the table above, we can see that most of these companies have experienced large drawdowns from their respective all-time high (ATH) prices. These drawdowns are mostly around 90% below ATH, with the exception of CANG, which has recovered from its 88% drawdown at the cycle low. While large drawdowns scare most investors away, we’re interested in finding these diamonds in the rough.

Turning to a couple of ratios from these companies’ income statements, we can see that all these companies are profitable, and most have substantial revenue bases. The sales multiples are all near or below one, with the exception of CANG again. A P/S of 4.9 might be common in American markets, but I would be cautious of the high net margins required to turn this revenue into income.

Moving on to earnings multiples, they are all quite reasonable. Every one of them is below 12, a rare valuation in American markets. Some companies, like XYF, have incredibly low P/E ratios. To understand how low a P/E of 1.76 is, we might calculate the earnings yield by taking its inverse, or 85%. It’s unlikely this income is sustainable, but it essentially means that net income could cover 85% of the market capitalization in a single year.

We’re not done yet—we haven’t even looked at the balance sheet, which is the focus of the value metrics defined above! Not only do most of these companies have a tangible book value large enough to cover twice the market capitalization, but the NCAV for the companies we have data on can also cover the market capitalization.

Drilling down into the magnitude of the margin of safety these book values offer, we can look at the discount factor. These range from 37% to 53%, meaning that these companies could discount their tangible assets by those degrees and still cover both liabilities and market capitalization.

Risks

Briefly, let’s touch on some risks inherent in these investments.

If we were Chinese nationals, there wouldn’t be much risk of losing our positions to politics, but as an American myself, the risk is higher that foreign investors could be left holding the bag. It’s unlikely but not impossible.

Furthermore, we have to trust that these companies are reporting accurate numbers in their financial reports. Read through their websites, listen to earnings calls, and analyze financial reports for any discrepancies to discern whether these companies are the “real deal.” A company like Xunlei is established, culturally relevant, and likely trustworthy, but we must do our due diligence for each of these companies.

China itself has experienced a profound deflationary recession that is scaring away a lot of foreign investment, but if we’re brave enough to take a contrarian stance in the face of these risks, the rewards could be significant.

Common “Wisdom”

To estimate how well our stock market investments perform, we often look to history as a guide. For most, simply knowing the historical average annual return is sufficient.

Using monthly close data from the S&P 500 Total Return index (SPXTR or SP500TR), we can calculate this mean return, but we must proceed carefully. A total return index is preferable to a simple price index because it includes dividends.

First, we’ll calculate the relative change between any given close value and the close 12 months prior. While many people are familiar with percentage change, this measure of relative change can bias our mean toward positive returns, so we should use log change instead. After doing this, we find an average return close to 11%.

To visualize performance over time, let’s look at a chart of annual ratios between the current close and the lagged close 12 months prior.

Considerations

Ratios provide a useful between interpretability and accuracy. The chart above is scaled logarithmically, making positive and negative returns symmetrical and mathematically appropriate if we calculate a mean. It’s also easy to read. To convert a ratio to a percentage change, simply subtract one and multiply by one hundred. For example, 0.6 becomes -40%, and 1.5 becomes +50%.

Our annual mean return appears as a blue line in the chart above, at a ratio of 1.11. The red line at 1 represents no annual change; values below 1 indicate years of loss and values above 1 indicate years of gain.

Notice the wide range of possible returns. In a given year, we might have turned a dollar into a dollar fifty, or just as easily, into 60 cents! Expecting $1.11 every year would be foolish.

We should also consider the years included in our annual return data. The chart above shows data from February 1989 to November 2024. While this may seem adequate, it leaves out over 100 years of additional stock market activity (Charles Dow began recording his index in 1885).

Refresher

In future articles, we’ll revisit this topic in more detail, but for now, here’s a summary of what we’ve learned about calculating total return:

• Use a total market index that includes dividends.

• Avoid calculating averages with percentage change.

• Instead, calculate averages with log change.

• An average alone does not capture the complexity of a distribution.

• Be cautious with data limited to a few market cycles.

Setbacks

Once again, Yahoo Finance has thrown a wrench into my data-gathering process. Now, not even the workaround I devised last month is working.

I won’t lie, this is getting frustrating.

There’s one last-ditch process that I can think of that would allow me to maintain the automated market reviews. If that fails, I may have to pivot to something else.

October in Review

With much exertion, I present October’s digest.

The definitive upward movements this past month were seen in the dollar, 10-year yields, gold, and the VIX. The definitive downward movements were seen in equity markets, notably in India and France. China’s SSE Composite index had a larger magnitude move down, but the move was not as clean as India’s decline.

With the recent gyrations of the U.S. dollar, the 10-year Treasury Note yield, gold, and the VIX, volatility is being introduced into the foundations upon which equity markets depend. There’s a lot of uncertainty in the markets, even if it doesn’t translate directly into the performance of the S&P 500, for instance.

You might find it interesting that 10-year yields are hovering around 4.3% now, and with money market funds paying approximately 4.75%, the spread between the two is narrow. If you invest in fixed income, consider opting for longer-term yields here.

Pause

I try not to indulge myself too much in this blog, but today is different.

If I’m being honest, it’s been over a month since I felt excited about finding something to write about each Sunday. I’ve lost the plot.

You, my dear reader, might have noticed.

Readership is dropping off a cliff, as if you can smell my uncertainty. I don’t blame you, the articles have been boring. If not exactly boring, then certainly lacking a genuine attempt to connect with you.

My apologies.

Loss

Though I did not know him well, I was included in an elderly man’s process of medically assisted dying.

He died last Tuesday.

I was able to see him the week before, the day before, but not on the day of his passing. Nor did I see his body, draped in the American flag as he was a veteran, because I did not want to.

He was terminally ill and he had his fill of life, so he wanted to go on his own terms.

I respect him for that, but it didn’t make seeing the grief of his wife, his daughter, and his granddaughter any easier. I feel humbled to have been part of such a profound and deeply personal event, even as a stranger.

His death affected me more than I expected.

Ennui

And when it gets too full, and I can’t close the lid
I spaz on my family and my closest friends

— Big K.R.I.T., “The Vent”

There’s an emptiness inside of me.

I’ve tried to feed it a diet of quick pleasures; it swallows them up. Even after one of my best ideas recently, the excitement fades. So, what can be done for this hole to feed?

Money doesn’t buy you most everything.

Hint

I promised myself I would write this blog regularly for at least a year.

After that, I will write when I have something to say.

Turnover

Simply put, turnover is the ratio between a given trading day’s volume and the total shares outstanding for a stock. This ratio can be smoothed with a moving average to make trends more apparent and easier to analyze.

When a stock has low turnover, it means that few people are trading the issue at that time. Conversely, high turnover indicates increased activity surrounding the stock. We can think of “quiet” and “loud” as good analogies for low and high turnover, respectively.

How does this help us trade? This part is a little more complex. High turnover, when a stock has been beaten down, is typically a sign of capitulation, where weak hands sell to stronger hands at a discount. Low turnover is more complicated; it can be a healthy sign if the price consolidates and then resumes an upward trend. However, it can also indicate weakness if the price is in its distribution phase, where strong hands sell to weaker hands at a premium.

Quiet as a Mouse

Using TradingView, we can request data on the daily volumes and total shares outstanding (TSO) for the ten largest public U.S. companies. We will sum each variable across all companies and then find the ratio between the summed volume and the summed TSO to establish a proxy for “market turnover.” While ten stocks are a small sample, they are the largest companies in a highly concentrated market, so the result should be fairly accurate.

Let’s plot what we found, highlighting certain levels of turnover on the S&P 500 cash index (SPX) alongside it.

Using the exponential moving average (EMA) of turnover, smoothed over 10 periods, we see that the current value is 0.61%. This means that, on average, 0.61% of a stock’s total shares are traded as volume in a day. The only other periods with such low turnover occurred at the end of 2021, specifically in August and October.

While turnover is rarely this muted, we can also highlight quiet periods below the 0.87% level, such as December 2017 and December 2019. It’s interesting that these periods are all near the end of the year, and for 2017, 2019, and 2021, they are spaced two years apart.

Notice that after all these “quiet times,” turnover eventually picks back up, sometimes dramatically, and that typically the price of the SPX suffers.

In contrast, consider that the 10-period EMA of turnover during March 2020 was 2.42%, or roughly four times higher than it is now! When the market goes south, it becomes active.

Keep this in mind.

Tool Around

This past week, I was delighted to stumble upon an interactive tool hosted on the CME Group’s website. It allows users to explore various datasets provided by the Commodity Futures Trading Commission (CFTC) in its weekly Commitments of Traders (COT) report.

I highly recommend clicking the link above and trying it for yourself.

“What’s so special about this report?” you might ask. Well, dear reader, it tells us how many contracts are long or short across a myriad of futures and options products. It also provides insights into which categories of traders hold these positions.

If this still sounds vague, there’s a reason for it. Futures and options products are diverse—you can trade them on stocks, bonds, commodities, and currencies!

For example, here’s the combined futures and options positioning for gold:

Notice how the tool breaks down trader classifications into producer, swap dealer, managed money, other reportable, and non-reportable. It appears that managed money traders are the most long in their net position, while swap dealers are the most short on their net position.

What I aim to do with this kind of data is to find groups that disagree with one another and pit them against each other to see which group is more accurate. I don’t think it’s much of a spoiler for me to tell you that dealers often come out on top.

For example, if we take the difference between the net position of managed money traders and the net position of swap dealers, we get the following:

Now, I don’t want to suggest that dealers always win and managers always lose—that’s simply not true. However, it’s fair to say that dealers tend to outperform, while managers are often suboptimal.

Gold is interesting, but here on this blog, we focus on stocks. So, let’s take a look at how U.S. equity index positioning has changed since we last reviewed it.

Equity Derivatives Net Exposure

On the interactive tool, there’s a useful sidebar for “Group Summaries.” One of these is simply labeled “Equity,” and it refers to the net position for combined futures and options contracts across the S&P 500 E-mini, NASDAQ 100 E-mini, and Russell 2000 E-mini futures products.

Here’s what it looks like:

Just like with gold, we see that dealers and asset managers are at odds in their net positioning. Managers are the most net long, while dealers are the most net short.

If we once again take the difference between manager and dealer net positions, we get the following:

This is the chart that keeps me up at night.

Managers are heavily long here, while dealers are quite short. It’s the most dramatic shift in positioning in such a short time within this series. The “dumb money” went from being extremely net short in 2022 to extremely net long in just a couple of years.

Who would you bet on? The house or the gamblers?

Resolution

A second contained within a minute and a minute within an hour… At this rate, it will be a while until we arrive at a millennium. In the same token, let us not forget at what resolution we wish to invest.

Do you want to stay up on the latest news, the subtlest gyration in price, and every word that comes out of the regulators’ mouths? Then your resolution is one of tactics, speed, and efficiency of decision. Think quick!

Would you rather not have to act every day, every week, every month, or perhaps even every year? Then your resolution is one of strategy, gradation, and resolution of action. Take it easy.

Plurality

Make no mistake, neither of these dispositions should be considered “better” than the other. After all, both cheetahs and sloths manage to feed themselves. Or, if we consider scale instead of speed, a leaf feeds a mouse as well as a bale feeds an elephant.

It would be a mistake, however, to confuse ease with passivity and speed with activity.

I’m sure you know of someone who, despite the sound and fury of their tempestuous movements, gets nothing done. Conversely, consider someone who does exactly as they plan to do and nothing more. We can remain active and engaged whether we resolve to trade on a small timeframe or invest on a larger one.

Indifference

When I degrade passivity, I don’t mean that we should not let the investment ride, or fail to relinquish some control over the outcome of our process. When we let the market do its thing instead of fighting it at every turn, good things come. But we should know what we’re getting into.

It sounds fancy to say that you invest in a “diversified, broad-market equity index,” but let’s not kid ourselves. You’re likely buying the Vanguard S&P 500 ETF (VOO) or a similar product. The VOO is not some magical fund, we can dissect it!

Here are the top 10 largest components of the VOO:

The VOO is not voodoo.

When you invest in an ETF, you own shares of actual companies, and you pay a fee to Vanguard for the convenience. These companies have earnings, or lack thereof, and their stocks have prices. Just because you don’t want to think about these details doesn’t mean your money isn’t tied to their success.

For example, if you were to invest only in the VOO, then almost 20% of the money you put into it goes to three companies: Apple, Microsoft, and NVIDIA. A measly 0.6% of the companies in the index hold 20% of the capital.

So much for diversity!

Engagement

My intent is not to denigrate VOO owners, it’s a perfectly fine ETF with low fees. My intent is to ensure that you know what you own. Simply put, most passive investors don’t know what they own, and I do think that’s a problem.

Your money is a symbolic representation of your labor. Don’t ever let your hard work go into an investment that you don’t understand simply because “everyone else is doing it” or because “finance is too complicated.”

Put in a little bit of effort, just a touch of hard work, and I’m positive that you can formulate your own opinion on where your money is best invested.

A suit from a tailor fits better than one from a warehouse.

Hiccup

I had been using Yahoo Finance’s historical quote data for my analyses until the service was shuffled behind a paywall this past month. Not wanting to pay $480 annually for a data source that I had found for free a couple of years back, I began searching for alternatives.

Nasdaq offers free quotes for the stocks found on their exchange, with a full decade of the daily open, high, low, and close prices being available. This will suffice for the short-term analyses performed in the monthly reviews here, but I was used to data that went back to the mid-1960s.

With Macrotrends, we can scrape entire histories of quote data for the stocks available, but not without a catch. I’ve already found examples of poorly adjusted prices on some stocks (due to splits or spin-offs), so I don’t fully trust this data.

These alternate sources of data are serviceable, but I do miss the free, full, and quality data from Yahoo Finance. Furthermore, these other providers don’t have quotes for indexes and futures like Yahoo does.

If I manage to find a perfect replacement, I’ll let you know. For now, there will be no stock screening section in this month’s review (sorry).

September in Review

Even with the workarounds I found for losing access to Yahoo’s data, we don’t have the FTSE data this month. If updates to the United Kingdom’s index were your personal favorite, apologies guv’nor!

What moved? In a word: China.

The Hang Seng and SSE Composite indexes exploded higher, gaining a massive +18% on the news that the Chinese financial regulators would begin stimulating their floundering equity markets. I have been watching several Chinese companies, but currently do not hold any positions in them. This appears to be a promising development, but I’m not an expert on Chinese markets, to say the least.

Gold gained +6.1% in the most definitive positive move this past month. This serves as an interesting contrast to crude oil, which fell -7.2% in the most definitive negative move. This divergence could be pointing towards deflationary fears in the market.

While oil is signaling continued disinflation, gold investors aren’t convinced of that signal. They would rather move into the safety of gold if there’s a deflationary slowdown in the economy. Otherwise, these prices tend to move in tandem if it were simply an indication of inflation.

Look to the 10-year yield for further evidence of a flight to safety. Investors are buying, pushing yields well below 4%. Think about the slew of harbingers this month: the Fed cuts rates, oil prices plunge, long-term bond yields are shriveling up, and the gains in gold continue.

Hypothetically, if I’m still mostly invested in mega-cap U.S. equities, which have performed remarkably well since October 2022, I could be looking to rotate into long-dated bonds while there is still yield to capture, gold before even more of its performance escapes me, or now even Chinese stocks as they appear to be bottoming.

In reality, I don’t believe these are the best investment vehicles to rotate to, but this is what the typical equity investor could be thinking right now.

ChatPYT

The Oracle of Delphi, Pythia, has held a shrine in the Temple of Apollo since at least the 8th century BC, amassing followers who looked to the priestess for divination. Today, we have a similar fascination with artificial intelligence (AI), a computational black-box model that interests us for its divination-like qualities.

Let’s ask ChatGPT about what it “thinks” of the current setup:

What economic condition could this be signaling: US dollar moving down, 10-year yield moving down, crude oil prices moving down, gold prices moving higher, US equity markets roughly flat, foreign equity markets moving higher?

Its response:

This combination of economic signals could indicate several underlying conditions:

    Deflationary Pressures: The falling crude oil prices and 10-year yields suggest concerns about lower inflation or even deflation. Investors might be expecting economic slowdown, leading to reduced demand for oil and lower yields as they seek safe assets.

    Safe-Haven Demand: Rising gold prices alongside a declining dollar can indicate a flight to safety, as investors seek gold as a hedge against economic uncertainty or currency devaluation. This could be driven by fears regarding the US economy or geopolitical concerns.

    Weakening Dollar: A declining dollar can make gold more attractive to foreign investors and can also imply a lack of confidence in US monetary policy or economic prospects, contributing to the rise in gold prices.

    Mixed Equity Market Sentiment: The US equity markets being flat while foreign markets rise might reflect a divergence in economic outlooks. Investors could be rotating into international equities due to better growth prospects elsewhere or due to relative valuations.

Overall, this scenario could signal a cautious outlook for the US economy, with potential concerns over growth, inflation, and the effectiveness of monetary policy, while at the same time reflecting a search for safe-haven assets like gold amid a more favorable environment for foreign equities.

Thus, the oracle has spoken. Ignore her “wisdom” at your own peril!

Jokes aside, compare my perspective to that of the AI’s and you’ll find that we’ve highlighted similar concerns with the current financial landscape. It’s only a snapshot though, and we could always be wrong in our interpretation of the data.

Personally, I’m not betting against this particular divination.

Go Long

We’ve all heard this little number: “buy and hold.” Given the price history of the S&P 500 cash index (SPX), this isn’t such a bad idea.

What does this strategy look like if we plot our long position in black?

Oh, it’s… just a chart of the SPX, duh. In that case, we did pretty well just by going long and holding that position!

Perhaps you noticed the legend in the upper left corner, which assigns a number to something mysterious called the “CV.” This is the coefficient of variation (CV) of the daily log changes between close prices.

A coefficient of variation is simply the standard deviation of a dataset divided by its mean. Why use this measure? For a technical refresher, check out my previous article entitled “Signal to Noise.”

In short, the lower the coefficient of variation, the less noise we have compared to our signal. For us, the signal is performance, our return on investment. We want as much return and as little volatility as possible when investing.

Therefore, we could try to minimize the CV using strategies beyond “buy and hold.”

What was that really popular finance movie called? “The Big Short,” yeah. We should totally try shorting the market! What could possibly go wrong?

Cut Short

Full disclosure, I’m about to cherry-pick.

I’ve spent a lot of time looking at the SPX, maybe even a little too long. Regardless, it’s occurred to me that the best times to buy into the market would have been 1932/1942, 1974/1982, and 2009. Conversely, 1929, 1968/1973, and 2000/2007 would have been great times to get out. I’ve included some optionality because it’s not immediately obvious which years are better.

How will the computer optimize the timing of our shorts? We’ll allow it to wiggle around the dates for when we go short and again when we cover and resume our long.

Let’s try it out, plotting the periods of when we’re short in red.

The computer decided to short in 1929, 1973, and 2007. Those shorts were covered and longs resumed in 1942, 1974, and 2009, respectively. Besides during the Great Depression, these shorts are extremely tactical, lasting only a couple years out of multi-decade cycles.

Notice the reduction in the CV. Its value was reduced from 50.42 to 25.32. Because shorting simply inverts daily performance, these reductions in the overall CV are due to increased performance. The simulation cheats, however, because fees are not paid on the short position.

Shorting can seriously improve the investment quality of a portfolio, but it requires perfect intuition for timing the market, which is exceedingly rare. Could there be another way to increase investment quality without having to be a superstar trader?

Cash Out

What if, instead of shorting, the computer can choose when to cash out? The gains or losses for that day will simply be nullified until the long is resumed.

Let’s check it out.

The computer decided to cash out in 1929, 1969, and 2000. It got back into the market in 1942, 1982, and 2009. Cashing out is far more forgiving than shorting, lasting about a decade for each period instead of a couple years.

The reduction in the CV was even more pronounced than with our shorting simulation, now down to 23.4, a marginal improvement. Most of these gains are due to decreased volatility or by avoiding losses, but not by adding performance like with a short position.

Conclusions

While “buy and hold” is a perfectly valid strategy, investment quality can be improved if we allow ourselves to tactically short the market or strategically cash out from it. I generally don’t recommend shorting; there are many disadvantages for the typical investor, and most of the time you will lose money doing so.

However, I do recommend cashing out or at least transitioning your portfolio into more liquidity if you believe the time is right. It’s a way of retaining the gains of your investment, avoiding periods of high volatility or losses, and can be psychologically comforting to know that at least a portion of your portfolio is not at risk.

Making the Cut

On Wednesday, the board of the Federal Open Market Committee (FOMC) voted to cut the overnight Federal funds rate by 50 basis points (this is 0.5%, if you didn’t know).

Whenever the FOMC comes out with a new statement, I like to compare it to their previous statement and play a game of “spot the differences.”

From the committee’s perspective, job gains are no longer moderating, they have definitely “slowed.” Furthermore, core inflation remains elevated above the 2% target.

Despite this, the FOMC sees the dual mandate of maximum employment and moderate inflation as “roughly in balance” here. Perhaps I am too bearish, but this sounds to me like the Fed is saying, “this is as good as it’s going to get.”

Notably, Michelle Bowman offered a dissenting vote. She wanted 25 basis points instead of 50. These dissenting votes are quite rare, the past three were during the June 2022, August 2005, and August 2003 meetings.

Why Now?

I mean damn, what would Jeromey Romey Romey Rome think?

— Kanye West, “Bound 2”

During the Q&A section of the FOMC presentation, Jerome Powell offers a slew of data sources: the Personal Consumption Expenditures (PCE) inflation report, the Quarterly Census of Employment and Wages (QCEW) employment report, and the Summary of Commentary on Current Economic Conditions (Beige Book).

If we were truly diligent, we would read all these reports — you probably should. Instead, we’ll cheat and simply look at a chart of the most relevant series!

The complete history of these series is included, but look to the past half decade.

During COVID, interest rates were slashed to zero and the economy was stimulated with an unprecedented amount of liquidity. Unemployment shot up as the economy locked down and wages were supplemented. Then, a wave of inflation hit as Americans spent, not saved, their stimulus checks. Finally, the Fed engaged in one of the steepest rate hiking cycles in its history, going from 0% to 5.33% in 18 months. This rate was held constant for 12 months, until it was cut this past week.

Think about it: the Fed started easing. This means that less pressure is being put on inflation going forward, despite the rate of inflation at 2.6% being above the target of 2%. Why risk inflation picking up again?

It must be the labor market. Unemployment is still fairly low at 4.2%, but this is well above the local bottom of 3.5%. It’s enough of a move to trigger the Sahm rule, which typically coincides with where the National Bureau of Economic Research (NBER) eventually labels the start of a recession.

Landing

It’s been 52 months since the funds rate has been cut by 0.5% or more. When else has that condition been satisfied? February 2020, January 2008, and March 2001.

I’ll let you make of that what you will.

Dealers are Short

The Commodity Futures Trading Commission (CFTC) releases a weekly report on the Commitment of Traders (COT) which gives us a view into futures positioning among different trading groups.

For example, we could take a look at the positions of dealers, who tend to be “smart money” in any given trade. If we were to take the ratio between their short and long positions on the S&P 500 E-mini futures (ES1) market, we get the following:

Currently, the ratio between dealer shorts and longs is 6.06. For every long position, dealers have on about 6 short positions. In the chart above, you can see the green highlight corresponding to when this ratio is at least above 5. Not every highlighted period demarcates a good place to be liquid, but the majority do.

It’s important to note the range of this ratio. At the bottom of the bear market in 2022, dealers had a short ratio of 0.3, which means that for every short position, dealers had on about 3 long positions.

I wonder, what other groups could be trading the S&P 500 futures?

Managers are Long

You may want to believe that asset managers are competent, but I don’t.

Somehow, managers routinely bet against the house. While dealers are comfortable being short here, managers have gotten even more comfortable being very long.

You may be wondering about what’s different in the chart above compared to the previous. Here, the ratio on the bottom plot is inverted. We’re instead taking the manager’s long positions over their short positions.

Currently, for every short position managers have, there are roughly 6.6 long positions. The previous highs of this range were around 3.4, so managers are almost twice as extended as they ever had been in this series.

Once again, let’s consider the bottom of the range, which is close to 1.1. This means that managers have a long bias, unlike the short to neutral bias that dealers have. Crucially, the dealers have not over-extended themselves over their previous range, but asset managers are incredibly bullish here.

Is there some way we could consider both positions at once?

Utraque Unum

We can balance the ratios by taking their product and then the square root of that product. Really, this is just a fancy way to take an average of ratios.

Using this new interaction ratio, we can see that a highish value would be above 3.6, but we’re all the way up to 6.32! The bottom of the range is all the way down at 0.67.

I drew some ascending trendlines on the ratio plot and then associated vertical lines to the breakdown of these trends. While we have not yet broken the current trend, we appear to be approaching an inflection very soon.

When a rising bullish positioning trend breaks, you can expect a diminishing reward to risk ratio into the future. We could have the patience to allow this positioning ratio to moderate before deploying our cash into equity indexes in a big way.

Refresher

In order to not retread old ground, you might want to take a look at my previous post “Nonlinear Regression” for an overview of the methods used in this article.

We will be fitting our Bézier regression model in a similar fashion, using an optimizer to minimize the sum of squared residuals between the model’s predictions and the original data. If that sounds confusing, please read the previous article!

Bézier Curves

Bézier is the last name of a smart Frenchman and is pronounced “BEZ-E-AY.”

The method of drawing curves that Bézier created allows for smooth interpolations between a series of control points. Zone out on this animation from Wikipedia for a minute and see if it “clicks” with you.

Looks pretty, doesn’t it?

Think of the “base layer” as the labeled control points, of which we have four. The next “layer” is the green points. As our t variable moves from 0 to 1, each green point progresses from one control point to another. There are three of these moving points: one that moves from P₀ to P₁, one from P₁ to P₂, and one from P₂ to P₃. We’ve now reduced the four original points into three new points.

This process is repeated until the series is fully reduced to a single moving point. Notice that the blue points are the reduction of the green points, and the black point which draws the red curve is the reduction of the blue points.

Ultimately, four control points are reduced into a single smooth curve. Nice!

Getting Technical

Without getting bogged down in the details, this is R code that reduces a list of points into a function that can be called on to produce the point of the Bézier curve at time t.

bezier_evaluation <- function(pts) {
    function(t) {
        lst <- as.list(pts)
        while (length(lst) > 1) {
            for (i in 1:(length(lst) - 1)) {
                lst[[i]] <- (lst[[i + 1]] * t) + (lst[[i]] * (1 - t))
            }
            lst[[length(lst)]] <- NULL
        }
        lst[[1]]
    }
}

I could ramble about how cool this function is, but we can just use it instead.

xy <- c(0+0i, 1+3i, 2+1i, 3+2i, 4+1.5i, 3+1i)
f <- bezier_evaluation(xy)

t <- seq(0, 1, length.out = 100)
curve <- f(t)

plot(xy)
lines(curve)

We first create a vector of complex numbers that we’re calling xy. We then enclose a function called f that will evaluate a Bézier curve over these control points. Our t values will always be bounded between 0 and 1, so we sequence 100 values between these bounds. Finally, we can evaluate the curve using our function f on the t vector.

Plotting both the control points and curve looks like the following.

Never mind the X and Y axes being labeled “Real” and “Imaginary.” I did that to stress that we’re using complex numbers to make working with X and Y points easier. Something that should be more obvious in the plot is that the order of the points does matter. The curve begins on the first point and ends on the last, but the influence of the middle points on the curve changes as t progresses from 0 to 1.

The Main Attraction

Let’s start with the model output and work backwards. We’ll be using the S&P 500 cash index (SPX) as we always do.

Briefly, let’s discuss how it’s made.

First, we fit a log-linear regression on the data, shown in red. Next, we set up our control points along this regression line, spaced at equal distances from one another. While optimizing the curve, we let the Y-axis component of the control points move freely but anchor the X-axis component. As the X axis moves from 1927 to 2024, our variable t is moving from 0 to 1. We can run the optimizer a few times on iteratively updated control points, depending on how tight we want the fit to be.

The final control point positions are shown by faint blue circles. The final Bézier regression curve is shown in blue. On a logarithmic scale, we are able to reduce the sum of squared errors to 39.9 using Bézier regression, down from 157.2 using log-linear regression.

It’s crucial to note that the control points at the very start and end of the curve have a lot of influence and overfit those periods. Because of this, we should take the model outputs near 1927 and 2024 less seriously than the outputs near the middle.

Market Cycles

In the same fashion as the previous post on “Nonlinear Regression,” we can determine long-term market cycles using our model. To do this, we take the quotient between the Bézier regression (blue curve) and the log-linear regression (red line).

If you look at the previous post, you’ll notice that the peaks and troughs highlighted by these methods have a fair degree of correspondence. Lows include the mid 1940s, early 1980s, and the mid 2010s. Highs include the 1960s and 2000.

Unlike the sine-wave model, we are unable to project the Bézier regression into the future. Nevertheless, it appears that we’re approaching another peak.

Perhaps this method allows us to be… ahead of the curve!