Holy Grail and trend-following

 

If a trend-following system is too slow, you risk a Type II error by missing a turning point.

If a trend-following system is too fast, you risk a Type I error by reacting to noise.

The Holy Grail of trend following is a dynamic system to adjust speed depending on market/economic conditions.

- from Campbell Harvey Regimes Notes 

I think this is the best way to think about the trend-following problem - a choice between making a type II or type I. You cannot escape this problem. Reduce type II and you take on more type I risk. Statisticians will often try and to set the type I to 5% and then have a suitable type II at 10-20%. Traders have to ask the question of what are the cost differences between type I and type II errors. If you are too fast, you will increase trading costs while if too slow, you will miss opportunities. Harvey in his regime work looks at four states of the world based on observable market regimes which fits nicely into the idea that trend-following only focuses on price information.  The bull market has short and long-term returns both moving higher. A bear market has short and long-term returns moving lower. The rebound has short returns positive while long-returns are negative, and corrections have short returns negative versus long-term rates.

His work on regimes shows that return performance can be sorted by these four states, yet further work can be developed to account for other state variables.  

See 

It is all about the market regime or similarity

Momentum and turning points - Working through corrections and rebounds

Currency factors as cluster approach

 

There has been extensive work on currency factors such as carry, value, momentum, and volatility, yet currencies may be unique from equities. The movement of returns in currency may be based on factors that are based on how they may cluster. In "Currency Factors", the authors focus on clustering of currencies into baskets and not traditional factors. They find that G10 currency co-movements can be explained by a limited number of clusters, a dollar currency and a European currency cluster. These clusters can be further extended to a commodity factor cluster and a world factor cluster based on trading volume. This suggests that a mental model of viewing currencies within their cluster and then within traditional factors may be a method to form quick judgments on the co-movement across currencies.

There are limits to the value from a crowd of economists

Is there a wisdom of crowds effect for macro forecasts? The answer is yes, per the new paper "On the wisdom of crowds (of economists)", but the impact of looking at more economists diminishes quickly. Whether the MSE, the change in the MSE from adding another economist, or looking at the relative improvement, the answer is all the same.  Check or average a few economists but the marginal impact of looking at a large group is minimal. Most economists seem to come up with similar forecasts which is not surprising. No economist wants to be an outlier relative to their peers, and most economists use the same models or frameworks which means they are likely to derive the same result. There is no value from looking at a big crowd of economists on the big macro questions. 

Robo-advisors - keep the rules simple

 

More investors are using robo-advisors to get investment advice. Relative to doing it yourself, the robo-advisor may be an improvement. Is this better than a financial advisor is a different question and remains to be answered. We know that the robo-advisor is cheaper, so the investor is receiving net savings versus the standard fees that are usually charged. 

Do you get more sophisticated advice? A recent study shows that the advice given is rather simple and focuses on only a few factors - what is your horizon, goal, and loss reaction are the top three. These simple rules are driven a lot of client money and will tie the movement of savings to a limited set of variables. See "What drives robo-advice?"

Data as a hedge fund edge

 “Simple models and a lot of data trump more elaborate models based on less data....

We don't have better algorithms; we just have more data.” Peter Norvig.

There are various types of edges that will generate extra return. One of the most important in the gathering and using of information that others don't have or do not use efficiently. Of course, there is public information that many investors have, but there is value-added through: 1. transforming the data, 2. mixing the data with other data. 3. getting the data and processing faster. Data can be delivered systematically which requires the systematic processing of the data. There is a desire for using more sophisticate techniques with analyzing data yet keeping it simple may be preferred. It is easier to tell the narrative, and it is easy to see where something may be going wrong.

Time is a scarce resource

Mr. Gilder noted, “The scarcest resource is time, which always becomes scarce as other things become abundant. It is human genius that transcends the scarcity of time.” That’s economic productivity in a nutshell. - Andy Kessler WSJ 3/31/25

A variation on the time value of money, but we have to value time, and it becomes more valuable when there are more conflicts or things trying to grab our attention and time. Depending on the activity, we want time to speed up or slowdown, but unfortunately, we do not often do a good job of putting a price on time.

Managed futures and portfolio construction - A good way of seeing value

The people at 3Fourteen Research of providing good simple analysis for the value of trend-following using modern portfolio with actual performance data. This is not anything new, yet the narrative is compelling. Add managed futures to the classic stock/bond mix and you push out the efficient frontier. No surprise, yet even a 10% allocation will improve the return to risk trade-off. The gain comes from reducing bond exposure and switching to managed futures. 

The expansion of the efficient from trend-following is significant even relative to adding asset to simple stock bond mix. You can expand the asset menu and see an efficient frontier expansion, but if you are looking for more risk reduction, add the trend-following piece. 

We cannot say that the future will be the same as the past, yet these charts reinforce a compelling case for a simple strategy addition. 

 

Trend-following, trading capacity, and diversification

The folks at Quantica Capital have generated a provocative study called, "When trend-following hits capacity: A case study on commodities, exploring the hidden opportunities of limited investment universe diversification". Given the growth in trend-following programs, it is important to think about the issue of capacity. This is once again the age-old question of how many markets should you trade and what is the value of trading more markets. 

Quantica finds that the top ten markets in liquidity represent about 70% of the total available commodity futures liquidity. It is not exactly clear how liquidity is measured but the intuition makes sense. Energy futures dominate commodity futures liquidity along with gold, silver, and soybeans. All the other markets will be harder to trade. This is important because many of the futures that are not in the top ten provide significant diversification. We can measure the value of diversification through a simple measure of the Sharpe ratio is sum of individual Sharpe ratios for each market times the diversification multiplier that is related to the average correlation across markets and the number of markets in the portfolio. There is significant value with moving beyond the top ten markets even though you may not have better trend characteristics and there is less liquidity. Diversification is a benefit unto itself. You pay with lower liquidity, but you get the strongest diversification benefits from commodities. There is no guarantee of higher returns from holding more markets, but you get a strong tailwind from diversification.  

CTA dispersion - Not an issue, now what

Some have argued that trend-followers and CTAs have increased in dispersion because of differences in how to implement their strategies and the markets used. The folks from CFM have developed with some simple charts to suggest that current dispersion is not out of the ordinary. We might be able to say that trend-following is becoming less diverse. 

There is more sameness with managers which has implications on how many CTAs to hold and what to expect from your portfolio. If dispersion is down, then you need fewer managers in this strategy. If there is less dispersion, then there should be a greater focus on cost. Add value through cutting your fund manage expenses. 

Will this continue? As we do more research on portfolio construction and disseminate the information, the same good practices will be past among managers. We should expect more sameness unless there is a new creative thinking.

Recoveries will age but look for bad behavior first

An old paper that I forgot about but still relevant today is "Will the Economic Recovery Die of Old Age?". Recoveries can get long in the tooth, but most recessions are man-made. The cause a recession can be Minsky speculative excess, poor policies, or just old age. They are created from bad policy choices or the accumulation of excesses, yet like mortality tables there is a life span for any recovery. Of course, through good policies, the length can be extended. We can expect that current recoveries will have a longer life past on history, but the likelihood of failure will still go up as we age. We are better at getting policy right, but we are not perfect. The current recovery can still continue and the likelihood of a recession is still likely below double digits, but watch for signs. 

Below the 200-day moving average - life is different

"Nothing good happens below the 200-day moving average!" maybe attributed to Paul Tudor Jones, but often used.

The asymmetry of markets is strong, and markets will start act differently when prices fall below the long-term average. Is there a theoretical reason for this? No. It is a long-term technical and volatility starts to gain on the downside. This may not be a hard and fast rule, but you cannot go wrong with, at the minimum, using this as a basis for reassessment. There are other rules like the death cross of the 50-day moving average crossing the 200-day moving average. You may not get rich following rules of thumb but having points of reassessment is helpful. Disciplined watchpoints will reduce anxiety.

It is all about the market regime or similarity

 

Perhaps it is not stupidity, but the connections between market indicators tell us something about a regime and we have a high degree of similarity across market indicators we are receiving a signal of a regime. The simplest regime is the business cycle or more specifically, a recession. A market downturn changes everything and is our key systematic risk which requires a risk premium. There has been significant work to refine different regimes based on volatility, policy choices, and market environment with different levels of success. What is found that whether asset class or factors, there different asset or risk factors will behave differently based on the regime. If an investor can predict correctly, the regime, he can rotate portfolio allocations and improve overall return.

A new paper has been added to the regime research, "Regimes". It uses a simple methodology for defining regimes and then trying to exploit the relationship. The authors look at a set of macro variables that have different correlation. They transform the data into z-scores and then use a distance function to form a similarity metric. This similarity measure can be used as the basis for making portfolio decisions. This works because the z-scores of the economic variables have a persistence. If we can find periods of similarity between current values and the past and then look at the returns associated with that period, we can then make a judgement on what may happen next period. 

The interesting part of this work is that it is consistent with how investors think about the market. We look at events today and then search for similar event periods in the past and then extrapolate what may happen in the future. This is done on an ad hoc basis but using a similarity score it can be quantified.

Uncertainty shocks and the business cycle

 

One of the key developments in macro finance has been the modeling of uncertainty as a variable that impacts macro and financial variables. The unknown influences decision-making. A good review is provided in "Uncertainty shocks and business cycle research". Uncertainty is not the same as volatility. Uncertainty is measured by news and economic forecast disagreement and uncertainty shocks will impact the business cycle by changing consumer demand and decisions to invest. Risk premia are impacted by uncertainty which changes the cost of capital.

One of the most important variables for thinking about market moves is uncertainty. Along with impacting the real economy, there is a financial effect that is separate from volatility. If we do not know what possible states will exist in the futures, there will be a movement to less risky assets which independent of is displayed in macro data.

Asymmetric financial shocks - always greater downside risks

Financial shocks are asymmetric, that is, the impact of a downward financial shock is significantly greater than a positive shock. The asymmetry is not with the size of the shock but the sign of the shock. A negative shock creates a stronger reaction than a positive shock. The size shows a symmetry. See the paper "Machine learning the macroeconomic effects of financial shocks". Using Bayesian Neural networks (BNN) based on excess bond premium for inflation, industrial production, and employment. This is a simple straightforward approach for a machine learning model. Perhaps many think that the result is obvious, but it reinforces protecting the downside and preparing for bad news. You will not get this result from a simple regression model, so it is important to see how non-linear thinking can be incorporated in trading.  

The ongoing mystery of the gold price move

 

Gold is now over $3000 an oz. and like any market that has a strong run-up in price there is talk of a bubble. The bubble talk becomes stronger when it is unclear what are the drivers for the increase in gold prices. The classic story is that gold rises as an inflation hedge, yet inflation has fall from its higher levels post-pandemic. We have seen the jump from $1250 pre-pandemic to the $2000 level at the end of 2023. Now we are seeing a move to $3000 which suggests that a new story is needed to explain the move. 

Erb and Harvey in their new paper "Is there still a golden dilemma?" offer a new view on gold which is both interesting and compelling. First, they show that inflation is not volatile enough to explain the variation in gold. Second, they find that there is a distinct change in the gold market based on the composition of buyers. The introduction of gold ETFs has made it easier to purchase a form of gold return which has increased the trend. Make it easier to purchase and demand will go up. Three, there has been an institutional change in gold demand based on de-dollarization. Given the dollar is being used as a form of financial statecraft to punish bad actors, there is a move by several countries to reduce their dollar holdings. 

Erb and Harvey offer an explanation for the rising prices but there is a problem. This explanation just suggests that there are new buyers that are pushing the real price higher based on economic or political expectations. There is no central driving force that will suggest prices should be higher. Expectations dominate bubble thinking, so it is hard to know whether a higher real price is sustainable or will be reversed as in the past.

QT slowdown and closet easing

 

"The Committee will slow the pace of decline of its securities holdings by reducing the monthly redemption cap on Treasury securities from $25 billion to $5 billion." Fed Chairman Powell

The Fed is over their pre-pandemic balance sheet by trillions of dollars, and it now want to cut the redemption cap by 80% to $5 billion a month or $60 billion a year. Why pretend that you are making any effort to adjust your balance sheet.

We don't want to go into all of the balance sheet dynamics. The Fed is still holding a large MBS portfolio that will not run-off because prepayments have slowed. The still pays balances on reserves, so it has a negative carry portfolio. The slowing of QT helps the Treasury and reduces pressure on interest rates. The overall effect allows more "money" in the baking system which will make it that much harder to get down to the magic 2% target. There may be pressure to not lower rates, but the QT at $25 billion will show resolve without strong market impact. Cutting to $5 billion shows no resolve and generates a signal that the Fed is not ready to get back to normal.

Corporate spreads and signaling - good time for credit restructuring

 

We are in an equity market correction with a decline of 10% from the high. There has been a widening of spreads in credit markets, yet the overall signaling in credit still suggests a stable market. One, credit spreads are off their lows but still below the disruptive period of September 2024 when the Fed believed there was a need to cut rates 50 bps. Two, credit spreads are much lower than the bank crisis period of February 2023 and the period of equity market correction in 2022. Three, spreads are lower than the period of higher inflation post-pandemic. 

This may be the beginning of a larger correction which means that investor still have the opportunity to reduce their credit risk. Credit adjustments, short of a crisis, are slow-moving, so there is the threat of being early with portfolio rebalancing, yet given the low level of spreads, pulling duration is an easy way of offering portfolio protection.

The power of observation - visual intelligence

 

How are your powers of observation? When you look, do you really see? if you are a good analyst, do you see what other do not? This is an important skill that is often underappreciated or not even discussed. When you look at financial data, what do you see? when you look at a price chart, what do you see? everyone has the same information, so the only difference between analysts is their ability to see what others do not. Many great works of art have been around for centuries, but a first-time observer or even an observer for decades may see something new or different. 

Amy Herman, the author of Visual Intelligence: Sharpen your perception, change your life, has been running classes for years on training detectives and other professionals on being more perceptive. She takes NYPD detectives to the Met and asks them to look at painting and describe what they see. She wants to sharpen their thinking through looking at art. For example, what do you see in the famous Manet painting, "A Bar at the Folies-Bergere"? 

The detectives can then become better aware of their biases and perceptual filters. We often see what we want to see or what we are told to see, so we have to adjust our thinking to be more open to what is in front of us. Try not to look at a painting based on what we think we see, but on what is actually present. Look without bias and separate fact from opinion. Focus on the details. Ask first, who, what, when, and where, then take the next steps to why. Do not make faulty assumptions and learn to first be objective. Looking without bias will often allow you to see things in plain sight along with missing details. Say what you see, not what you think - the classic separation of fact from opinion.

If you are being unbiased, you need to avoid subjective words and phrases like, "obviously, clearly, never, always, actually or it goes without saying...". Investors should think about their forms of communicating through the 3 Rs: repeating, renaming, and reframing. This will clarify for the sender and receiver of information. You need to provide clarity to others what you are seeing. To avoid your biases, follow Herman's three rules: 1. become aware of your biases and boot the bad ones, 2. don't mistake biases for facts; instead use them to find facts, and 3. run conclusions past others to get a second opinion. 

This may be an odd conversation for an investment blog that often focuses on quant research, but good research is about better seeing of the facts. If you are a good observer, you will be good researcher. 

Future Babble - we are not good forecasters

 

Dan Gardner wrote a devastating book on the quality of forecasts, Future Babble: Why expert predictions are next to worthless, and you can do better. Can we make good predictions on the future? Gardner pours on the evidence that the answer is no, not in the least. 

Karl Popper states, "The course of human history is strongly influenced by the growth of human knowledge, but it is impossible to predict, by rational or scientific methods, the future growth of scientific knowledge." Our global and economic changes are based on changes in knowledge. Predictions about shortages are often wrong because experts do not account for technical changes. They underestimate the impact of knowledge.

It is an unpredictable world and that is especially for the experts. Experts may be good at describing the past and the present, but they do not have any edge on judging the future. At best, experts will say that we can expect more of the same. They are hard pressed to answer questions about change. 

Gardner described the difference between two forms of experts: the foxes and hedgehogs. Hedgehogs know one thing very well while the foxes are generalists. The generalists will often beat the specialists, so look to generalist thinking for predictions. 

While the author may leave the ready with a sense of hopelessness, we can control uncertainty through relying on diversification and planning for change. While we cannot make good predictions, we can always assume that the world will change. 

Geopolitical risks are in the rise

 

Geopolitical risk is on the rise. In fact, we are at the highest level since the beginning of the Russian invasion of Ukraine. The BGRI focuses on brokerage and news reports concerning geopolitical risks and the news is filled with uncertainty concerning the reaction to US policy and President Trump. The players of geopolitical risk may be following the same policies but the uncertainty concerns what will be the response of the US. The risk is focused on new responses which mean new reactions.  Markets abhor uncertainty, so we are seeing the results of this uncertainty.

Are these risks priced in the market? I hate this question because it is so open-ended. What does it mean that stock market has priced a major cyberattack? if there is a 5% chance over the next year, what does that mean in terms of a risk premium.  How do we price a terrorist attack? That said, it is the responsibility of global macro investors to think about this pricing effect and think about how it can be modeled. I may not agree with the risk map, but it can be viewed as a point of departure for deeper discussion and modeling.

Risk Barometer - It is not macro, but cyberattack that is the worry

 

I was reading the Allianz Risk Barometer for 2025 expected that there would be big changes in the ranking. I was thinking that, as a macroeconomist, there would be an increase in macro developments, political risks and changes in regulation as the top drivers. 

I was surprised that cyber incidents and business interruptions were the number one and two most important risks. However, upon closer reflection, many investors and business can hedge or trade around macro risk. They cannot hedge or trade if there is a cyber event. If there is a business disruption, there is not much you can do in the marketplace to solve or to find a profit. Step back and think about what you are doing to protect against these business risks.

Trends exist at almost all time scales

The new paper "Trend and Reversion in Financial Markets on Time Scales from Minutes to Decades" does an exhaustive analysis of trend behavior from the shortest time possible to long time periods and find good evidence of trend and reversion based on the strength of the trend. Weak trends persist and strong trends are likely to reverse. Reversal coefficients are relatively stable while trends show the strongest pattern in the 3-6 month period which is consistent with most trend-following managers. There is a lot going on with this paper given the large number of markets and time horizons tested and the number of transformations used to tease-out the results.

 The overall conclusions are summarized as follows:

• Intraday horizons at the tick level show strong reversal properties based on tick size
• Time horizons up to 15 minutes - weak trends reverse and strong trend persist
• Time horizon 30 minute to few hours - start to show trend patterns 
• Time horizon of hours to several days - small trends persist and strong trend reverse
• Time horizon from days to one year - trends persist if weak and a reversal of strong trends
• Longer time horizon - trend pattern exist out to two years then there is reversal of weak and persistence of strong trend which reflect longer-term cycles.

I learned from this paper is that weak trends will persist, but strong trends will reverse. Wait until the trend is very clear and you will be whipsawed. Look for the weak persistent trends but you will be subject to noise and the maddening effects of drawdowns and periods of poor performance as you wait to accumulate return.

The laws of simplicity should be applied to quant investing

 

I finished rereading The Laws of Simplicity. It is a short book that can be finished in one sitting, yet it is full of good ideas that take some time to internalize. I try and use these concepts with all of my research, but it is not always easy. Attempts to add alpha seem to be getting more and more complex with a lot of moving parts. The idea of simplicity is to eliminate the clutter and focus on core drivers, the more that must be estimated, the more that is likely to go wrong or subject to estimation error. 

The idea of trend-following is one of the simplest strategies that works. We find that momentum also works. The idea is simple, yet implementation can grow in complexity with risk management, positioning and exception analysis. The job of the research is to ask the question, "Is this feature really needed?", or "Can this model be made simpler?". 

Below are the ten laws of simplicity by Maeda. Do a walkthrough of any model and ask if it can be adjusted to be simpler.

Portable alpha - A useful tool, but questions

 

We have always been in favor of portable alpha as the best way to gain both beta and alpha exposure. Stop thinking about picking managers and pick the amount of beta and potential alpha. Don't pay a premium for beta. Pay for your alpha only. The following graphs are from Blackrock and Pimco and provide the basic framework.

The idea that you can engineer extra value always has appeal for investors, yet there is a need to look at the risk portion of the equation and the fact that alpha generation must meet a minimum to offset the costs of the engineering. 

One, the risk is based on the amount of leverage used. Do you want to have 100% equity exposure plus alpha or do you want to undertake this strategy without a higher notional value? The correlation for the alpha strategy with the market matters. Second, given there are costs with the strategy, there must be an expectation that alpha will be both positive and above the financing costs. There is also a need for the alpha to be uncorrelated with the market return so that any beta shortfall will be offset by alpha and lower risk.