Top Traders Unplugged Sep 19, 2026 1h 15m 46m saved
With Richard Brennan, content producer at ATS Trading Solutions, who runs its trading program analytics and portfolio compilation
Richard Brennan spent three years testing the claim that trend following stopped working after the financial crisis, and found that the risk-adjusted payoff at a 20-day holding horizon had fallen to roughly a quarter of what it used to be while the 200-day horizon got stronger.
Allocators read the first half of that and concluded the strategy had been arbitraged away. Brennan read both halves and concluded the opportunity had simply moved to a slower part of the market.
"Well, from that research we undertook, we concluded that the edge had not disappeared, but rather it had moved."
Brennan, content producer at ATS Trading Solutions, on Top Traders Unplugged, runs the firm's trading program analytics and portfolio compilation, and has just finished a book built on about four decades of futures data across asset classes.
The full episode is covered here so you can skip it. 76 minutes of audio, 29 minutes of reading.
Here are the 17 insights that matter.
Key Takeaways
The edge in trend following moved rather than disappeared, migrating from fast signals to slow ones
Since 2020 the strongest risk-adjusted results sit around a 200-day horizon; the 20-day horizon pays about a quarter of what it once did
Large trends are structural features of markets, not anomalies waiting to be competed away
Leverage, forced activity and shifting capital flows produce moves that feed on themselves
Taking partial profits buys a smoother equity curve by selling off the positive skew that pays for everything else
He says the historical sample holds too few outliers to settle the question either way
A forecast everybody acts on destroys itself, which is the point of Brian Arthur's El Farol bar problem
Attendance averaged about 60 while the rules people used to predict it never stopped changing
Equilibrium models assume the pieces stay still; the Santa Fe Institute asked what happens while they keep moving
Markets contain both negative feedback, which pulls prices back, and positive feedback, which pushes them further
Only the second one explains why some moves travel far past their original cause
Increasing returns is why VHS beat the better tape format, and why the same dynamic shows up in trends
In a simulated market that let investors swap forecasting rules, the rules that survived were price-based trend-following rules
Every system with rules and incentives gets gamed, and the boundary looks strongest right before it breaks
A backtest is evidence about what happened, not a promise about what comes next
1. The Edge Moved, Not Gone
Brennan opened on the book he and the show's host have spent three years writing, due later this year. When they started in 2020, he said, the received view among allocators was that trend following's edge had weakened after the financial crisis and that the industry had grown too crowded. Neither of them believed it, but neither wanted to answer an opinion with an opinion, so they went to the data.
The research found a relocation, not a disappearance
Well, from that research we undertook, we concluded that the edge had not disappeared, but rather it had moved.
Richard Brennan
They examined futures markets across asset classes using about four decades of data. The tendency of those markets to trend was stronger in the most recent decade than in any of the three before it — strengthening, he said, over exactly the period the industry spent worrying about its disappearance.
What changed was where the payoff sat.
The strongest results are now at the slow end
So since 2020, the strongest results have concentrated around the medium to long term 200 day horizon
Richard Brennan
Fast signals lost about three quarters of their risk-adjusted payoff
such as the 20 day horizon that had fallen to roughly one quarter of its earlier level
Richard Brennan
That reconciles the two accounts. Managers trading faster horizons felt a real deterioration, Brennan said, and the investment community mistook the decay in one part of the landscape for the collapse of the whole thing.
The host agreed that slower models had been favored in recent years, and said the surprise for him was that the evidence kept strengthening over the three years it took to write the book. The data runs to 30 April 2026, and the manuscript is now in typesetting. Buyers will be able to register on the show's website for a free annual update of the figures, rebuilt each year out to 2030.
2. Outliers Are Structural
The deeper argument of the book, Brennan said, is about what outliers are.
Big trends are a feature of market structure
Well, I think the book argues that outliers, or these big trends, they're actually structural features of markets.
Richard Brennan
He named the ingredients that produce them.
Four things that make moves feed on themselves
So markets contain, interacting participants, leverage feedback, forced activity and changing flows of capital. Those ingredients, they create movements that feed upon themselves and travel much farther than anyone expects.
Richard Brennan
The structure persists; the timing and the location drift. That is a mixed blessing for anyone trading it.
The same uncertainty that preserves the opportunity makes it hard to hold
So that uncertainty actually helps preserve opportunity. But it does also make trend following difficult to practise because the returns don't arrive when and where the investor necessarily wants them to.
Richard Brennan
The host added his own view alongside it: that every portfolio should carry a meaningful allocation to trend following, sized for the investor's goals and their ability to stay with it, because nobody gets to pick the next market regime but anyone can own a process that adapts when it arrives. He also said he finds it strange that a strategy with more hard evidence behind it than any other still attracts the most skepticism.
3. Trend Up, Traditional Down
The show records a day early, so the weekly numbers were as of Tuesday 15 September, with the trend indices lagging to Monday night. The host put the TTU Trend Barometer at 52 the previous day.
The main drivers were energy and fixed income, with equities and most other sectors struggling. On the managed-futures side the month was strong across the board: "Soc Gen CTA Index up 3.13% so far this month, up 14.59% so far this year," and "The Trend Index up 3.23% in September, up 14.67% so far this year and the Short term traders index up 36 basis points in September, up 6.15% this year."
Traditional assets went the other way. The host read out "MSCI World down 1.55% so far this month, up 11.65 for the year," then the developed-markets index outside North America: "The EAFE meaning ex US and Canada down 2.41% so far this month up 11.49%." And then the part that mattered most: "Then we have the US Aggregate Bond Index also down 1.13% so far and for the year down 1.07%. So no offset, no risk mitigation there." Equities gave nothing back either: "The S&P 500 total return is down 1.23% in September and it's up 11.74% so far this year."
On the news behind the fixed-income moves, the host cited a "95% probability according to Reuters that the Fed will hike later today," with the Bank of England meeting the same week and not expected to move, and a possible Bank of Japan decision after that. He put European 10-year yields "sitting at 5%" and "3.55 on the 10 year bund yield," said "I noticed that Brent now is at 108," and flagged a second Reuters item he was not sure how to read, that "European physical cargoes for Brent is above $130 per barrel." He tied that back to Adam Rozencwajg's argument on an earlier episode that losing a billion barrels of expected oil supply eventually has to hurt. These levels were quoted live on air and are reproduced as spoken.
Brennan's read was short.
Higher rates are bad for economies and good for his strategy
The Fed no doubt they will put up interest rates. I think also December, they probably will put it up again. So it's a challenging period of time for the economies, but I certainly think it's going to be beneficial for trend following.
Richard Brennan
He added the obvious joke: maybe it is a good time to publish a book about trend following.
4. Why He Won't Take Profits
A listener named John asked whether taking partial profits earlier on trades that are up, say, 200% could outweigh the give-back on the few that run to three or four times the entry, given that closed profit can be reinvested with limited downside.
Brennan called it a sharp question that goes at his own philosophy, and conceded the mechanics work.
Partial profits do what the questioner says they do
So taking partial profits does convert open profit into closed equity that can therefore be redeployed and it can also reduce drawdowns and produce a smoother return path. But we are not creating that smoothness for free.
Richard Brennan
The cost is the shape of the return distribution.
You are selling the skew that pays for the failures
We're exchanging some positive skew for a smoother equity curve. So that's not wrong, but it's just a different strategy.
Richard Brennan
It could be tested historically — same entries, same initial risk, same reinvestment rules, then compare compound growth, drawdown and skewness. He does not think the test would settle anything.
There are too few outliers in the record to draw a conclusion
But the problem is we have very little statistical confidence in this history because the outliers we are depending on are so infrequent and few and far between.
Richard Brennan
One extraordinary trade can flip which approach looks better. And he expects the future to hold bigger events than the record contains, in both directions, because a larger sample explores more of what he called the fractal tree of markets — the big branches rather than the small twigs visible in the last 10 or 20 years.
He has looked further back than the track record goes
I've certainly done research back to over 125 years of research and I've seen some absolute doozies in the past
Richard Brennan
So his answer to John was no.
The error he is most afraid of is the one that costs nothing visible
So my worry is always what you call the type 2 error, missing those beauties, those outliers.
Richard Brennan
5. The El Farol Bar Problem
The host introduced the Santa Fe Institute for listeners who had not come across it: founded in 1984, bringing together physicists, economists, biologists and computer scientists to study complex adaptive systems — settings where many participants interact, learn, and change the system they are trying to understand. Investors form expectations, act on them and move prices; the new prices change the next round of expectations.
Brennan started with the El Farol bar, a real bar in Santa Fe that was busy on Thursday nights when it had Irish music. W. Brian Arthur, the economist who led the institute's work on the economy as a complex adaptive system, turned the problem of whether to go into a thought experiment.
The rule that makes the decision impossible
If fewer than 60 attended, the evening was enjoyable. But if more than 60 arrived, it became too crowded and those who stayed home made the better choice.
Richard Brennan
Everyone can see the attendance history. What nobody can work out is what that history implies for this Thursday. If all 100 people predict a quiet night, they all go and it is crowded. If they all predict a crowd, they all stay home and it is quiet.
Your answer is only as good as everyone else's answer
So how clever your choice is depends entirely on what everyone chooses, not what you choose.
Richard Brennan
Being contrarian is no escape, he said, if everybody turns contrarian at the same time.
We think we are forecasting; collectively we are deciding
We believe that we are using our models to forecast an outcome by ourselves. But where in fact, it's everyone's models which are actually deciding the outcome.
Richard Brennan
There is no answer sitting there to be found
The future price is not waiting to be discovered. Like the answer to a puzzle, it emerges from all the participant decisions, not just your own.
Richard Brennan
6. Rules That Defeat Themselves
The host pushed back: people forecast markets every day, so what does the bar actually add?
Arthur's answer was to give each simulated person several simple forecasting rules, such as this week resembles last week, or take an average, and let each one use whichever of their own rules had worked best lately. The diversity of rules was the point.
The same history, read a dozen different ways
This is how often we treat a market. Lots of people applying their different models to the same history.
Richard Brennan
Actual attendance then became evidence for choosing the next rule. When too many rules predicted a quiet bar, too many people went and those rules failed; rules predicting a crowd gained ground, and so on.
The average was stable while the rules underneath it were not
what he found, on average, was that the bar approximated 60 attendances on average, over the long term
Richard Brennan
No rule can stay on top, because using it changes the thing it measures
So no rule became permanently superior because its use of that rule helped change the environment in which it operated.
Richard Brennan
The conclusion Arthur drew from it
So what Arthur's experiment concluded was that the market's not a fixed outcome with one correct answer. The outcomes change partly because people keep trying to solve it and there's this reflexivity in the market.
Richard Brennan
7. Reflexivity And Bank Runs
The host recalled George Soros writing about reflexivity, and not understanding it on first reading. Asked to go deeper, Brennan gave the definition and then an example.
What reflexivity actually names
So reflexivity describes the loop between beliefs, actions and outcomes.
Richard Brennan
A solvent bank can be talked into insolvency in four days
So a bank may be fiscally sound on Monday, but a rumour spreads and depositors withdraw their money to protect themselves. Those withdrawals drain the bank's cash and by Friday, their collective actions may have helped create the very collapse they feared.
Richard Brennan
The market version runs through collateral: a rising price improves reported performance and strengthens collateral, which attracts capital and credit, which produces more buying and another rise.
Nobody is uncovering the future; everyone is writing it
So the future's not sitting there fully formed, waiting to be uncovered. Participants are helping write it step by step, if that makes sense.
Richard Brennan
The host summarized the loop back to him in five steps — expectations produce actions, actions change price and incentives, new conditions alter expectations, and the process can stabilize, reverse or reinforce. Brennan confirmed it and added the part that matters for where returns come from.
Trends and mean reversion both come from inside the market
The collective patterns associated with that create all of the features that we see in the markets. And that's all endogenous.
Richard Brennan
The host connected it to an earlier TTU conversation with Jean-Philippe Bouchaud, who had published work on endogenous and exogenous drivers of price. Most people assume a big trend needs a big news event, he said; the evidence says otherwise.
8. Respond, Don't Predict
If the future is being written rather than revealed, the host asked, how does a trader deal with that?
The whole method in one word
So I believe it's by responding rather than predicting.
Richard Brennan
An outlier hunter accepts that most attempts will be small, Brennan said, because a few exceptional moves drive most of the long-run result.
Which market produces the next one is unknowable in advance
Now we can't know which market will produce the next outlier or how far that outlier is going to travel.
Richard Brennan
So the process is built to survive being wrong most of the time
So we participate broadly, keep each initial exposure small and accept failed attempts as part of that search. But when a trend develops, we use a trailing exit that allows it to continue until the market shows us that the movement has changed.
Richard Brennan
And that is why there is no target price
So that's why I don't use personally fixed profit targets at all.
Richard Brennan
The decisions that shape the later stages of a move have not been taken yet, he said, which is an argument for a responsive strategy rather than a predictive one.
9. The Equilibrium Objection
The host raised the efficient market hypothesis and his early recording with Andrew Lo, who proposed the adaptive markets hypothesis and told him markets can be efficient, just not all the time. Why does a story about a bar challenge established economics?
Brennan tied efficiency to the older idea underneath it.
Equilibrium has been the organizing idea in economics for centuries
So by the mid-1980s, equilibrium had been a central organizing idea in economics for a very long time, hundreds of years.
Richard Brennan
He gave the textbook version: a fish stall where unsold fish push the price down and quick sales push it up, until supply roughly matches demand. Useful in many settings.
The trouble starts when it becomes the default assumption
But the problem begins when equilibrium is treated as the natural starting point for every market, particularly one in which participants differ.
Richard Brennan
The bar exposes why.
A forecast everybody follows cannot stay correct
There can be no single forecast that everyone can successfully follow because if everyone adopts it, their collective actions change the outcome and make the forecast wrong.
Richard Brennan
Santa Fe started from a different question. Rather than asking what balance looks like once the forces have settled, it asked what emerges while the pieces themselves keep changing.
A building site where everyone keeps redrawing the plans
So imagine people arriving at a construction site with different plans, so each can see what the others are building. And they revise their own plan. Every decision changes the site confronting the other builders.
Richard Brennan
Two approaches that start in different places
One starts with the conditions needed for balance. That's the equilibrium model. The other starts with interacting participants and asks what patterns emerge from them.
Richard Brennan
Arthur's own description of what an economy is
And Arthur described the economy less as a machine and more as an ecology in which the strategies compete, lose effectiveness and they evolve.
Richard Brennan
Brennan was careful not to overclaim. Equilibrium models are useful when the important relationships are reasonably stable, he said, and the Santa Fe group never said conventional economics was simply wrong.
10. Two Feedbacks, Both Real
Asked whether equilibrium still has a place, Brennan said yes — it is where mean reversion lives.
The forces that push back against a move
But markets certainly contain forces that resist movement. A high price can attract supply and reduce demand. High profits attract competitors which reduce margins.
Richard Brennan
Economists call that negative feedback. In plain terms, the system pushes back against the change, and mean reversion sits comfortably inside that framework.
The forces that push in the same direction as the move
But Brian Arthur showed that markets can also contain forces that reinforce change. A rising market may attract buyers whose buying pushes the price higher and attract still more capital.
Richard Brennan
Real markets carry both, and they are open systems: capital, participants, technology, regulation and instruments all change.
A resting point that never quite arrives
So a market can settle temporarily into an apparent or a quasi equilibrium without arriving at a permanent resting point.
Richard Brennan
Why the distinction decides whether trend following can work at all
So if every movement were quickly pulled back towards equilibrium, that would be a mean reverter's paradise.
Richard Brennan
Reinforcement, he said, is what gives a rational basis for expecting some moves to travel much further than anyone anticipated.
11. VHS Beat The Better Tape
Arthur's contribution that took longest to land was increasing returns, set against the familiar idea of diminishing returns. Add workers to a fixed patch of land and each new worker eventually adds less; the constraint dominates and pulls the system toward a limit.
He was ignored for years for saying the reverse can happen
But Brian Arthur studied situations in which the opposite can occur. And for many years he was totally disregarded.
Richard Brennan
How an early lead turns into a permanent one
So with increasing returns, expansion makes further expansion more attractive, and early advantage changes the environment in ways that reinforce that advantage.
Richard Brennan
His example was the format war. Betamax was widely regarded as the better picture.
Adoption, not quality, picked the winner
So as more households bought VHS machines, video stores stocked more VHS tapes, and the larger selection made VHS more attractive to the next household.
Richard Brennan
A small head start locked in the whole market
So a small early advantage created that environment that continued to favor the same technology.
Richard Brennan
The host had a personal stake in that outcome. In the late 1970s his youth football team went on an exchange to the United States, and the coach of the American side turned out to be Sony's marketing manager there — a man who had also been press secretary to Robert Kennedy at the time of the assassination. When the American team came to Denmark, the coach stayed with the host's family, talked about Betamax, and arranged for them to buy one of the first machines. Recording a football match and watching it again was, at the time, extraordinary.
Then the rental shops filled up with VHS. "So we stuck with this machine that could only really record stuff and we would have to watch the same football games over and over," the host said, adding that he had backed the wrong horse.
The host also drew the line forward to today's market. Arthur struggled for years to get the idea of increasing returns accepted, he said, and "Yet it is so predominant in the tech stocks of today." Windows won a dominant position while plenty of people argued the product was inferior, and the same mechanism sits under the technology sector now.
12. Path Dependence In Price
Asked whether increasing returns is just the modern network effect under an older name, Brennan agreed it is, and then widened it.
Network effects are the same mechanism
Yes, it is. It's definitely network effect.
Richard Brennan
Amazon is his example of the wider version. Without the bookstore and the reach it built first, he questioned whether the company would be the size it is. Artificial intelligence would not have arrived without the internet before it, because the models draw on internet resources and search.
Each step depends on the step before it
So each thing is contingent on another thing. It's a path dependent process. And increasing returns is where adoption helps create the winner.
Richard Brennan
In markets the same reinforcement runs through positions rather than products.
The mechanism works in both directions
So a rising price improves the performance of existing holders that may attract fresh capital to those holders of that capital. It strengthens collateral, it encourages further buying
Richard Brennan
Falling prices trigger stops, margin calls and forced selling, which is the same amplification pointed the other way. None of it says which move becomes an outlier or how far it runs — only why a market need not snap back to equilibrium once participants start reacting to the move itself.
The host asked whether reflexivity and increasing returns are two names for one thing. They are related and distinct, Brennan said.
Two different parts of the same process
So reflexivity is the loop between beliefs, actions, and outcomes. Increasing returns is a reinforcing process in which success or adoption changes the environment so that further success becomes more likely.
Richard Brennan
What they share is path dependence
So it's saying that how we arrived at this point in time can affect what happens next. The eventual outcome is not contained in the starting conditions alone.
Richard Brennan
That is why he holds no view on how long a trend will last. The very next iteration might contain the reinforcement that keeps it going.
The practical instruction it gives him
So it tells me not to assume that a movement must end simply because it's traveled a long way.
Richard Brennan
A move that can reinforce itself will not necessarily do so — the forces extending it always compete with the forces resisting it.
Which is why the positions are many and small
So I therefore spread many small positions across many markets because I don't know where the outlier will appear.
Richard Brennan
The line that sums up his method
So this is why I believe trend following is a response, it's not a prophecy.
Richard Brennan
13. An Artificial Stock Market
The Santa Fe group went further than the bar and rebuilt the problem with prices.
A simulated market where everyone brought their own forecasting rules
So they created a simulated market populated by investors using different forecasting rules.
Richard Brennan
Forecasts determined whether each investor wanted to buy or sell; combined orders moved the price; the new price became evidence for judging which rules had worked. When the simulated investors adapted slowly, the market stayed calm.
Searching harder for better rules created the turbulence
But when they searched more actively for better rules, calm periods were interrupted by clusters of turbulence.
Richard Brennan
And one family of rules won.
Trend following emerged from the competition on its own
Price based strategies emerged. Trend following strategies started to become preferred.
Richard Brennan
The investors were handed a whole range of forecasting rules at the start and settled on the price-based ones, which survived the competition between strategies.
What a simplified model was able to show
So complicated market behaviour did not have to come entirely from outside news. It was also being created and amplified by interactions among these participants.
Richard Brennan
Asked when this was done, Brennan dated it to the institute's founding in 1984, recommended M. Mitchell Waldrop's book Complexity for the history, and named Murray Gell-Mann, author of The Quark and the Jaguar, as one of the eminent figures in the story. The work was revisited recently at a Santa Fe commemorative event, and Brennan wrote a series about it on the ATS Trading Solutions site.
14. All Systems Will Be Gamed
Arthur also wrote an article arguing that every system gets gamed. Brennan was clear about what the phrase does and does not mean.
Gaming is usually done inside the rules, not against them
So gaming doesn't necessarily mean breaking the rules. It can mean following them in a way that earns a reward while defeating their purpose.
Richard Brennan
He tied it to Goodhart's law.
Once a measure becomes a target, people manage the measure
So this is when a measure becomes a target, people begin managing the measure rather than the result it was intended to represent.
Richard Brennan
The bonus scheme that stops measuring anything
Someone discovers that opening several accounts for the same customer earns several bonuses. So the account total rises, but genuine customer growth doesn't.
Richard Brennan
Markets are full of the same material — benchmarks, risk limits, margin rules, expected policy responses.
Participants reshape the system by organizing around it
Participants organize around them, they exploit them and they protect themselves from their consequences. And through that adaption they change the system itself.
Richard Brennan
He compared it to a dam wall: keep testing it and the crack gets found eventually.
Every agent in the system is hunting for the same edge
The agents in it are always trying to compete in that system and get the benefit or the edge.
Richard Brennan
Asked whether that is how a stable market turns into an unstable one, he said yes, and described the sequence.
Calm is what builds the positions that break
So for example, a long period of stability encourages people to assume it's going to continue. So under that circumstance, leverage rises, hedges shrink and positions accumulate around boundaries that have held for years.
Richard Brennan
A disturbance then exposes the dependence: stops trigger, margin calls arrive, lenders withdraw and buyers step back, each response moving price and activating the next layer.
An outlier is a change in relationships, not a big fluctuation
An outlier may therefore be more than just a large ordinary fluctuation. It can be the visible result of relationships changing as participants respond to one another.
Richard Brennan
The sentence to take out of the section
And a boundary can look strongest immediately before the behavior built around it helps produce that break.
Richard Brennan
15. Forty Years Is Not A Law
The host came back to the length of the record: if something has worked for 40 or 50 years, surely that counts for something.
It counts, Brennan said, but not for as much as people want it to. A track record comes out of an environment with particular participants, incentives and costs, and those conditions do not stay fixed. A successful pattern attracts capital; other participants copy the rule, anticipate its trades, or hunt for its weak points.
The bar problem again, in the form that matters to an allocator
Average attendance can appear stable while the rules underneath it are continuously changing. A pattern can persist while the reasons producing it evolve.
Richard Brennan
That, he said, is non-stationarity in practical terms: relationships change over time. More history helps you judge whether an idea has survived different conditions, but it cannot convert an evolving relationship into a permanent law.
What a long record does and does not buy you
So the backtest, it provides evidence, but it's certainly not a promise for the future.
Richard Brennan
The host asked the obvious follow-up: how do you stop non-stationarity becoming a convenient excuse for ignoring results you dislike? By being disciplined about what evidence can and cannot tell you, Brennan said. Historical testing does reveal turnover, drawdowns, concentration and reliance on exceptional trades, and it does reject a rule that repeatedly fails to produce the behavior it was designed to capture.
A realized average return is not a constant of nature
An average annual return calculated from one realised history is evidence about what occurred. It's not a physical constant.
Richard Brennan
The rule he applies to himself
It should never become an excuse for ignoring evidence we don't like.
Richard Brennan
16. Roulette Balls Don't Learn
The host raised a comparison with the casino. Ed Thorp and Claude Shannon built a computer in the 1960s to predict where a roulette ball would land. A decade or so later, Doyne Farmer, Norman Packard and their colleagues ran their own version using physics — measuring the wheel and ball while the table was spinning, with a small digital computer eventually concealed inside a shoe.
It gave them a zone, not a pocket
it didn't predict the exact pocket, but it determined with a degree of confidence which zone it would fall into
Richard Brennan
Then came the difference that makes markets a harder problem.
The wheel never adapts to the people betting on it
But the thing was, in this roulette experiment, the roulette ball doesn't learn, unlike our reflexive markets.
Richard Brennan
Three things a market can do that a ball cannot
The roulette ball can't copy the methods. It can't anticipate the bet or change its behavior to remove the advantage. But this is where a financial market can.
Richard Brennan
He reached back to an earlier episode for the image.
The physicist's version of the same complaint
And remember, in our podcast we had with Dave Dredge, we talked about Richard Feynman's famous quote about imagine if electrons had feelings, how difficult physics would be.
Richard Brennan
Farmer carried the same measure-model-test approach from the roulette table into markets, Brennan said, but the object of study had changed underneath him. In markets, learning, adaptation and competition are part of the system.
His verdict on trying to forecast one
And that's what these complex adaptive systems are. They're menaces to try and predict.
Richard Brennan
17. What A Backtest Can Prove
The host asked what a backtest can honestly establish if markets adapt.
It establishes one thing and one thing only
Well, really a backtest shows what specified rules would have done through a known historical sequence.
Richard Brennan
It shows turnover, drawdowns, failed trades and exceptional winners.
And it cannot establish the thing everyone wants from it
But what it can't establish is that the same environment will recur in the future.
Richard Brennan
So he uses it as a design check, not a forecast
So I personally use it to test whether the rules produce the behavior I want out of my models. I don't treat it as a map showing where the next outlier is going to appear or how large it will become.
Richard Brennan
Asked how a trend follower prepares for a future that cannot be mapped, he answered by shrinking the amount of forecasting the system has to get right.
Fewer things have to be predicted correctly
So the best thing we can do is reduce how much needs to be predicted correctly using our system.
Richard Brennan
Search broadly, keep each loss small, use trailing exits so an exceptional move has room to run, and keep the rules mechanical so they respond to price without the trader rewriting them. Research is where the design gets reconsidered; daily execution is where it gets followed.
Modesty about forecasting is not permission to be undisciplined
And complexity gives us a reason to be modest about prediction. It's not an excuse to abandon discipline.
Richard Brennan
With about ten minutes left, the host asked for the three things a listener should take away.
One: you are inside the system you are forecasting
So first, market participants are inside the system, so their beliefs and actions help create the outcomes they are trying to anticipate.
Richard Brennan
Two: some moves feed themselves, and that explains trends without locating them
The second is that some market movements actually feed themselves. So that helps explain these persistent trends and exceptional outliers, but it doesn't tell us where the next one's to going to appear.
Richard Brennan
Three: history is evidence and nothing more
And third, history is evidence, but it's not a promise for the future.
Richard Brennan
Which leaves one response available
So our response is therefore to participate broadly, keep failed attempts manageable and leave room for the movement that travels much farther than anyone expected.
Richard Brennan
And the picture he would replace the machine with
So that's why the market is not a machine, in my opinion. And I'm much more like Arthur. I believe it's innovative, evolving ecology of participants that respond to a world their own actions are helping to create.
Richard Brennan
Bonus Insights
Japan now has more than 100,000 people aged 100 or over
The host mentioned the figure as a piece of news that hit his screen that morning. Brennan's explanation was one word, twice: "It's got to be diet. It's got to be diet, I'm sure of it." The host's follow-up was that they need a strategy for the long run.
Neither of them could place the phrase about stability
Asked whether "stability begets instability" is a Minsky moment, Brennan said, "It might be a Minsky moment. It certainly sounds like it should be."
The book has a firm data cut-off and a soft publication date
Brennan would not give a date beyond later in the year. The figures run to 30 April 2026, and the book is in typesetting and production.
The reading recommendation is a history, not a textbook
Brennan plugged Waldrop's Complexity as the way into the Santa Fe story, naming Brian Arthur, John Holland and Stuart Kauffman as the people it brings to life: "So it's not only a story about science, it's the human story about them and the struggles they had in getting their eyes and ears across."
The host is off for a fortnight
Alan Dunne takes over the series for the next couple of weeks, with Nick Baltas expected as the co-host on the first of them. Listener questions still go to the show's usual email address.
Richard Brennan's bottom line
Markets keep producing large trends because participants react to each other rather than to a fixed set of facts, which means the payoff cannot be located in advance — so the only workable response is to spread small bets widely, lose often and cheaply, and leave the exit open on whatever starts running.
Products, Companies & Tools Mentioned
ATS Trading Solutions (Brennan's firm; he publishes his series on the Santa Fe Institute's work there, and runs its trading program analytics and portfolio compilation)
The Santa Fe Institute (Founded 1984 to study complex adaptive systems; the source of both the El Farol bar problem and the artificial stock market Brennan described)
The SG CTA Index, SG Trend Index and SG Short Term Traders Index (The managed-futures benchmarks the show reads out each week; all three were up in September and up double digits for the year)
The TTU Trend Barometer (The show's own trend gauge, which the host put at 52 — a strong reading, in his description)
MSCI World and MSCI EAFE (Both down for the month on the host's numbers, against a managed-futures sector that was up)
The S&P 500 (Down on a total-return basis in September on the host's figures, and up about 11.74% for the year)
Sony (Maker of Betamax, which Brennan used as the case study in losing a format war despite the better picture; the host bought one of the first machines through Sony's US marketing manager)
Amazon (Brennan's example of path dependence — he questioned whether the company reaches its current size without the bookstore and the reach it built first)
Microsoft Windows (Raised by the host as a second case of increasing returns beating product quality)
The Federal Reserve, the Bank of England and the Bank of Japan (All three meeting the same week; Brennan expects another increase in December as well)
Reuters (Source of both the Fed-odds figure and the report on European physical cargoes the host read out)
Books & Resources Mentioned
Complexity: The Emerging Science at the Edge of Order and Chaos – M. Mitchell Waldrop (Brennan's recommendation, twice: the history of the Santa Fe Institute and its fights with economic orthodoxy)
The Quark and the Jaguar – Murray Gell-Mann (Named as the work of one of the physicists in that story)
All Systems will be Gamed: Exploitive Behavior in Economic and Social Systems – W. Brian Arthur (The article behind Brennan's section on benchmarks, risk limits and Goodhart's law)
Increasing Returns and the New World of Business – W. Brian Arthur (Arthur's own popular statement of the idea that took years to be accepted)
The Trend Following Manifesto – Niels Kaastrup-Larsen and Richard Brennan (The three-year study behind the first half of the conversation; data to 30 April 2026, publication expected later this year, with free annual updates to 2030 for buyers who register)
Adaptive Markets – Andrew W. Lo (The hypothesis the host recalled from an early recording: markets can be efficient, just not all the time)
The Billion-Barrel Shock Has Not Arrived Yet ft. Adam Rozencwajg (The earlier TTU episode the host referred back to on energy, and the source of the missing-billion-barrels argument)
Volatility Series ft. Jean-Philippe Bouchaud (The earlier episode on endogenous versus exogenous drivers of price that the host said he had underrated at the time)
Why the Best Portfolios Are Built to Be Wrong ft. David Dredge & Richard Brennan (Where the Feynman line about electrons with feelings came up)
Listen to the full episode
🟢 Spotify | 🟣 Apple Podcasts | 🔗 Episode page
Listen to the full episode:
More Richard Brennan interviews and podcast appearances
If this was worth your time, send it to someone who has to have a view on this.
Get the latest market chatter as it happens: