Piece One reached an uncomfortable conclusion: markets are complex systems, they cannot be modelled with precision, and any framework that claims otherwise is mismatched to the thing it describes.

This is not a reason to give up on models. It is a reason to be clear about the mental framework you are using and to be aware of its limitations.

You cannot operate in a market without one. The question was never whether to have a model — it is whether the one you are using is robust and whether you are holding it the right way. The investor who says they don’t use a specific mental scaffolding to describe the investment environment is not being honest. The model is simply unexamined.

An unexamined model is dangerous. Faulty assumptions go unnoticed. And it is unknowingly rigid against change. Eventually it will break, usually at considerable cost.

A mental model does not yield formulas or forecasts. It is a working representation of a system — a structured way of understanding how the parts relate and what to expect when conditions change. It is the map, not the territory. Ancient seafarers knew this well. Beyond the limits of their maps, dragons dwelt.

That distinction is seemingly obvious, yet there are a great many examples of serious investors mistaking their models for reality — usually around the point reality began significantly diverging from modelled outcomes.

A map is necessarily a simplification. It incorporates the key features and omits detail. That is not a flaw. It is the whole point. A map you trust is useful. A map you have stopped questioning is dangerous. It implies you’ve forgotten it is just a map.

For two decades, government bonds and equities tended to move in opposite directions. When shares fell, bonds rose, and a portfolio holding both was steadier than one holding either alone. The standard policy portfolio is built on this observation. Some investors absorbed it as a permanent property of markets. It is not.

It was a feature of a particular regime — a particular inflation environment, a particular central bank stance. In 2022 that regime changed. Bonds and equities fell together, and the diversification a great many portfolios were counting on was not there. The relationship had been a useful map of one period, not of the whole territory.

And this particular fact is not buried in arcane economic theory. It is clear from both market history and a straightforward understanding of the real risk profile of fixed income instruments. A belief in a fixed relationship to equity returns is a misunderstanding of both.

Charlie Munger spent decades making one argument: that investing fails more often from poor thinking than from poor information. His answer was what he called a latticework of mental models — a deliberately multi-disciplinary toolkit the investor brings to every problem.

The instinct of analytical training runs the other way. It rewards depth in one domain. Professionals specialise. But under complexity and inherent uncertainty, a single framework used alone becomes a trap. Any framework can be genuinely useful and genuinely blinding at the same time, because by design it filters out everything it was not built to see.

The latticework works differently. Take one situation — a market that has climbed for several years, with the consensus firmly bullish — and put several lenses on it.

Psychology explains why the crowd is comfortable: recency bias, social proof, the comfort of company. History asks whether this configuration has occurred before and how it tended to resolve. Ecology contributes adaptation and crowding — the strategy that worked is now owned by everyone, which is exactly what makes it fragile. Physics contributes the concept of a critical state: a system that looks stable while conditions for a sudden change quietly accumulate. And careful attention to the language being used often reveals more than the explicit investment case does.

None of these lenses gives a forecast. Each gives a different part of the picture. The latticework is not a collection of theories to admire. It is a set of working lenses, each one sharpened by use.

Building one is the easy part. The harder discipline — the one that separates the investors who use models well from the investors a model eventually destroys — is how you hold it.

The central principle is one sentence. A model of a market must be able to change.

It sounds obvious. In practice it is one of the hardest disciplines in investing. Markets are adaptive systems — the participants change, the instruments change, the regime changes — so a model built for one environment is not automatically right for the next.

One thing has to be true before any of that is possible. The model has to be explicit. You cannot examine, stress-test or update a framework you have never actually put into words. Making the model explicit is the precondition for everything that follows.

Here is the mechanism. A model that works earns your trust. The longer it works, the more it earns. And a model you deeply trust is a model you slowly stop examining — not by decision, but by habit. Success quietly converts a working tool into an article of faith.

That conversion is invisible while it happens. You discover it only when the model has stopped working and you are still acting on it.

This failure has a name: model rigidity. The investor with a rigid model is often not wrong in their analysis — the analysis is internally sound, carefully done, consistent with everything the model says. The problem is that the model is describing a market that no longer exists.

Take an investor who has watched a cyclical sector for thirty years. Every downturn in that sector has been followed by a recovery. The investor builds a model: when this sector falls, buy it, because it always comes back. Then the sector is disrupted structurally — the way printed media was disrupted by the internet — not cyclically. The fall that looks like every previous fall is not like every previous fall. But the model cannot see the difference, because it was built before the distinction existed.

Model rigidity is dangerous precisely because it does not feel like error. It feels like discipline. This is not only an individual failing. Whole institutions ossify around a model. LTCM was partly this. So was every bank that, in 2007, ran a model in which national house prices did not fall.

The obvious correction is also a mistake. If model rigidity is the error of holding a model too long, the opposite error is abandoning a sound model too early. An investor with a good framework hits a bad stretch. The positions move against them. Under that pressure, they conclude the model is broken and drop it — usually near the point of maximum pain, which is often the point where the model was about to come good.

Both errors come from the same place: an inability to tell the difference between a model that has genuinely stopped working and a model that is simply having a hard time.

Conviction matters. Holding a sound position through the discomfort it will inevitably produce is not stubbornness. It is the discipline that makes investing work.

The resolution is that conviction and flexibility are not opposites. They are answers to different questions. Conviction is the right response to noise. Flexibility is the right response to genuine change. The job is not to choose between them as a personality setting. It is to work out, in each case, which of the two situations you are actually in.

How do you tell them apart?

You cannot tell them apart by how much it hurts. A sound model in a rough patch and an obsolete model in terminal decline produce the same sensation — money going out, conviction under strain. The feeling is identical, which makes the feeling useless as a signal.

The right question is never simply ‘is the price down?’ It is: ‘has the thing my model depends on actually changed?’ If the model rests on a particular mechanism — a cycle, a structural bid, a relationship between two variables — the question is whether that mechanism is still operating. Not whether the price is comfortable. Whether the mechanism is intact.

This is far easier to ask if you have done the work in advance. The most useful version: decide, while you are calm and not losing money, what specifically would tell you the model is wrong. Write it down. Name the conditions that would falsify it. Then, when the position is against you and your judgement is at its worst, you are not improvising a verdict under pressure. You are checking reality against a list you made when you could think clearly.

The best forecasters, studied formally over many years, are not the ones with the boldest convictions. They update in small, frequent increments — treating the model as a live estimate to be revised, not a position to be defended and then abandoned in one painful lurch. Updating early and often is not indecision. It is how a model stays alive.

There is no formula. No threshold, no indicator, no rule that turns the judgement into a calculation. There is only the habit — kept on a schedule, and applied hardest when the pressure is greatest — of asking whether the model you are acting on still describes the market you are actually in.

Sections One and Two described what a model is and how to hold it. The practical question remains: what should an adaptive model of a market actually contain?

Not a forecast. A working model is not a prediction machine. It is the structured understanding you bring to every market you look at — the thing that lets you interpret what you see rather than simply react to it.

Four things belong in it.

A working model is assembled from three things. Study: the latticework — psychology, history, the structure of markets, the nature of complex systems — comes from deliberate reading and thinking. There is no shortcut. Then experience, which only counts if it is recorded. Memory is not a reliable witness. It rewrites the past into something more flattering and more coherent than the past actually was. The decisions have to be written down at the time: what you did, what you expected, what you feared, what actually happened. And finally, the discipline of making the model explicit — actually stating what you believe about how markets work, and why. It is uncomfortable, because written beliefs can be shown to be wrong. That discomfort is the point.

The building never stops. A working model is a permanent draft. It is reviewed deliberately, on a schedule, and reviewed hardest after a large market move, after you have been clearly wrong, and most dangerously after you have been clearly right. A run of success is when a model quietly hardens from a tool into faith.

Piece One described the problem. Markets are complex adaptive systems. They cannot be modelled to precision, they generate genuine uncertainty, and the frameworks that promise otherwise fail in the conditions that matter most.

Piece Two has described the response. Not a better model in the sense the equilibrium school means — a more complete set of equations — but a working model in a different sense entirely. A map: multi-disciplinary, built from study and honest experience, held with conviction against noise and flexibility against change, and understood for what it is — a platform to stand on, not an oracle to obey.

The four components — the nature of the system, the structure of the market, its cycles, and yourself — are the foundation the rest of this body of work is built on. These two pieces are the groundwork beneath it.

Investing is a craft, not a formula. The uncertainty does not go away. What you can build is something adequate to meet it — a clearer map, held the right way, of territory that will never sit still.