The Map is Not the Territory: The One Thing Markets Will Never Price
Here’s an esoteric but useful way to think about financial markets today: they resemble Google Maps. Not the interface itself, but the experience of trusting it to always be accurate, and to make good decisions on one’s behalf.
The people who understand this best aren't the ones following the map - they're the ones who've seen enough rerouted traffic to know when the system is being inaccurate to everyone at once. This month, I chatted with Thomas Kuehn, Director of Enhanced Analytics, Global Structured Finance at Fitch Ratings. He’s someone whose career only makes sense once you hear how little of it was planned. I have a feeling that his experience in this regard is not uncommon.
Thomas didn’t set out to become a specialist in structured finance, credit modeling, or systemic risk. He’s quite explicit about that. His career began in market risk, then to managing a virtually credit-risk-free but market-risk-heavy US municipal bond investment portfolio, and then unfolded largely by accident. After resettling from Dublin - where he’d been working in European fixed income funds at Pioneer Investment - he found himself unemployed and looking for a role in treasury risk management.
What happened next is worth noting. One of the Big Three credit rating agencies needed a very specific combination of econometric skill and subject-matter expertise for what was supposed to be a short, one-off project.
“The project took a couple or so years longer to wrap up than anticipated,” Thomas told me, “and somewhere along the way, a contract role quietly became permanent employment.”
Since then, he’s moved across teams rather than up a single ladder - a path that’s given him what he describes as “a relatively unique ability to bring things together, draw from different fields, and think holistically.” It’s also why his answers resist tidy categorisation. He’s seen risk, modelling, regulation, and markets from enough angles to be suspicious of clean stories and absolute certainties all the time.
That matters, because our conversation kept coming back to a simple but unsettling idea: Markets don’t fail to be efficient because they lack sophistication - they fail because they’re working in a system where people don’t agree on what’s right, and responsibility is spread across institutions. This may sound familiar. In The Principles of Banking, I wrote that,
“Markets are dynamic, and the environment is constantly changing. Adaptability and an ability to respond quickly and effectively to events, whilst still retaining the long-term view, is the key to maintaining success in banking." —Choudhry, M., 2022. The Principles of Banking, 2nd ed. Singapore: John Wiley & Sons, preface p. xl
Which brings us, oddly enough, to Google Maps…
When Everyone Takes the Same Shortcut
Google Maps is a very efficient tool. It ingests vast amounts of real-time data, recalculates constantly, and optimises routes faster than any human could. And yet, if you’ve ever been sent down a “shortcut” that turned into a dirt road, a closed bridge, or a ten-minute detour through a residential rabbit warren, you know the problem. The system optimises for what it can measure, not for the unexpected.
Markets behave the same way.
As tools become more powerful - AI-driven research, automated models, ESG scoring systems, portfolio optimisers - we expect markets to converge toward efficiency. Instead, they keep producing traffic jams in new places. Capital piles into narrow channels, reverses violently, and creates opportunities that look obvious only in hindsight.
The system optimises for what it can measure, not the unexpected
The reason is not a lack of data or computation. It’s that the map is not the territory, and the territory keeps changing. Alfred Korzybski coined that phrase nearly a century ago, but finance continues to learn the hard way.
ESG as the Ultimate “Optimised Route”
ESG, Tom says, is a perfect example of map-induced distortion.
The premise sounds reasonable enough: compress the unmappable complexity of corporate behavior - supply chains, externalities, stakeholder impacts - into something investors can actually use. But the translation itself becomes the problem.
"ESG is a loosely defined concept," Tom says, and the looseness isn't a bug to be fixed. It's fundamental to what ESG is trying to do - turn continuous ethical reality into discrete investable categories.
"It's almost like organic labels, there are three or four different organic certifications for what's considered organic,” Tom explains. “The same underlying reality, multiple incompatible frameworks, each one arbitrary at the margins.”
In ESG, this means that a company two degrees removed from weapons manufacturing gets flagged. Another, equally connected but structured differently, passes through. (Not to mention the additional complication that, depending on a given ESG methodology’s underlying ethical system, certain weapons might score positively.) The classification feels technical, neutral. But it isn't describing the world, it's rearranging it. And as Tom says,
"Then suddenly, when companies get classified this way, there's a huge wealth transfer."
Not gradual, not proportional to underlying virtue. It’s sudden. Because capital doesn't flow toward goodness, it flows toward legibility. And when frameworks shift, whole categories of companies get repriced overnight.
This is ESG's deeper paradox: the very tools meant to make ethics investable create new forms of opacity. So more disclosure doesn't always reduce information asymmetry or arbitrage; it can concentrate it.
Just as Google Maps turns every driver into a traffic problem for the next street over, ESG channels capital into the same narrow paths of acceptability. This leaves everything outside the frame radically mispriced, not because it's better or worse, but because it's unmapped.
Whenever a framework redraws reality in this manner, someone will be compensated for identifying where the map is incorrect.
Arbitrage Is the Detour You Didn’t Know You Were Taking
Classic arbitrage (the kind that academic textbooks describe) got paved over decades ago. Obvious price discrepancies between identical assets don't survive seconds. But arbitrage itself didn't vanish; it migrated to wherever the map remains crude relative to the territory beneath it.
High-frequency trading found arbitrage in latency, but Tom's interest runs in the opposite direction. Not faster, but longer, messier, and more human.
"I still think the value is in long-term illiquid investments that take into account geopolitical transitions."
This is arbitrage at Google Maps' outer limits: assets with decade-long horizons, locked capital, outcomes sensitive to forces no algorithm wants to touch. For example, infrastructure in politically unstable regions. Forest estates exposed to climate policy shifts. Anything where the real drivers - regulatory change, demographic drift, institutional decay - live outside the spreadsheet entirely.
Markets can't be fully efficient, Tom argues, because the world they're supposed to price isn't. This is the knowledge problem Hayek identified: not that information is scarce, but that the most critical information is dispersed, contextual, tacit - the kind that never makes it into the dataset. (IMHO, he’s spot on 100%!).
Markets can't be fully efficient, because the world they are supposed to price isn't
And now we're handing that dataset to machines.
AI: A Better Map That Encourages Worse Driving
As Tom suggests, we've dramatically lowered the barrier to entry in quantitative finance. Models can be generated, coded, and tested by people who’ve never studied partial differential equations. He himself uses AI constantly, but he harbours no illusions about what it actually changes:
"I use AI and LLMs for coding; it handles about 80% of the Python work."
But here's what democratisation actually means: not that everyone gets the same edge, but that more people rely on the same abstractions. When the tools converge, the advantage migrates elsewhere - to proprietary data, institutional insight, and most critically, judgment about when the output is misleading or simply inaccurate.
Generative AI of the currently dominant LLM flavour excels at the divergent stage: searching, synthesising, even creating prototypes. Then it's Google Maps at its best. But like GPS, it has no understanding of why a route might be unsafe, politically sensitive, or temporarily blocked by something it can't observe.
It delivers answers without consequence, intellect without intuition.
Einstein warned that: "We should be careful not to make the intellect our God."
AI gives us intellect on demand, but what it doesn't give us is the feel for when something's wrong. And Tom is concerned that there's a quieter structural risk that compounds over time.
Junior analysts need to learn by doing
Junior analysts used to learn by navigating terrain manually - checking assumptions, reconciling inconsistencies, building an internal map through repetition and failure. Tom’s question is what happens when that formation process gets bypassed:
"The limit is really about how much you “cognitively” offload: If machines do all of that, organisations end up with pristine outputs and no one who knows how the system actually works." Tom further observes that, "Even the automation of mechanical routine tasks comes at a price, as the typical knowledge worker benefits from such quasi ‘downtimes’, which provide headspace for the processing of the trickier, value-adding problems they are working on.”
The danger isn't simply AI hallucination, though Tom's seen plenty; the bigger risk is institutional: losing the people who know when something doesn't make sense. Not because they ran the numbers again, but because they've developed enough feel for the territory to distrust the map.
Why Efficiency Breaks at the Edges
Markets fail to become efficient for the same reason GPS fails in edge cases: the boundaries matter more than the averages. Crises don’t respect neat categories, yet our models assume they do.
Tom saw this clearly during COVID. Households prioritised mortgages over credit cards while governments rewrote fiscal rules overnight, but financial institutions continued to model banks, insurers, sovereigns, and corporates separately - as if shocks politely stayed in their lanes. As Tom puts it,
“Everyone maintains separate modeling approaches… but there’s no total integration within a hierarchical economic model that encompasses them all. That’s something we could easily solve.”
Tom recalls pre-GFC US municipal bond ratings as the high watermark of sector-analytical silos devoid of an overarching cross-sector macro model: With BBB muni ratings displaying historical default rates comparable to AAA corporates, a fertile arbitrage ground was laid for structurers of novel financial products to exploit.
What stops us from pursuing a more holistic approach isn’t mathematics; it’s that integration would disrupt existing mandates and power structures. As Tom argues, efficiency isn’t blocked by ignorance, it’s blocked by incentives.
Those incentives don’t just prevent better models, they actively preserve mispricing. When reality is fragmented by institutional design, no single model captures the whole transmission mechanism, and gaps persist between how risks actually propagate and how they’re priced. That’s where opportunity survives.
The best investors use the map, but don't obey it blindly
This is why “alpha” doesn’t disappear as markets get smarter. The best investors, Tom believes, resemble experienced drivers: they use the map, but don’t obey it blindly. They know when a detour reflects reality and when it’s merely an artifact of the model.
“It’s really all about imagining scenarios,” Tom says, “and understanding the transmission mechanism.”
Markets aren’t inefficient despite better tools; they’re inefficient because those tools simplify a world that refuses to be simple. And like Google Maps, they’re most dangerous when they’re almost right. That is, trusted enough to follow, crude enough to mislead, and influential enough to send everyone in the same wrong direction.
Going forward….
In banking, having technical knowledge is one thing; knowing how to apply it under pressure is another. The Principles of Banking includes real-world examples, technical frameworks, and universal takeaways that professionals can immediately apply in their day jobs. But there’s no need to take my word for it:
“The Principles of Banking is easily the most important text for anyone in banking today and should be required reading for all personal development plans. When I was a regulator at the UK Financial Services Authority, managing the Change In Control team, I was responsible for assessing and granting regulatory approvals for complex banking transactions, such as Virgin Money’s takeover of Northern Rock. I relied heavily on Professor Choudhry’s text as a reference throughout the banking license approval process.
“Since then, I have referenced his book often, as a guide continuously during my career in banking as regulator, consultant, investor, CRO and now on the strategy/commercial side, and crucially while setting up a new bank and going through the bank license approval process myself.” —Nihar Mehta, Chief Corporate Development Officer, Monument Bank Ltd, London
For beginners and veterans alike, this book will act as a reference point, guide, and friend.
📚 Explore The Anthology 💥 20% off with code ANT20
📘 DiscoverThe Principles of Banking 💥 30% off with code PBC2E
