All on the Line · Mathematics
Financial models compress an uncountable world. Mistaking compression for completeness is where fragility begins.
Models and compression
3 March 2026
6 minutes
There are more numbers between 0 and 1 than there are whole numbers in existence. This may feel counterintuitive, but it’s a fact. The whole numbers stretch without end: 0, 1, 2, 3, 4, continuing forever. And the interval between 0 and 1 feels narrow, confined, almost trivial by comparison. Yet in 1874, Georg Cantor proved that the infinity contained within that small interval is strictly larger than the infinity of all whole numbers combined, and in doing so he forced mathematics to confront the fact that infinity is not a single undifferentiated idea but a hierarchy with structure. Some infinities are larger than others, and one infinity can fit inside another.
The infinity of whole numbers is what mathematicians call countable. That term does not imply finiteness; it means that the elements of the set can be placed into a sequence where each element corresponds to a natural number. Even though the list never ends, the pairing exists. Cantor’s breakthrough came when he examined the real numbers between 0 and 1, which include every possible decimal expansion, terminating decimals, repeating decimals, and those that continue without pattern forever. Through what is now known as the diagonal argument, he demonstrated that no matter how long your list is, no matter how many decimal expansions you write down, you can always construct a new number by changing the nth digit of the nth entry. That new number will differ from every number in the list at least at one decimal place. The consequence is not merely that the list is incomplete, but that completeness is impossible. The set is uncountable. Its cardinality exceeds that of the natural numbers.
The result becomes even more unsettling when one recognizes that the number of real numbers between 0 and 0.1 is equal to the number between 0 and 1, equal to the number between 0 and 100, and equal to the entire continuum of real numbers altogether. When mathematicians compare sizes of infinite sets, they don’t use length, they use the idea of a one-to-one correspondence (a bijection). If you can pair every element of set A with exactly one element of set B, with nothing left over on either side, the two sets have the same cardinality. For example, take every real number x in the interval (0,0.1), multiply it by 10, and you get a real number 10x in the interval (0,1). Multiply it by 1000, and you get a real number 1000x in the interval (0,100). Every number in (0,0.1) pairs with exactly one number in (0,1) and (0,100), and vice versa.
So shrinking the interval in the real numbers realm does not shrink the infinity. Between any two distinct real numbers, there are infinitely many more. Density persists at every scale. Cantor did not merely expand mathematics; he destabilized our instinctive understanding of magnitude. Large ceased to mean what it seemed to mean. Infinity was no longer a horizon, but a layered landscape.
In last week’s column, I examined the cost of perfect optimization through the Traveling Salesman Problem. A route with 30 stops yields 30 factorial possible permutations, approximately 2.65 × 1032 distinct routes. To evaluate each one at a trillion computations per second would require more than 600 times the age of the universe, which itself is approximately 13.8 billion years. The lesson was not that optimization is futile but that brute-force perfection becomes computationally infeasible long before problems appear intuitively large. Discrete combinatorics produces explosion at a pace that overwhelms hardware.
That example dealt with finite but astronomically large sets. Cantor’s insight exposes something more profound: certain infinities are not just large, they are unenumerable. Financial markets operate inside continuous state spaces, which fall into the unenumerable infinity space. Asset prices, interest rates, exchange rates, volatility surfaces, and credit spreads are represented mathematically as real-valued processes evolving through time. In the framework of modern quantitative finance, these processes are modeled using stochastic calculus, where the space of possible trajectories is not simply vast but uncountable.
A Monte Carlo simulation with 10,000 paths may feel thorough. Increasing it to 100,000 improves convergence. Pushing it to one million refines estimates further. Yet each simulation remains a countable subset of an uncountable universe. Between any two simulated price paths, there exist infinitely many trajectories that were not generated. This means that we are just able to model a thin slice of possible trajectories, and that gap is not technological. Even with infinite computing power, one cannot list an uncountable set. Cantor’s distinction is structural, not computational. Optimization problems confront limits of speed, but continuous systems confront limits of logic.
Financial models do not attempt to enumerate every possible future. They compress an infinite continuum into tractable representations. Measure theory provides the mathematical foundation for this compression by allowing us to define probability measures over continuous spaces and compute expectations without listing individual outcomes. Instead of cataloging every trajectory of an asset price, we integrate across a distribution and extract summary statistics: expected returns, variances, tail probabilities, scenario outcomes. The mathematics is rigorous and internally coherent. The abstraction is powerful.
Compression, however, is not completeness. Increasing the number of simulations reduces sampling error but does not alter the underlying structure of the possibility space. Between any two simulated scenarios lie infinitely many that remain unobserved. No expansion in computational resources closes that structural gap. A model remains an approximation because the territory it describes is richer than any representation we construct.
The necessity of compression becomes obvious through a simple analogy. Imagine building a map that records every grain of sand, every pebble, every microscopic contour of a landscape. Such a map would replicate the territory with perfect fidelity, but its very completeness would render it useless. If your map included every grain of sand, you could not fold it, carry it, or deploy it for navigation. A model that attempted to incorporate every conceivable detail would collapse under its own weight. Therefore abstraction is not a defect, it is the condition of usability.
The discipline lies in understanding what the compression omits. It is tempting to say that models fail to incorporate events such as 9/11, the subprime mortgage collapse, or COVID because those events were unimaginable, but that explanation is too simple. Pandemics were known, systemic banking crises were documented repeatedly across centuries and correlation breakdowns under stress were observable in emerging market crises long before 2008. The issue was rarely ignorance of possibility; it was calibration and structure. Most financial models are anchored to historical data drawn from relatively stable regimes. Correlations are estimated from periods in which liquidity is abundant, volatility is contained, and institutions behave predictably. Tail events, however, are often regime-shifting phenomena. They alter correlations, liquidity dynamics, and behavioral feedback loops simultaneously. When the regime changes, the distribution itself changes. The tails were there, it’s just that they were thin because the compression assumed continuity of structure. What was mischaracterized was not mathematics, but the persistence of the regime.
There is also an institutional dimension that is harder to discuss but impossible to ignore. Assigning meaningful probability mass to severe downside scenarios has immediate economic consequences: expected returns decline, capital requirements increase, leverage becomes less justifiable, growth projections tighten. An ultra-conservative model may be mathematically defensible, but it produces numbers that appear less attractive in board presentations and investor decks. Institutions compete on performance, not on prudence that never gets tested. The pressure to present robust earnings and efficient capital allocation subtly incentivizes thinner tails and optimistic compression. This is not necessarily the result of recklessness; it is the natural byproduct of incentives interacting with uncertainty. The danger, therefore, is not that models fail to include alien invasions. It is that plausible discontinuities, events within the structural domain of the system, are assigned negligible weight because incorporating them seriously would materially alter today’s reported outcomes.
Recognizing that we operate within an uncountable continuum does not imply that every imaginable event must be modeled. An extraterrestrial invasion is logically possible but economically irrelevant within the defined scope of most macro models. Its probability mass is, for all practical purposes, negligible. The challenge is not to incorporate every conceivable scenario but to define the boundaries of the model with clarity and to assign appropriate weight to regions of the outcome space that carry structural significance. The real danger lies in low-probability, high-impact events that reside within the plausible domain of the system and whose probability is small but not vanishing.
Cantor’s hierarchy of infinities offers more than a mathematical curiosity; it provides a lens for thinking about uncertainty. Financial systems resemble the uncountable more than the countable. The possible paths of an economy, a balance sheet, or a market through time belong to a continuum that cannot be exhaustively listed. Increasing simulation counts enhances precision within a defined frame, but does not eliminate structural incompleteness. Between any two scenarios in a model, infinitely many remain unmodeled.
The objective of modeling is not to conquer infinity but to navigate within it responsibly. Combinatorial explosion taught us that perfect optimization is computationally infeasible. Continuous explosion teaches us that perfect representation is logically impossible. Approximation is not a concession; it is the only rational strategy available in systems defined by uncountable possibility. Once that perspective is internalized, claims of comprehensiveness become more cautious, and confidence becomes more disciplined. We allocate capital with buffers, stress-test assumptions rigorously, and treat outputs as guides rather than guarantees.
Cantor revealed that infinity has layers. Financial history reminds us that risk has layers as well. Recognizing this does not weaken modeling; it deepens it. The world is defined by a continuum that cannot be fully enumerated, so disciplined compression coupled with structural humility becomes the foundation of sound judgment.
There are more numbers between 0 and 1 than there are whole numbers in existence. This may feel counterintuitive, but it’s a fact. The whole numbers stretch without end: 0, 1, 2, 3, 4, continuing forever. And the interval between 0 and 1 feels narrow, confined, almost trivial by comparison. Yet in 1874, Georg Cantor proved that the infinity contained within that small interval is strictly larger than the infinity of all whole numbers combined, and in doing so he forced mathematics to confront the fact that infinity is not a single undifferentiated idea but a hierarchy with structure. Some infinities are larger than others, and one infinity can fit inside another.
The infinity of whole numbers is what mathematicians call countable. That term does not imply finiteness; it means that the elements of the set can be placed into a sequence where each element corresponds to a natural number. Even though the list never ends, the pairing exists. Cantor’s breakthrough came when he examined the real numbers between 0 and 1, which include every possible decimal expansion, terminating decimals, repeating decimals, and those that continue without pattern forever. Through what is now known as the diagonal argument, he demonstrated that no matter how long your list is, no matter how many decimal expansions you write down, you can always construct a new number by changing the nth digit of the nth entry. That new number will differ from every number in the list at least at one decimal place. The consequence is not merely that the list is incomplete, but that completeness is impossible. The set is uncountable. Its cardinality exceeds that of the natural numbers.
The result becomes even more unsettling when one recognizes that the number of real numbers between 0 and 0.1 is equal to the number between 0 and 1, equal to the number between 0 and 100, and equal to the entire continuum of real numbers altogether. When mathematicians compare sizes of infinite sets, they don’t use length, they use the idea of a one-to-one correspondence (a bijection). If you can pair every element of set A with exactly one element of set B, with nothing left over on either side, the two sets have the same cardinality. For example, take every real number x in the interval (0,0.1), multiply it by 10, and you get a real number 10x in the interval (0,1). Multiply it by 1000, and you get a real number 1000x in the interval (0,100). Every number in (0,0.1) pairs with exactly one number in (0,1) and (0,100), and vice versa.
So shrinking the interval in the real numbers realm does not shrink the infinity. Between any two distinct real numbers, there are infinitely many more. Density persists at every scale. Cantor did not merely expand mathematics; he destabilized our instinctive understanding of magnitude. Large ceased to mean what it seemed to mean. Infinity was no longer a horizon, but a layered landscape.
In last week’s column, I examined the cost of perfect optimization through the Traveling Salesman Problem. A route with 30 stops yields 30 factorial possible permutations, approximately 2.65 × 1032 distinct routes. To evaluate each one at a trillion computations per second would require more than 600 times the age of the universe, which itself is approximately 13.8 billion years. The lesson was not that optimization is futile but that brute-force perfection becomes computationally infeasible long before problems appear intuitively large. Discrete combinatorics produces explosion at a pace that overwhelms hardware.
That example dealt with finite but astronomically large sets. Cantor’s insight exposes something more profound: certain infinities are not just large, they are unenumerable. Financial markets operate inside continuous state spaces, which fall into the unenumerable infinity space. Asset prices, interest rates, exchange rates, volatility surfaces, and credit spreads are represented mathematically as real-valued processes evolving through time. In the framework of modern quantitative finance, these processes are modeled using stochastic calculus, where the space of possible trajectories is not simply vast but uncountable.
A Monte Carlo simulation with 10,000 paths may feel thorough. Increasing it to 100,000 improves convergence. Pushing it to one million refines estimates further. Yet each simulation remains a countable subset of an uncountable universe. Between any two simulated price paths, there exist infinitely many trajectories that were not generated. This means that we are just able to model a thin slice of possible trajectories, and that gap is not technological. Even with infinite computing power, one cannot list an uncountable set. Cantor’s distinction is structural, not computational. Optimization problems confront limits of speed, but continuous systems confront limits of logic.
Financial models do not attempt to enumerate every possible future. They compress an infinite continuum into tractable representations. Measure theory provides the mathematical foundation for this compression by allowing us to define probability measures over continuous spaces and compute expectations without listing individual outcomes. Instead of cataloging every trajectory of an asset price, we integrate across a distribution and extract summary statistics: expected returns, variances, tail probabilities, scenario outcomes. The mathematics is rigorous and internally coherent. The abstraction is powerful.
Compression, however, is not completeness. Increasing the number of simulations reduces sampling error but does not alter the underlying structure of the possibility space. Between any two simulated scenarios lie infinitely many that remain unobserved. No expansion in computational resources closes that structural gap. A model remains an approximation because the territory it describes is richer than any representation we construct.
The necessity of compression becomes obvious through a simple analogy. Imagine building a map that records every grain of sand, every pebble, every microscopic contour of a landscape. Such a map would replicate the territory with perfect fidelity, but its very completeness would render it useless. If your map included every grain of sand, you could not fold it, carry it, or deploy it for navigation. A model that attempted to incorporate every conceivable detail would collapse under its own weight. Therefore abstraction is not a defect, it is the condition of usability.
The discipline lies in understanding what the compression omits. It is tempting to say that models fail to incorporate events such as 9/11, the subprime mortgage collapse, or COVID because those events were unimaginable, but that explanation is too simple. Pandemics were known, systemic banking crises were documented repeatedly across centuries and correlation breakdowns under stress were observable in emerging market crises long before 2008. The issue was rarely ignorance of possibility; it was calibration and structure. Most financial models are anchored to historical data drawn from relatively stable regimes. Correlations are estimated from periods in which liquidity is abundant, volatility is contained, and institutions behave predictably. Tail events, however, are often regime-shifting phenomena. They alter correlations, liquidity dynamics, and behavioral feedback loops simultaneously. When the regime changes, the distribution itself changes. The tails were there, it’s just that they were thin because the compression assumed continuity of structure. What was mischaracterized was not mathematics, but the persistence of the regime.
There is also an institutional dimension that is harder to discuss but impossible to ignore. Assigning meaningful probability mass to severe downside scenarios has immediate economic consequences: expected returns decline, capital requirements increase, leverage becomes less justifiable, growth projections tighten. An ultra-conservative model may be mathematically defensible, but it produces numbers that appear less attractive in board presentations and investor decks. Institutions compete on performance, not on prudence that never gets tested. The pressure to present robust earnings and efficient capital allocation subtly incentivizes thinner tails and optimistic compression. This is not necessarily the result of recklessness; it is the natural byproduct of incentives interacting with uncertainty. The danger, therefore, is not that models fail to include alien invasions. It is that plausible discontinuities, events within the structural domain of the system, are assigned negligible weight because incorporating them seriously would materially alter today’s reported outcomes.
Recognizing that we operate within an uncountable continuum does not imply that every imaginable event must be modeled. An extraterrestrial invasion is logically possible but economically irrelevant within the defined scope of most macro models. Its probability mass is, for all practical purposes, negligible. The challenge is not to incorporate every conceivable scenario but to define the boundaries of the model with clarity and to assign appropriate weight to regions of the outcome space that carry structural significance. The real danger lies in low-probability, high-impact events that reside within the plausible domain of the system and whose probability is small but not vanishing.
Cantor’s hierarchy of infinities offers more than a mathematical curiosity; it provides a lens for thinking about uncertainty. Financial systems resemble the uncountable more than the countable. The possible paths of an economy, a balance sheet, or a market through time belong to a continuum that cannot be exhaustively listed. Increasing simulation counts enhances precision within a defined frame, but does not eliminate structural incompleteness. Between any two scenarios in a model, infinitely many remain unmodeled.
The objective of modeling is not to conquer infinity but to navigate within it responsibly. Combinatorial explosion taught us that perfect optimization is computationally infeasible. Continuous explosion teaches us that perfect representation is logically impossible. Approximation is not a concession; it is the only rational strategy available in systems defined by uncountable possibility. Once that perspective is internalized, claims of comprehensiveness become more cautious, and confidence becomes more disciplined. We allocate capital with buffers, stress-test assumptions rigorously, and treat outputs as guides rather than guarantees.
Cantor revealed that infinity has layers. Financial history reminds us that risk has layers as well. Recognizing this does not weaken modeling; it deepens it. The world is defined by a continuum that cannot be fully enumerated, so disciplined compression coupled with structural humility becomes the foundation of sound judgment.
— Carlos E. Mora
I wake up, I build, I repeat. No guarantees.
I work like it’s all on the line, because it is.
Family is the only true legacy.
Your name is your currency, and it must be earned daily.
The arithmetic in these essays is the arithmetic the practice runs on a mandate.
Discuss a mandate →