SPATIAL-COGNITIVE ATLAS · VOYAGE SIX

Cities That Calculated Chance — From Interrupted Games to Probability Axioms

Probability was not completed one day by counting coin tosses. Stakes in an interrupted game, a fair price for a gamble, urban death records, annuities, errors in observing stars, and state statistics changed what it meant to handle the unknown with numbers in different places.

QUESTION FOR THE ROUTE

Why did the same language of chance become a fair share at the gaming table, survival risk in insurance, observational error at an observatory, and an average for the state?

WHAT THE LINE DOES NOT CLAIM

The line is neither proof that one probability theory travelled unchanged from Paris to Moscow nor a ranking of cities or civilizations. It is the viewer’s edited itinerary across letters, print, mortality bills, societies, observatories, state statistics, and universities that made different problems repeatable; every pin states its own location basis and limit.

The camera rests on each city while you read, then eases through the runway between scenes. Select any marker or scene link to travel in either direction.

The same route, four questions

A lens never hides a scene or proves a cause. It changes which places you compare first, and the URL preserves your choice.

Whole route Translation networks Commercial demand Observation and institutions Patronage and power

Patronage and power · READING QUESTION

Who made time and access possible, and whom did those arrangements support or constrain?

Causes of death defined by church and city, Royal Society publication, population categories made by states, and access through universities and publishers shaped which data and people became visible. Greater precision carries responsibility to examine classification and resource allocation rather than becoming a view outside power.

Start with:London Wrocław London Brussels

Patrons are not lone creators, and patronage is not romanticized apart from power, war, or inequality.

Live map · 지도를 불러오는 중…

01 / 11 · 1654 CE

Paris

01 · 1654 CE
Paris · Letter sent
The Game Stops but the Money Remains — Pascal Counts Possible Futures
When two players stop before either reaches the target number of wins, dividing the stake by the current score alone is not fair. Pascal worked backward through possible future wins and the value of each state to calculate each player’s expected share, connecting the problem with combinations in his arithmetical triangle. The question posed by de Méré and earlier discussions by Pacioli, Cardano, and Tartaglia prevent this from becoming a story in which Pascal alone completed modern probability from nothing.
PAUSE AND ASK
If a first-to-five game stops at 3–2, should the stake be divided by wins already earned or by wins and losses that could still occur?
How the idea changed
Working backward through all possible finishes and their values rather than dividing by the present score made fairness calculable from the structure of future cases. Chance became not a prediction of the result but a relation for valuing a share now.
What this place made possible
Parisian gaming culture, de Méré’s practical question, Pascal’s work on the arithmetical triangle, and a mathematical correspondence network including Carcavi made one problem calculable and debatable. Neither salon nor gaming table generated the method automatically.
How it moved
The interrupted-game stake and earlier arithmetic discussions → Pascal’s recursive division and combinatorial calculation → letters moving from Paris to Fermat in Toulouse in 1654 → comparison of different solutions
Do not overclaim
The Pascal–Fermat letters are not probability’s sole birth certificate. Pacioli, Cardano, and Tartaglia had treated earlier versions, contemporaries including de Méré and Roberval participated, and the two correspondents did not formalize all of modern probability.
Evidence sources
- Bibliothèque nationale de France — Pascal and Fermat correspondence
  Supports: The four published letters of 1654, de Méré’s interrupted-game question, the correspondents’ different calculations, and the history of an expected share
- MAA Convergence — The Problem of Points and Perseverance
  Supports: The historical and pedagogical context of the 1654 problem-of-points exchange and the non-instantaneous development of a solution
Stable link to this scene
ParisToulouse
02 · 1654 CE
Toulouse · Letter sent
A Calculation Goes Paris to Toulouse while Another Returns
Fermat calculated the division by directly enumerating equally possible ways the game could finish. In their short correspondence during the summer of 1654, Pascal’s recursive reasoning and Fermat’s combinatorial reasoning illuminated the same answer; one letter explicitly describes a proposition travelling from Paris to Toulouse while another went the opposite way. Roughly five surviving letters form a powerful starting point, not a textbook already containing independence, conditional probability, and continuous distributions.
PAUSE AND ASK
Can the same fair share be found by directly counting possible finishes rather than reasoning backward recursively?
How the idea changed
Fermat extended the needed play to an imagined finish and enumerated equally possible outcomes. As different methods met by post and supported the same number, confidence could arise from agreement among independent representations rather than one person’s intuition.
What this place made possible
Fermat combined judicial office in Toulouse with mathematics and connected chiefly by letter to Parisian scholars. Distance from a central institution brought isolation, while correspondence created a research space for delayed reflection and remote checking.
How it moved
Pascal’s question and calculation moved Paris → Toulouse ↔ Fermat’s enumeration and figurate-number proposition moved Toulouse → Paris → preservation through intermediaries including Carcavi and later editions of the letters
Do not overclaim
The line marks an exchange attested in the letters, not the exact itinerary of every item or one-way transmission. “Equally possible cases” still require justification about which outcomes may be treated alike and do not transfer automatically to every real risk.
Evidence sources
- Fermat to Pascal, 29 August 1654 — critical-edition transcription
  Supports: The place, date, and wording of the 1654 letter that explicitly describes propositions travelling Paris to Toulouse and Toulouse to Paris
- MacTutor — Pierre de Fermat biography
  Supports: Fermat’s activity and office in Toulouse, renewed correspondence with Paris mathematicians in 1654, and the scope of his probability exchange with Pascal
Stable link to this scene
ToulouseLeiden
03 · 1657 CE
Leiden · Publication
What Is This Game Worth Now? — Huygens Prints Expectation
Huygens’s De ratiociniis in ludo aleae gave propositions for the fair present value of games with several possible outcomes. Printed in Latin at Leiden as an appendix to van Schooten’s mathematical exercises, problems that had circulated in letters became public exercises other readers could solve and criticize. Calling it the first printed probability treatise does not erase earlier manuscripts and problem traditions or claim that the modern formalism of random variables and expectation was complete.
PAUSE AND ASK
If a game whose outcome is unresolved is transferred to someone else now, what is its fair price?
How the idea changed
Combining each payoff with its equal chances gave the game itself a present value. Expectation began not as a number that predicts the future, but as a fair exchange value acceptable when players swap roles under the same conditions.
What this place made possible
Mathematical teaching at Leiden, van Schooten’s network around Cartesian geometry, and the Elzevir press turned Huygens’s Dutch manuscript into a Latin appendix reusable by scholarly readers. The place of composition is not collapsed into the publication city.
How it moved
Huygens encounters Pascal–Fermat problems in Paris → Dutch manuscript → Latin translation and editing by van Schooten → appendix to the 1657 Leiden Exercitationes mathematicae → later problem networks including Bernoulli and de Moivre
Do not overclaim
The label “first printed probability treatise” does not erase Cardano’s earlier unpublished manuscript or other gaming calculations. Huygens’s value or expectation is not equated with every modern random-variable, probability-measure, or risk-neutral-pricing concept.
Evidence sources
- ETH-Bibliothek e-rara — Schooten and Huygens, Leiden 1657
  Supports: The 1657 Leiden Elzevir edition, Huygens’s probability treatise appended to van Schooten, and the digitized primary copy
- Leiden University — Huygens, facts and figures
  Supports: The Dutch text and Latin translation, 1657 publication date and expectation propositions, and bibliographic relation to the Dutch edition of 1660
Stable link to this scene
LeidenLondon
04 · 1662 CE
London · Publication
From One Person’s Death to a City’s Frequency — Graunt Rereads the Bills
Patronage and power · lens spotlight
London’s weekly Bills of Mortality were administrative notices of burials and epidemic deaths. John Graunt assembled years of births, deaths, and cause classifications to compare recurring patterns in sex ratios, disease, and survival and to construct an early life table grounded partly in real data. Records with omissions, misclassification, and little age information were neither a complete census nor a law predicting every individual life.
PAUSE AND ASK
Although different people die each week, do recurring urban patterns appear when many years of mortality bills are layered together?
How the idea changed
Counting births, burials, and disease classifications to compare proportions and stability replaced predicting an individual death with describing group-level regularity. Probabilistic reasoning moved from symmetric gaming cases to imperfect but observed population records.
What this place made possible
Weekly parish mortality bills, plague surveillance and administration, commercial print, and the emerging Royal Society made years of records available for rereading. The recording system created data and also created categories, omissions, and errors.
How it moved
Parish burial and christening records → London’s weekly Bills of Mortality → Graunt reclassifies and compares years of tables → 1662 Natural and Political Observations → expansion into population, life-table, and public-health calculation
Do not overclaim
Cause labels in the bills are not modern diagnoses and the records contain missing ages, places, nonconformists, and recorder judgment. Graunt’s tables are not presented as a complete census, precise individual prediction, or the lone invention of modern statistics.
Evidence sources
- The Royal Society Catalogue — John Graunt
  Supports: The 1662 publication, Graunt’s London civic and Royal Society associations, and the record of an early life table based partly on real mortality data
Stable link to this scene
LondonLyon
05 · 1663 CE
Lyon · Publication
A Manuscript Written Earlier Arrives Later — Cardano’s Gambling Book
Cardano had written or revised De ludo aleae, counting dice totals and favourable cases, by about 1564 but did not publish it in his lifetime. Its inclusion nearly a century later in the Lyon Opera omnia makes the difference between doing a calculation and making it a public text unusually visible. Late publication does not mean Cardano worked in Lyon in 1663 or prove that the manuscript directly influenced Pascal and Fermat.
PAUSE AND ASK
If a calculation was written earlier but printed later, does its history begin when it was thought or when other readers could encounter it?
How the idea changed
Cardano’s manuscript compares favourable and total cases and explores multiplication across independent trials while visibly correcting errors. Posthumous print separates discovery, composition, publication, and influence into different times, folding a simple linear chronology of probability.
What this place made possible
Lyon’s international book trade and Charles Spon’s editorial work assembled Cardano’s manuscripts into the ten-volume Opera omnia. The city is not where the calculations originated, but where a private manuscript became a public edition.
How it moved
Cardano’s gambling experience and notes from the 1520s onward → unpublished manuscript revised by about 1564 → Spon gathers autograph materials after Cardano’s death → printing in volume one of the 1663 Lyon Opera omnia → later reassessment in histories of probability
Do not overclaim
The Lyon pin and 1663 date are publication evidence, not the place or date of composition. No proven route carries the manuscript to Pascal and Fermat, and Cardano’s calculations did not complete modern axioms, independence, or conditional probability.
Evidence sources
- University of Milan Cardano Project — Opera omnia
  Supports: The 1663 Lyon printing, editors and publishers, and the searchable bibliographic record for De ludo aleae in volume one
- MacTutor — Girolamo Cardano biography
  Supports: Cardano’s gambling experience, probability content, and the boundary between probable completion around 1563 and posthumous publication in 1663
Stable link to this scene
LyonWrocław
06 · 1693 CE
Wrocław · Main activity
Pricing an Annuity from a City’s Death Records — The Breslau Life Table
Patronage and power · lens spotlight
Edmond Halley used birth and death data collected in Breslau by Caspar Neumann to estimate survivors by age and propose a method for pricing life annuities. Linking an unpredictable individual lifetime with group survival frequencies and present value made probability a tool of insurance and finance. The Wrocław pin marks where the records were produced, while the paper was published by the Royal Society in London; a few years from one city are not a universal survival law for every period and class.
PAUSE AND ASK
Can a few years of births and deaths in one city be used to estimate a thirty-year-old’s survival and a fair annuity price?
How the idea changed
Turning deaths by age into a table of a hypothetical cohort declining over time made an unknowable individual lifetime tractable as group survival probability. Adding present values of future payments made probability a tool for comparing life risk and long contracts.
What this place made possible
Halley regarded Breslau as relatively stable and less affected by migration, while pastor Caspar Neumann systematically collected monthly births and deaths and sent them to the Royal Society. Those judgments and categories are both strengths and limits of the sample.
How it moved
Breslau church and civic birth and burial records → Neumann’s monthly tables → data sent to the Royal Society in London → Halley’s age-specific life table and annuity calculation → later actuarial and demographic use
Do not overclaim
The Wrocław pin marks where the data were produced; Halley calculated and published in London. A few years from one city have class, migration, and recording limits, and a life table neither predicts an individual death date nor transfers unchanged to every population.
Evidence sources
- The Royal Society — Halley’s paper on the Breslau Bills of Mortality
  Supports: The 1693 manuscript and Royal Society record, Breslau birth and funeral tables, and evidence connecting a life table to annuity prices
Stable link to this scene
WrocławBasel
07 · 1713 CE
Basel · Publication
How Much Repetition Is Enough to Trust? — Bernoulli’s Law of Large Numbers
Jakob Bernoulli’s posthumous Ars Conjectandi extended Huygens’s games and combinatorics and proved that, under repeated trials with the same conditions, observed frequency could approach an unknown probability with chosen accuracy and reliability. It did not remove chance; it related sample size, tolerated error, and confidence. “Large numbers” do not promise certainty after merely trying many times and require assumptions such as independence and stable conditions.
PAUSE AND ASK
When the true probability is unknown, how many trials are needed before an observed frequency deserves a chosen degree of trust?
How the idea changed
With enough repetitions, the chance that frequency lies within a chosen error of probability can be raised to a desired level. Repetition became a quantitative relation among sample size, error, and reliability, opening a path from observations back to probability.
What this place made possible
The University of Basel, Bernoulli family teaching, rivalry, and letters, and Nicolaus Bernoulli’s posthumous editing turned Jakob’s manuscript into a book. The family network preserved knowledge while carrying competition over credit, access, and position.
How it moved
Huygens’s games and combinatorics → Jakob Bernoulli’s manuscript and proof of a law of large numbers → posthumous editing by his nephew Nicolaus → Basel publication of Ars Conjectandi in 1713 → later frequency and statistical inference
Do not overclaim
A law of large numbers is neither a promise of exact average after finitely many trials nor a way to recover gambling losses. It requires assumptions such as stable probability and independence and did not complete every modern theorem or inference.
Evidence sources
- MacTutor — Jacob Bernoulli biography
  Supports: Bernoulli’s Basel activity, posthumous publication of Ars Conjectandi in 1713, and the scope of his combinatorial, probability, and large-number work
Stable link to this scene
BaselLondon
08 · 1763 CE
London · Presentation
Seeing Results and Asking Backward about Causes — Bayes and Price
Patronage and power · lens spotlight
A manuscript left by Thomas Bayes treated the inverse problem of using observed successes and failures to find how likely an unknown success probability was to lie within an interval. Richard Price prepared it, communicated it to the Royal Society in 1763, and added an interpretation about inferring stable causes in nature. Bayes did not single-handedly complete today’s theorem, priors, and computational statistics; Laplace and later interpreters greatly extended the framework.
PAUSE AND ASK
Given only repeated outcomes from a coin with unknown success probability, can we reason backward about the range in which that probability may lie?
How the idea changed
The direction from known cause to possible result was reversed to derive a distribution over an unknown probability from observed successes and failures. Evidence updates possibilities, while the choice of a prior assumption does not disappear outside the calculation.
What this place made possible
Bayes’s private manuscript reached readers through Price’s editing and interpretation, a Royal Society reading, and journal print. The London society made a posthumous manuscript public while acting as a gate over who became remembered.
How it moved
Bayes’s posthumous manuscript → discovery, editing, and additions by Richard Price → Royal Society reading in London in December 1763 → Philosophical Transactions publication → Laplace’s generalization and later Bayesian interpretations
Do not overclaim
The essay is neither the full textbook theorem for every case nor a completed Bayesian philosophy. Price and Laplace remain visible, and a uniform prior is not made a universally neutral rule whenever information is absent.
Evidence sources
- Royal Society Philosophical Transactions — Bayes and Price (1763)
  Supports: The title and 1763 journal record of Bayes’s posthumous essay, Price’s role as communicator, and the inverse-probability problem
- MacTutor — Earliest uses: Bayes and inverse probability
  Supports: The relation of Bayes’s problem to a beta posterior and the boundary between the name “Bayes theorem” and Laplace’s generalization
Stable link to this scene
LondonGöttingen
09 · 1809 CE
Göttingen · Composition
From Conflicting Observations to the Most Plausible Orbit — Gauss Calculates Error
At Göttingen, Gauss’s Theoria motus explained least-squares calculation and an error curve in the problem of combining astronomical observations to estimate an orbit. Disagreement among measurements became not discarded failure but a search for the value that best explains the full dataset under an error model. Legendre published least squares first in 1805, and de Moivre and Laplace preceded Gauss on the normal curve, so neither method was Gauss’s lone invention.
PAUSE AND ASK
When repeated observations of one body all differ, how can the most plausible orbit be chosen without simply discarding measurements?
How the idea changed
Choosing parameters that minimize the sum of squared differences between observations and model predictions adjusts the full dataset together. Under a particular error assumption, its connection to a normal curve made an “exact value” not one errorless reading but an estimate supported by data, model, and assumptions.
What this place made possible
Göttingen’s observatory and university supplied instruments, books, students, and correspondence for computing observations made across Europe and comparing orbital and geodetic problems over time. The pin marks the research center named in Gauss’s 1809 preface, not the production site of every observation.
How it moved
Error curves of de Moivre and Laplace and Legendre’s 1805 publication of least squares + astronomical observations including Piazzi’s → Gauss’s computation and justification at Göttingen → 1809 Theoria motus → spread through astronomy, geodesy, and statistical error theory
Do not overclaim
Gauss was not the first to publish least squares; Legendre did so in 1805. The normal curve also has earlier work by de Moivre and Laplace, and real errors do not universally have to be independent, symmetric, and normally distributed.
Evidence sources
- MacTutor — Gauss, Theoria motus preface
  Supports: The Göttingen-dated preface to the 1809 Theoria motus and the composition context for estimating orbits from actual observations
- MacTutor — Earliest symbols in probability and statistics
  Supports: Gauss’s 1809 least-squares and error-curve work, predecessors in de Moivre and Laplace, and the scope of error theory in astronomy and geodesy
Stable link to this scene
GöttingenBrussels
10 · 1835 CE
Brussels · Composition
Placing an Error Curve over Society — Quetelet’s Average Man
Patronage and power · lens spotlight
Working through the Brussels observatory, state statistics, and international scholarly networks, Adolphe Quetelet compared means and distributions of height, weight, marriage, crime, and other group records and made the “average man” central to social physics. A curve used for random observational error moved into the description of recurring social patterns. A mean is neither a real ideal person nor a moral norm, and the state institutions defining categories and records also possess powers of bias, stigma, and control.
PAUSE AND ASK
When records of height, marriage, or crime across different people are gathered into averages and curves, does society become clearer or do real differences disappear?
How the idea changed
Applying a distribution used for random observational error to recurring group patterns turned a mean from a calculation into a reference for social comparison. It also created a danger: reifying the “average man” as cause, ideal, or norm can hide variation and the institutions that produced the data.
What this place made possible
The Brussels Observatory, Belgian state statistics, the Royal Academy, and international statistical congresses joined astronomy, administrative records, and scholarly classification. A new state’s demand to know its population expanded data and the power to define measurement categories.
How it moved
Astronomical error and probability curves → French, British, and Belgian population and moral statistics → Quetelet’s comparison and composition in Brussels → 1835 Paris printing of Sur l’homme → spread through social science, anthropometry, and policy statistics
Do not overclaim
The 1835 pin rests on the Brussels-dated preface and activity; the first edition was printed in Paris. A group mean proves neither a real “normal person” nor a desirable body or conduct, and it does not erase the power embedded in state and scholarly categories of crime, race, or sex.
Evidence sources
- Académie royale de Belgique — Quetelet, Physique sociale edition
  Supports: The Brussels-dated 1835 preface, Quetelet’s distinction between empirical books and the theory of the average man, and the bibliography of the Paris first edition
- MacTutor — Adolphe Quetelet biography
  Supports: Quetelet’s Brussels observatory and statistical activity and the scope of applying probability to social and vital statistics
Stable link to this scene
BrusselsMoscow
11 · 1933 CE
Moscow · Composition
A Grammar of Sets and Measures for Chance — Kolmogorov’s Axioms
At Moscow University, Andrey Kolmogorov used measure theory to organize sample spaces, events, non-negativity, total probability one, and countable additivity into an axiomatic system. Publication of the German monograph by Springer in Berlin in 1933 made a shared rigorous language available across different probability problems. The axioms did not settle every philosophical interpretation of probability or erase earlier axiomatization and measure theory.
PAUSE AND ASK
Can gaming cases, life-table frequencies, and continuous errors become objects of one mathematical system rather than separate tricks?
How the idea changed
Treating events as subsets of a sample space and probability as a measure of total value one placed finite, continuous, and infinite settings under common axioms. Probability gained a relational grammar stating which operations are permitted rather than relying only on intuitive possibility.
What this place made possible
Moscow University seminars and networks in function theory, measure, and probability around Kolmogorov, Khinchin, and Alexandrov provided a setting for unifying results across fields. German-language Springer publishing connected the Moscow work to international readers.
How it moved
Measure theory of Borel and Lebesgue, Hilbert’s sixth problem, and several axiomatization attempts → Kolmogorov’s synthesis in Moscow → 1933 German monograph published by Springer Berlin → a common foundation for modern probability and stochastic processes
Do not overclaim
Kolmogorov’s axioms govern calculation but do not settle whether probability is objective frequency, degree of belief, or another interpretation. Earlier measure theory and attempts by Hilbert, Bernstein, von Mises, and others remain visible, and Berlin publication is not confused with Moscow research.
Evidence sources
- MacTutor — Andrey Kolmogorov biography
  Supports: Kolmogorov’s 1931 Moscow University professorship and the scope of the measure-theoretic axiomatization in his 1933 probability monograph
- London Mathematical Society — Kolmogorov obituary and bibliography
  Supports: Springer Berlin publication of the 1933 Grundbegriffe der Wahrscheinlichkeitsrechnung and the chronology of Kolmogorov’s work
Stable link to this scene

TOUCH THE MATHEMATICS

Cities That Calculated Chance — From Interrupted Games to Probability Axioms

Eleven scenes connect letters about an interrupted game between Paris and Toulouse, fair expectation in Leiden, mortality tables in London, Cardano’s much older gambling manuscript printed late in Lyon, a life table from Wrocław, the law of large numbers in Basel, inverse probability in London, observational error at Göttingen, the social average in Brussels, and axioms in Moscow. Probability appears not as one formula invented at one moment, but as changing answers to what it means to calculate what is not known.

Continue with the 88-second cinematic journey

OPEN THE FULL MAP

Follow the river of probability on the world map

Leave the edited eleven scenes and compare other people, problems, hubs, and routes in probability and statistics on the live map.

Explore the full map

11 scroll-controlled map scenes

The Game Stops but the Money Remains — Pascal Counts Possible Futures

A Calculation Goes Paris to Toulouse while Another Returns

What Is This Game Worth Now? — Huygens Prints Expectation

From One Person’s Death to a City’s Frequency — Graunt Rereads the Bills

A Manuscript Written Earlier Arrives Later — Cardano’s Gambling Book

Pricing an Annuity from a City’s Death Records — The Breslau Life Table

How Much Repetition Is Enough to Trust? — Bernoulli’s Law of Large Numbers

Seeing Results and Asking Backward about Causes — Bayes and Price

From Conflicting Observations to the Most Plausible Orbit — Gauss Calculates Error

Placing an Error Curve over Society — Quetelet’s Average Man

A Grammar of Sets and Measures for Chance — Kolmogorov’s Axioms

Cities That Calculated Chance — From Interrupted Games to Probability Axioms

Follow the river of probability on the world map