EN

Automated trading

Timeline: from Bachelier to AI agents

More than a century of quantitative and AI trading in 14 milestones. Filter by theme or open a milestone for the explanation and sources.

  1. 1900The birth of mathematical financeTheory

    French mathematician Louis Bachelier defends his thesis Théorie de la spéculation. He is the first to use Brownian motion to describe price movements and option prices on the Paris Bourse mathematically, decades before computers existed.

    Sources

  2. 1952Modern portfolio theoryTheory

    Economist Harry Markowitz publishes Portfolio Selection. For the first time, risk and return are defined mathematically, through the correlation between assets, rather than by gut instinct. In 1990 he receives the Nobel Prize in Economics for this work.

    Sources

  3. 1969The first quantitative hedge fundFunds

    Mathematician Edward O. Thorp, known for his card-counting method for blackjack, launches Princeton Newport Partners. Using computers and statistical arbitrage on convertible bonds and warrants, he shows that a systematic, mathematical approach can work in the market.

    Sources

  4. 1971–1973NASDAQ and the Black-Scholes formulaTheoryMarket structure & rules

    In 1971, NASDAQ launches as the first electronic stock market. In 1973, Fischer Black, Myron Scholes and Robert Merton publish their option-pricing model. Banks start hiring physicists and mathematicians in large numbers: the first “quants”.

    Sources

  5. 1976Electronic orders on the NYSEMarket structure & rules

    The New York Stock Exchange introduces the DOT system (Designated Order Turnaround), succeeded by SuperDOT in 1984. Brokers can now route orders electronically instead of through floor runners. This makes “program trading” possible: automated rules that buy or sell whole baskets of stocks at once.

    Sources

  6. 1982–1987Volatility models, pairs trading and Black MondayTheoryCrashes

    Robert Engle introduces the ARCH model in 1982 to forecast changing volatility, and Nunzio Tartaglia’s team at Morgan Stanley develops automated “pairs trading”. On 19 October 1987, Black Monday, automated “portfolio insurance” strategies amplify the sell-off: the Dow Jones loses 22.6% in a single day. It is the first major shock in which algorithms play a leading role.

    Sources

  7. 1988Renaissance Technologies and the Medallion FundFunds

    Mathematician and former codebreaker Jim Simons launches the Medallion Fund at Renaissance Technologies. Instead of traditional analysts, he hires astrophysicists, cryptographers and speech-recognition specialists who apply statistical pattern recognition to the market. The fund becomes one of the most successful investment funds ever.

    Sources

  8. 1998–2005Electronic trading venues and the rise of HFTMarket structure & rules

    In 1998, the US regulator SEC adopts Regulation ATS, allowing electronic trading networks to compete with traditional exchanges. In 2001, US markets move from fractions to decimal prices in $0.01 steps, compressing spreads. With Regulation NMS (2005), electronic execution at the best price becomes the norm. It is the starting signal for high-frequency trading, where firms compete in microseconds.

    Sources

  9. 2001Software agents beat human tradersAI & data

    IBM researchers show in controlled experiments that adaptive software agents outperform human traders in an auction market. The idea of the “trading machine” gains a scientific footing.

    Sources

  10. 6 May 2010The Flash CrashCrashesMarket structure & rules

    A large automated sell order in S&P 500 futures sets off a chain reaction among high-frequency algorithms. In around 36 minutes, nearly a trillion dollars in market value disappears before largely recovering. Regulators then introduce stricter circuit breakers and controls on algorithms.

    Sources

  11. 2011–2015Cloud backtesting and open sourceAI & data

    Platforms such as Quantopian and QuantConnect give programmers browser-based access to historical market data and testing environments. Together with open-source machine learning libraries such as scikit-learn and Google’s TensorFlow (2015), individuals can now build their own models too.

    Sources

  12. 2017–2020Transformers and alternative dataAI & data

    In 2017, Google researchers introduce the Transformer architecture, which transforms how language is processed. Funds use such models to analyse “alternative data”, from satellite images of car parks and shipping movements to earnings calls and social media. JPMorgan deploys LOXM in 2017, a self-learning system for executing large orders with minimal market impact, and in 2019 FinBERT appears, a language model built specifically for financial texts.

    Sources

  13. 2023–2024Generative AI and financial language modelsAI & data

    After the launch of ChatGPT, Bloomberg unveils BloombergGPT, a 50-billion-parameter language model trained on financial data. Trading platforms integrate AI assistants, so that even people who cannot code can describe a strategy in plain language and have it tested.

    Sources

  14. 2025–2026Agentic AIAI & data

    AI trading evolves from predictive models to collaborating AI agents: one agent analyses filings and macroeconomic news, another looks for technical patterns, a third monitors portfolio risk. Techniques once reserved for large institutions are moving ever closer to individual investors. Regulators are closely monitoring the implications for financial stability.

    Sources

The sources link to external websites. Vermeeren & Van Haute is not responsible for their content. This timeline contains general information and is not investment advice.

Back to the article on automated trading