Simple Explanation (ELI5)

Think of it like buying lots of groceries from different stores to avoid raising prices: big funds do the same with stocks, splitting orders. Scientists tried to verify this theory using only public data, like spotting a thief's shadow without seeing their face.

Key Advantages & Current Limits

🟢 Pros: Using public data makes research transparent and replicable by anyone, without secret keys. 🟢 Pros: Paves the way for open, accessible quantitative finance for more researchers. 🔴 Cons: Anonymous data cannot independently confirm LMF theory, only show compatibility. 🔴 Cons: The trade-off between model accuracy and theory fidelity reveals intrinsic limits of synthetic reconstruction.

Future Outlook & Takeaways

In the next 3-5 years we'll see more open-source tools to analyze markets with public data, making finance more transparent. However, the lack of detailed data will push the use of advanced techniques like machine learning to improve reconstruction of market behaviors.

How Public Data Reveals Hidden Large Orders?

Article

100%

Imagine wanting to buy a mountain of shares without pushing the price up. What do you do? Split it into many small orders, like grocery shopping in instalments to avoid emptying your wallet all at once. This strategy, called a "metaorder", is daily bread for large investment funds. For years scientists have suspected that this very splitting creates those mysterious long-range correlations seen in markets: the LMF (Lillo-Mike-Farmer) theory.

Until now, testing it required proprietary data with trader identities. But a research team tried to replay the game using only public data from the Johannesburg Stock Exchange (JSE): 239 stocks, three years of trades (2023-2025). They built a metaorder "search engine": crossing transactions and quotes, trying thousands of configurations to see which reproduce known market behaviours.

Surprising result: if they tune the model to mimic average metaorder impact, the LMF theory fails. If they force it to satisfy LMF, average impact loosens. This means anonymous data alone cannot independently confirm the theory: they only show compatibility, not definitive proof. It's like recognising a thief by their shadow: the shadow is there, but you don't see the face.

The study, however, opens a path: with open data and smart algorithms we can probe deep market mechanisms without secret keys. A step forward for transparent, replicable quantitative finance.

Bonus

100%

The study uses transaction and quote (TAQ) data for the 239 largest-cap JSE stocks as of 13 March 2026, covering 1 January 2023 – 31 December 2025. Synthetic metaorder reconstruction employs a clustering algorithm grouping executions close in time, on the same side (buy/sell), with compatible price characteristics. A grid search is performed over key parameters: maximum time window between consecutive executions (Δt_max), minimum volume participation threshold (π_min), and price tolerance (δ_p).

Each configuration is evaluated on two fronts: (1) adherence to metaorder impact stylised facts (power-law temporary impact curve, logarithmic permanent impact decay, size distribution); (2) fidelity to the LMF relation, linking market order variance to metaorder size with a 1/2 exponent. Results show a structural trade-off: configurations minimising impact stylised fact errors (Configuration A) yield an estimated LMF exponent ≈ 0.35, significantly different from 0.5. Configurations minimising LMF discrepancy (Configuration B) recover the 0.5 exponent by construction but show weaker single-stock fits for execution shortfall and impact decay.

The authors conclude that anonymous data do not allow an independent test of LMF theory: reconstruction is "underdetermined". However, compatibility observed in Configuration B suggests the theory is not falsified by public data. Future work could integrate L2 order book data or use machine learning techniques to improve identification.