Statistical arbitrage & module discovery

Graph-conditioned residuals, pair selection, and a constrained search harness for testing alternative research modules.

PythonGraphsVECMStatistical arbitrageSkyDiscoverDeterministic evaluators

Built systemRank 02 of 12 in this field ↗

Why it sits here. Placed by the scope of implementation, the available contribution evidence, and the distinct technical capability it demonstrates.

06 / Two sides. One uncertain future.STUDY IN SPACE

01 / IMPLEMENTATION & CONTRIBUTION

What the work involves

Donald implemented graph, residual, expected-net-alpha, pair-selection, and execution layers, plus a bounded task/evaluator pack for exploring alternative research modules.

Technical depth

Dynamic peer graphs, efficient-price filtering, VECM residual experts, pair-quality admission, expected-net-alpha decomposition, portfolio constraints and a separation between search evaluation and promotion validation.

The project family

pair-tradesky_discover_pair_trade

02 / RESULTS

What came out of it

Six constrained search tasks sit alongside graph and residual-model components, with separate promotion and full-system validation entrypoints.

03 / SUPPORTING EVIDENCE

Follow the source

Implementation notes, project records, and supporting artifacts.

Source context & project scope

The design document mixes plans with implemented work; only source-backed components are claimed complete. Minute-bar search does not establish order-book execution fidelity. No gains from automated search are asserted.

pairability.py

Pairability graph, forward-correlation estimator and graph builder are implemented.

SOURCE · 2026-09-17

net_alpha.py

ExpectedNetAlphaModel and score decomposition source.

SOURCE · 2026-09-17

production_engine.py

Integration source exists; runtime deployment not verified.

SOURCE · 2026-09-17

README.md

Six EVOLVE-BLOCK tasks and explicit separation of search harness, promotion gate and integrated validation.

SOURCE · 2026-09-17