A Durant Labs research project
ALPHAROUTER
Adaptive intelligence
for prediction markets.
The inquiry
How does information
become price?
Alpha Router is an independent research system investigating the relationship between information, probability, and market prices.
Its architecture is designed to connect settlement modeling, market forecasting, probabilistic reasoning, systematic discovery, execution analysis, and risk-aware capital allocation within one continuously evaluated research process.
Research architecture
Reality and markets.
Separate what is likely to happen from how markets are likely to price it.
The Truth Model studies event-settlement probabilities in the context of contract rules, source evidence, and calibration. The Market Model studies future prices, information propagation, and the persistence of pricing differences. The observation layer organizes contracts, rules, order books, trades, source observations, and outcomes with explicit timing, provenance, and evidence quality.
Structure and discovery.
Study the relationships between contracts, then test what those relationships can tell us.
The Global Shadow Market examines logical consistency, conditional probabilities, and joint outcomes. Automatic Alpha Discovery investigates hypotheses, feature interactions, and changing market regimes through out-of-sample and forward evaluation. Active Intelligence asks when additional information could improve a decision enough to justify its cost.
Execution and uncertainty.
A modeled opportunity must be examined in the conditions where it would be executed.
Market Ecology studies observable liquidity, quote behavior, fill likelihood, adverse selection, and competitive conditions. Execution Intelligence extends that inquiry to order placement, price improvement, waiting, cancellation, partial fills, slippage, and costs. The capital-allocation design weighs estimated edge against uncertainty, correlation, liquidity, capacity, capital requirements, and portfolio drawdowns.
The integrated process
One research loop.
Continuously questioned.
- Observe
- Model
- Discover
- Validate
- Allocate
- Execute under controls
- Measure
- Learn
The intended loop connects evidence, models, decisions, and measured outcomes. Its research objective is to discover and validate new advantages as existing relationships change or deteriorate.
Research discipline
Evidence before
conviction.
Evaluation is designed to expose failure as well as identify promise. Hypotheses, forecasts, decisions, and outcomes inform the next experiment.
Probabilistic estimates remain distinct from logical guarantees. Historical patterns require independent predictive evidence. Model confidence does not remove uncertainty.
Simulation, forward paper evaluation, and real trading are separate evidence classes. A research result is not a realized financial return; hypothetical opportunities must still survive the economics of execution.