HODOLOGICContact

Research

Researching better ways to understand and decide.

Hodologic Research conducts independent and applied research into the methods behind good analysis: how to model complex systems, how to represent what is uncertain, and how evidence should inform choices.

Browse publications Research areas
A long, symmetrical aisle of library shelves

Research areas

The methods behind good decisions.

Our research agenda follows the chain from data to decisions. Each area is chosen because it changes what an analysis can honestly claim.

Data

Measurement, quality and provenance of the evidence itself.

Modeling

Representing mechanisms at the right level of detail.

Complex systems

Networks, feedbacks and behaviour that emerges from interaction.

Uncertainty

Representing, propagating and communicating what is not known.

Prediction

Forecasts that are calibrated, not only accurate.

Causality

Telling what drives an outcome from what merely accompanies it.

Simulation

Exploring outcomes that cannot be observed directly.

Optimization

Good choices under constraints and competing goals.

Decision science

How analysis should, and does, inform choices.

Artificial intelligence

Where learned models help, and where they mislead.

Scientific methodology

Reproducibility, validation and honest reporting.

Research produces

  • Papers
  • Methods
  • Open-source tools
  • Datasets
  • Technical reports
  • Benchmarks

Research and consulting

Each one feeds the other.

Research develops methods and tests them in the open. Consulting brings the real questions that show which methods matter, and where they break.

Research

Produces new methods and knowledge.

Consulting

Brings real-world questions and applications.

Models

All models simplify reality. Good models simplify what matters.

Too simpleA straight line misses the structure.
UsefulCaptures the signal, ignores the noise.
Too complexFits every point, including the noise.
Fig. 1Three models fitted to the same sixteen observations. Only the middle one separates the signal from the noise.

Models as instruments

Explanation
Why an outcome occurs, and which mechanisms drive it.
Prediction
What is likely to happen next, with an honest range.
Simulation
What could happen under assumptions not yet observed.
Comparison
How alternatives differ on the criteria that matter.
Optimization
Which option performs best within real constraints.
Decision-making
What to do, given everything above and what is still unknown.

Methods we draw on

  • Statistics
  • Machine learning
  • Operations research
  • Econometrics
  • Simulation
  • Bayesian methods
  • Causal inference
  • Optimization
  • Agent-based modeling
  • Mathematical modeling

The method follows the question. We choose the simplest model that can support the decision, then test whether it does.

Uncertainty

Better decisions do not require certainty.

They require understanding what is known,what is uncertain,and what changes the decision.

04812020406080100Time →Outcome indexNOWDecision thresholdDECISIONt ≈ 81: upper boundcrosses thresholdOBSERVEDESTIMATEDPREDICTEDUNCERTAIN90% intervallower bound

Fig. 2Illustrative forecast. The central prediction never reaches the threshold, but the 90% interval does, so the decision is driven by the uncertainty rather than the average.

Observed
What was measured.
Estimated
What the model infers from the measurements.
Predicted
What the model expects next.
Uncertain
The range the model cannot rule out.
Decision
The point where the right choice changes.

Current research

Work in progress.

Research · In-house project

EdgeQuant · Forecasting and decisions in betting markets

Betting markets are a fast, unforgiving test of probabilistic forecasts. Every prediction has a price, and every outcome is known within days.

EdgeQuant is our test bed for turning forecasts into decisions under uncertainty. It combines model predictions, tipster consensus and live odds, keeps only selections with a measurable edge, and builds portfolios of bets with Markowitz optimization and fractional Kelly sizing. Monte Carlo simulation shows the range of capital paths and drawdowns before any stake is placed.

Data
39,000+ historical matches
Allocation
Markowitz, fractional Kelly
Risk
Monte Carlo, drawdown analysis
Validation
Brier score, closing line value, backtests
Open the app
0%3%5%8%10%0%5%10%15%20%Risk (standard deviation) →Expected returnMIN VARIANCEMAX SHARPEEFFICIENT FRONTIER
Fig. 4Each point is a possible portfolio of five bets. The frontier marks the best expected return for each level of risk. Illustrative figures.

Publications

Research outputs

Papers, reports, methods, datasets and software from Hodologic Research.

TypeTitleYear
Research paperAnalytical methods for territorial planning policyHodologic Research · 21 September 2026 · GIA · PDF · FR2026
White paperLe Kiosque: a sovereign document-intelligence platform for the Gabonese pressHodologic Research · July 2026 · Le Kiosque · PDF · FR2026
ManifestoLe Kiosque: principles for reading the pressHodologic Research · July 2026 · Le Kiosque · PDF · FR2026

All publications

Collaborate

Working on a related question?

We welcome research collaborations, data partnerships and reviews of our methods. Write to us with the question and what you have tried.

Contact Hodologic Research