Abstract data visualization background

Where forecasts once ended, duration now begins

Markets have always been read in chapters — bull runs, bear retreats, the long pauses between. For decades, analysts measured what was happening. Few asked how long the chapter would last. Fruelazcovtemmue builds models that estimate the expected length of prevailing market conditions, turning regime detection into regime duration.

12

Research quarters

40+

Markets modeled

From detection to duration

A regime identified is a regime half-understood. The other half is knowing how long it might endure.
Regime duration curves
Multi-market coverage
Weekly model refresh
Transparent methodology

Our story

The project began in a small research unit where two questions kept surfacing: how long does a market regime typically last, and can machine learning estimate that before it ends.

What the research uncovered

  • Regime boundaries are fuzzy: Markets do not switch states at a single price point. The transition zone can stretch for weeks, and models must account for that uncertainty.
  • Duration follows patterns: Historical regime lengths cluster around certain ranges. Machine learning can learn these distributions and update them as new data arrives.
  • Context changes everything: The same price pattern means different things in different macroeconomic settings. Models that ignore context misread duration by wide margins.
  • Transparency builds trust: Black-box forecasts serve no one. Every model output is paired with an explanation of the factors that shaped the duration estimate.

Research goals

The research program pursues three parallel objectives: improve the accuracy of regime duration estimates across equity, fixed income, and commodity markets; develop interpretable models that analysts can inspect and challenge; and publish findings in peer-reviewed venues to advance the broader field of quantitative finance. The work is ongoing. Each quarter brings new data, new model variants, and new questions about what drives markets to persist or pivot.

What sets the approach apart

Unlike tools that flag regime changes after they occur, these models estimate expected remaining duration while a regime is still active. The difference is the gap between diagnosis and prognosis — between knowing what happened and estimating what remains.

Working principles

Six commitments that shape how research is conducted, how models are built, and how findings are shared with the market.
01 Intellectual rigor
Every model is stress-tested against historical data, alternative specifications, and the possibility that the underlying structure of markets has shifted.
1
02
Method transparency
Users deserve to know how a duration estimate was derived. Model architecture, input variables, and update logic are documented and available for review.
2
03
Client partnership
Research improves when it is challenged. The team works closely with practitioners who use the forecasts, incorporating their feedback into model refinement.
3
04 Continuous learning
Markets evolve. Models that performed well last year may falter this year. The research program treats every forecast as a learning opportunity.
4
05 Honest uncertainty
Duration estimates come with confidence intervals, not point predictions. The team reports what the model knows and what it does not.
5
06 Research integrity

Findings are reported as they are, not as they might be wished. Negative results are documented and analyzed with the same care as positive ones.

6

Company trajectory

A young research organization with a focused mission and a growing record of published work, client engagement, and model refinement.

The numbers reflect early momentum, not final achievement. They are offered as markers of progress, not proof of arrival.
120

Research papers read

The team maintains a disciplined reading program across machine learning, econometrics, and market microstructure literature.

18

Model iterations

Each major revision incorporates new data sources, architectural changes, or feedback from practitioners in the field.

6

Quarters of operation

The organization was founded in early 2025 and has released weekly forecasts since its third quarter.

9

Team members

A small group of researchers and engineers with backgrounds in quantitative finance, statistics, and software development.
47

Client conversations

Direct dialogue with portfolio managers, risk officers, and quantitative researchers informs every model update.

The people behind the models

A compact team of researchers and engineers who build, test, and refine the duration forecasting system. Each member brings a distinct specialization, and the group functions as a collaborative unit rather than a hierarchy of isolated experts.

Dr. Eleanor Vance

Dr. Eleanor Vance

Lead Research Scientist

Dr. Vance spent twelve years in academic research before moving to industry. Her work on Bayesian duration models forms the theoretical backbone of the forecasting system. She oversees model architecture and ensures that every release meets rigorous validation standards.

Core specializations

Bayesian time series Regime-switching models Model interpretability Academic publishing
Marcus Chen-Whitmore

Marcus Chen-Whitmore

Head of Quantitative Engineering

Chen-Whitmore designs the infrastructure that turns research code into stable, scalable forecasting pipelines. His background spans hedge fund technology and cloud architecture, and he insists that models must be as robust in production as they are elegant in theory.

Core specializations

Machine learning systems Distributed computation Data pipeline design Production reliability
Priya Nandakumar

Priya Nandakumar

Senior Market Analyst

Nandakumar bridges the gap between model output and market context. She tests forecasts against real-world conditions, translates technical results into practitioner language, and gathers feedback that drives the next research cycle.

Core specializations

Macroeconomic analysis Cross-asset research Client communication Forecast validation
Modern office workspace interior