INTELLIGENCE / MODELS

A model strengthens a hypothesis. It does not replace one.

Models in kquant act as forecasts, regime filters and meta-labels around formal strategy logic. Their incremental contribution is measured separately and faces the same independent validation.

Launch research Explore the details ↓

SYSTEM DOMAIN02 / DISCOVER

Search, rank and freeze a candidate

DISCOVER
OPTIMIZE
VALIDATE
RUN
OBSERVE

Distinct roles

Forecasting, regime filtering and meta-labeling answer different questions instead of collapsing into one AI score.

Incremental value

A model-free anchor is compared with the candidate to measure whether the model adds durable quality.

Time-series discipline

Training, features and evaluation stay separated through time; future values never enter the current decision.

Honest rejection

An unstable model can be removed while preserving the baseline strategy and the complete experiment history.

DECISION FRAMEWORK

Did the candidate improve evidence or only the training score?

Read the candidate against its anchor, neighbouring parameters, sample size, costs and complete trial history.

Bound the search space

Compare several objectives

Freeze before opening holdout

OUTPUT

Candidate record

Important

Historical, statistical and model results do not promise future returns. kquant provides computational tools and technical execution, but does not assess whether a specific trade is suitable for a user.