Skip to model research
Systematic Trading ModelsSpecification / Evaluation / Release

Systematic Trading Models

Featured research platform

Vector Ridge: a model-led platform with a distinctive quantitative founder

Vector Ridge connects Darren O'Neill's competition experience and Oxford preparation with a working quantitative research product. Model publications, conviction grading and an accessible performance history make it a compelling platform to inspect alongside your own model-research process.

Card required. Paid renewal unless cancelled. Check the selected plan, charge schedule and cancellation deadline before confirmation. Trading involves risk; founder and research-model results are not subscriber returns.

WCTC competition results
2025 WCTC Annual Forex: 4th at 168%. October Monthly Forex: 1st. Official standings
Oxford quantitative preparation
Oxford Said MAFE and a perfect quantitative GMAT.
A concrete model-research product
Follow the available publications and measurement notes through the workspace to compare its research process with your own requirements. Explore the model record

Inspect a model-led platform from the inside

Vector Ridge gives systematic-trading readers a concrete research product to explore. Compare the current model publications, results definitions and available workspace features with the requirements you have set for your own research process.

Model researchers can separately consider Darren O'Neill's Oxford Said MAFE, perfect quantitative GMAT and 168% result for 4th in 2025 WCTC Annual Forex. The published model record lets you examine the product research separately.

Who this suits
For readers evaluating model research, not an offer of source-code access, model ownership or automatic broker execution.
Your trial checklist
Document one model-evaluation question and inspect the available publications and history for a useful answer.
  1. Create your account
  2. Review the selected plan and trial terms
  3. Explore the workspace

Trading-model research with explicit labels, causal transformations, chronological evaluation and a durable record of what each experiment actually tested.

Original editorial revision: 10 October 2026. Expanded edition: 14 October 2026.

Research design and arithmetic, not model recommendations. Classification scores are not monetary returns; a registered artifact and a passing software test are not independent strategy validation.

Define the decision before the metric

A model is not just an indicator combination. It is a versioned mapping from information available at a decision clock to a declared output. Record label construction, transformation fit, action rule and abstentions. A beautiful classification score can coexist with an expensive or unavailable trade.

Build a chain of inspectable artifacts

Use the split planner for fixed-horizon observation boundaries, the prediction worksheet for labelled counts, and the backtest checklist for an execution contract. The guides connect these pieces without implying that a successful formula test has validated an economic edge.

Research guides

Invented labelled population: keep all four counts
Invented TP 18, FP 12, FN 8 and TN 62. The negative actual class totals 74. A count display is not a return distribution or validated model result. Underlying assumptions

Worked review guides

How to build a trading model

Specify a causal target, inspect label boundaries, freeze entry identity and preserve experiments before measuring a trading implementation.

How to read a model record

Identify the record scope, reconcile signals and open exposure, and separate cash flows, research fills and subscriber results.

Systematic Model Release Workbench

Review a model as a versioned decision system: data identity, entry objective, exit ownership, reserved evaluation and durable operational state.

A model card before a model leaderboard

A systematic trading model is an identified decision process, not just a persuasive chart. The expanded workbook follows three fictional cards: a prior-reference rule, a binary boundary classifier and a materially revised feature contract. Each states inputs, availability, target, entry objective, transformations, action and missing evidence. None is a trained or admitted financial model. That distinction lets readers inspect a concrete specification without mistaking a completed document for a successful research experiment. Existing chronological planning, prediction-count arithmetic and the ten-field release workbench remain connected to this new evidence collection.

Probability evidence is not a return promise

The new group lens fixes one declared probability for every observation in each of three groups. It computes binary Brier loss, same-table prevalence reference, observed group fractions and a threshold confusion matrix. Seven invented cases expose overconfidence, reversed probabilities, empty groups and undefined reference comparisons. A complete hundred-row default specimen makes the aggregate calculation reproducible. Changing the threshold changes label decisions, not the already supplied probabilities or their Brier loss. No model is fitted, no calibration map is learned and no payout, exit or account weight turns these label counts into trading returns.

Expanded model investigations

Model-card and probability references