Lab research — Tradour

What we tested

Within the Consumer Discretionary rule, did RISE-labelled observations match their frozen direction more often than DROP-labelled ones?

The specification was frozen before this run. Neutral and unevaluated cases do not enter the directional success-rate denominator.

What we found

54.3% matched the frozen rule.

334 matched and 281 did not, from 615 directional cases.

Eligible observations: 747
Distinct tickers: 69
Evidence period: 22 Nov 2022 – 7 Aug 2026

How strong is the evidence?

Keep the range beside the headline.

The 95% Wilson interval is 50.4% to 58.2%. The archive was already exposed, so this is not a clean holdout result.

0%50% comparator100%

See it

Where the result changed

DROP52.7%

450 directional cases

RISE58.8%

165 directional cases

What this does not prove

A useful historical result is not a trading instruction.

The outcome is the frozen rule-aligned legacy directional result, not a canonical market return.
This archive was available before the Lab test was registered, so the result is exploratory rather than clean out-of-sample evidence.
Repeated tickers and shared market events may reduce the effective sample size.

What could we test next?

Fork the question; never rewrite this result.

A variation becomes a new frozen research object linked back to this study.

Test a variation

Research recipe

Dataset: bsp-evidence-2026-08-19
Archive through: 2026-08-19
Engine: tradour-lab-engine-v1
Outcome: frozen_rule_directional_success
Minimum sample: 20
Functions: filter_observations → calculate_sample_summary → compare_groups → wilson_confidence_interval

Original specification preserved for reruns. Post-result variations create a new study.

{
  "spec_version": "tradour-lab-spec-v1",
  "question": "Within the Consumer Discretionary rule, did RISE-labelled observations match their frozen direction more often than DROP-labelled ones?",
  "hypothesis": "Consumer Discretionary RISE-labelled observations have a different directional success rate from DROP-labelled observations.",
  "universe": "canonical_bsp_rule_observations",
  "rule_ids": [
    "consumer-discretionary-follow-displayed-v1"
  ],
  "displayed_directions": [],
  "source_sessions": [],
  "sectors": [],
  "industries": [],
  "min_total_indicators": null,
  "min_rd_confidence": null,
  "date_start": null,
  "date_end": null,
  "group_by": "displayed_direction",
  "minimum_sample": 20,
  "chronological_split": "2024-12-31",
  "outcome_metric": "frozen_rule_directional_success",
  "benchmark": "balanced_direction_comparator_50_percent"
}