TradeLean: the one-indicator hypothesis / full0.702 zone study
English editorial review, 25 October2026

This is a code-and-derived-results package. The cached raw OHLCV histories are not embedded. Re-running numerical studies requires the input cache files and universe metadata recorded in the manifests, in the same project layout. No exchange or account credentials are included.

The full zone is [0.618,0.786] in outward pivot-range coordinates, symmetric on the support and resistance sides. Its width is0.168 times the previous reference-period range. Whole-band comparisons use adjacent bands of that same width.

zone_evidence.py produces full-band profiles, close-based episodes, reference-period/arrival observations, and85 parameter configurations from17 directional or momentum families. These are not85 independent indicators or100 complete trading strategies. ATR, standard deviation and efficiency ratio are not scored as if they were directional extremes.

An observed-price witness uses a recorded open, high, low or close inside the band. RSI at that price is reconstructed from the previous completed native-bar Wilder state. The OHLC-compatible ceiling is a range bound, not a confidence interval or a claim that all intermediate prices traded. Right-censored close episodes and incomplete reference-period arrival samples remain separate.

rsi_zone_correspondence.py contains the earlier completed-close and nominal-centre correspondence study. Its centre-contact input is a derived subset of the archived universality event table. Those older nominal contacts are not the primary full-zone observation unit.

benchmark.py is the separate20-family ×5-window exploratory direction screen. Its target, train/test split and limitations are in benchmark-specification.json. The100-tool-evidence-audit.csv describes coverage for every named catalogue entry; it does not certify universal replacement.

Relevant verification commands from a project root with Python dependencies installed:
python -m pytest research/pivot-indicator-audit/test_zone_evidence.py -q
python -m pytest research/pivot-indicator-audit/test_rsi_zone_correspondence.py -q
python research/pivot-indicator-audit/zone_evidence.py

Dependencies include Python3, NumPy, pandas, scikit-learn and a parquet reader such as pyarrow. Data sources, coverage and source hashes are in the included manifests. Selection is an availability sample, not a random census of every market. The research does not establish universal indicator superiority or trading profitability.

Selected case studies: case_studies.py scans BTC, ETH, SPY and AAPL from 2025 onward. Four positive examples are editorial selections after viewing outcomes. All 57 close/reclaim candidates, the full selection specification, source hashes, per-bar indicator values and validation are included. Confirmation price changes are not executable returns. Conditions absent by week-end are censored there; indicators already satisfied at the price response are retained with zero extra bars.

Matched full-zone audit (26 October 2026): paired_zone_test.py compares 1,725 original executable setups and 20,700 matched policy rows. Each of six conditions is an immediate gate or a wait-up-to-six-bars policy. Same original stop, pivot target, terminal native bar and original baseline risk unit; skips score cash zero. No positive-outcome selection. Costs 0/10/25 bp are hypothetical scenarios, not calibrated asset-specific execution costs. Both-hit candles remain with stop-first and target-first sensitivity. 11 unit tests cover next-open timing, adverse gaps, both-hit handling, symmetry, unchanged risk units, wait cancellation, causal feature state and gap restart. Summary, coverage and all paired rows are in this download. The data are reused cached history capped before 25 October 2026, not independent prospective validation.
