WH Alpha
Wealth Horizon Quantitative Research
Research and market tools
U.S. historical market views. Separate A-share research. No admitted Alpha or model.
Guest access is temporarily unavailable.
Guest and signed-in sessions currently receive identical product capability.
65 formal trials consumed; no admitted Alpha or model input.
TF01 coverage repair stopped after engineering errors left insufficient source-stage budget. Full coverage remains incomplete and prediction untested; trial 66 is not registered.
Ledger V63 · 2026-10-14
AMIHUD’s four configurations retain bounded research-survivor admission, limited to verdict §3 (V1-001). No tradable Alpha.
Strict PIT, execution and regime dependence remain unproved. Model, strategy, Product and live permissions stay closed.
Verdict V1-001 · 2026-09-30 / Verdict V1-002 · 2026-10-01 / Verdict C-000 · 2026-10-01
Trial 65 completed; the fixed expression did not earn an upgrade or a recommended money route. User review pending.
Ledger V63 / MR01 · 2026-10-14
Earnings and leaderboard clues did not replicate; margin-finance screening is inconclusive.
LHB-002 / EP-003 / FIN-001 · 2026-10-02–03
Research standard
Inspect each model’s data clock, assumptions, costs, weaknesses, and current authority. Uncertainty is reported as a state—not hidden by presentation.
Stable instrument identity, historical membership, lifecycle evidence, corporate actions, and filing availability are governed without projecting today backward.
Development, validation, sealed holdout, and prospective observation remain distinct. Attractive in-sample results do not become proof.
Entry timing, turnover, spread, impact, capacity, and instrument-specific risk belong in the evaluation—not in a footnote after the result.
Logic, source fields, formulas, parameters, exclusions, evidence, counterevidence, fingerprints, and invalidation remain open to review.
Append-only progress record
Dated historical records. Negative results and closed gates remain visible.
Historical record: Phase S closed (Verdict 090). Corporate-action acquisition completed 10,592/10,592 (Verdict 096); OHLCV captured 5,292/5,296 with zero quarantine (Verdict 102). Verdict 104 conditionally opened the first closing-price/volume batch on 2026-09-29. F1–F4 execution closed on 2026-09-30.
Trading-status formal acquisition complete: V2 4,166/4,166. Source-clock methodology review passed (Verdict 091).
The U.S. line is unfrozen. V1.5 missed its frozen start and is terminal; the negative record is pending final publication wording. V1.6 is in design, not started. A-share Phase S formal acquisition started with 10,592 exchange-direct queries.
The 2,377-document result set and 599 locked quarantine records remained unchanged. The separate 2016–2021 retrospective acquisition is on hold pending confirmation of its execution state.
V1.5 cannot start late or be backfilled. The U.S. executor hardening passed independent review.
Choose a module, inspect its summary, then open the evidence. Earlier material is dated history.
Framework definitions, not completed stages or active model capabilities.
Finite factor tests retain rejected trials and unresolved limits.
Model construction requires admitted inputs with incremental evidence; no active model is admitted.
A locked model must precede entry, exit, sizing, costs, capacity and risk rules.
Fixed protocols separate Development, Validation and Holdout; later stages remain closed.
Prospective observation requires a surviving model and frozen implementation; not started.
Product activation requires a separate human decision; research survival grants no trading permission.
Ledger V5 · 2026-09-16
TF01 coverage repair stopped after engineering errors left insufficient source-stage budget. Full coverage remains incomplete and prediction untested; trial 66 is not registered.
Only research produced inside the factor → model → strategy architecture is presented as current Product research. Earlier pre-architecture methods remain internal audit history.
Each future ranking must identify its active model, market fit, rank drivers, entry readiness, chase risk, and invalidation conditions.
The current Candidate Baseline V1 remains visible and transparent, but is explicitly unvalidated and is not presented as an expected-return model.
Strict PIT, execution and regime dependence remain unproved. Model, strategy, Product and live permissions stay closed.
The planned AI research layer extends traditional quantitative research through bounded hypothesis generation and adversarial review. Isolated data stages, finite budgets, deterministic gates, and retained failures constrain the search.