KabuForge
Menu
GitHub日本語

Research Methodology and Limitations

What KabuForge visibility checks and local simulations establish, and what they cannot certify.

Availability at decision time

In v0.1.1, inputs declare a timezone-aware decision timestamp and snapshot identity. check_point_in_time and check_lookahead inspect available_at, count future-visible rows and require a visibility gate. Missing availability timestamps invalidate this check; declared timestamps do not prove genuine historical availability.

A visibility gate has limits

This does not certify source provenance, vendor timing, historical revisions, survivorship, corporate actions, universe construction or economic validity. Factor diagnostics do not establish out-of-sample performance or investability.

Simulation semantics

Research marks and execution quotes are separate inputs. Order intents are plans; simulated fills are local records rather than broker fills. Costs, liquidity, capacity, recovery and real broker behavior need independent validation.

Release validation is bounded

The v0.1.0 notes retain an independent 29/30 P0 acceptance result. The full Ubuntu suite did not complete; a reused synthetic case matched selected outputs. These historical release statements are not current website validation or a claim that the remaining gate passed.

Bring appropriately licensed data

KabuForge grants no vendor-data access or redistribution rights. Validate external datasets and their history independently before interpreting any research result. The synthetic demo cannot measure investment performance.

Applies to: v0.1.1 · Public source docs/en_US/RESEARCH_METHODOLOGY.md

Reviewed · KabuForge