Use a transparent hierarchy for hidden-stem membership, layer, season, position, exposure, and branch integrity without presenting one modern score as universal doctrine.
Hidden-stem weight is a methodology choice, not a single universally accepted percentage table. A defensible system first preserves canonical branch membership and order, then applies disclosed layers for primary or secondary qi, month command, season, pillar position, exposure, roots, and active branch relations. OpenFate separates source facts, engine coefficients, and interpretive conclusions so each can be audited.
At a glance
Stable layer
Branch membership and hidden-stem order
Method layer
Primary, middle, and residual coefficients
Context layer
Month command and seasonal phase
Topology layer
Position, exposure, roots, and interactions
Governance
Versioned coefficients and reproducible inputs
Limit
No universal 60/30/10 rule for every chart
Freeze canonical data before adding weights
Start with the twelve-branch hidden-stem table and its order. These membership facts should be identical across language versions and product surfaces.
Do not let a score remove a hidden stem or invent one that the table does not contain.
Keep source data immutable.
Test locale parity.
Declare every coefficient and its purpose
If the model distinguishes primary, middle, and residual qi, publish the coefficient source or state that it is product policy. Do the same for season, month, pillar, and spatial adjustments.
Avoid presenting a convenient modern percentage as a verbatim classical rule when no primary source supplies it.
Version the weighting policy.
Separate tradition from implementation.
Apply interactions after the base snapshot
Store an unmodified hidden-stem and element snapshot, then calculate combination, clash, punishment, or break effects in a separate layer. This preserves traceability.
Use bounded changes and record which branch and hidden stem produced each adjustment. Harm or Void should not inherit numerical penalties unless the policy explicitly defines them.
Keep before-and-after values.
Attribute every adjustment.
Validate outputs, not only formulas
Test canonical tables, total-mass conservation where applicable, edge cases, multilingual labels, and whether a display-only factor leaks into core strength or useful-god logic.
Publish the method as an estimate inside a traditional framework, not as scientific measurement or professional advice.
產品數值模型須公開輸入、基準、季節與位置修正、上限及版本，不把係數冒充古典原文OpenFate separates deterministic chart calculation, traditional interpretation, and modern editorial explanation.
