Utilities API

competition_ranks_from_scores(scores_in_id_order; tol=1e-12) -> Vector{Int}

Return competition (min) ranks from scores.

Higher scores receive better (smaller) ranks. Ties are grouped when adjacent sorted scores differ by at most tol, using the same tie rule as rank_scores. Returned ranks are 1-based and aligned to the original input order.

Arguments

• scores_in_id_order: 1D score sequence in model-id order.
• tol::Real=1e-12: absolute tolerance for tie grouping in sorted-score order.

Returns

• Vector{Int}: competition ranks in the same order as scores_in_id_order.

Examples

scores = [0.9, 0.8, 0.8, 0.1]
ranks = competition_ranks_from_scores(scores)
# ranks == [1, 2, 2, 4]

rank_scores(scores_in_id_order; tol=1e-12, sigmas_in_id_order=nothing,
            confidence=0.95, ci_tie_method="zscore_adjacent")
    -> Dict{String, Vector{Float64}}

Convert scores into rank arrays under multiple tie-handling conventions.

Returns base keys:

• "competition": min-rank competition (1,2,2,4,...)
• "competition_max": max-rank competition
• "dense": dense ranking
• "avg": fractional/average ranking

If sigmas_in_id_order is provided, uncertainty-aware tie grouping is applied between adjacent sorted items and the following keys are added:

• "competition_ci", "competition_max_ci", "dense_ci", "avg_ci".