POOL IQ IS CLOSED (2026-09-23). This board is kept as the record of a published result, not as a
live tool: the pre-registered gate resolved, the frozen v5 scorer did not clear it, and the
FAIL below stands exactly as published on 2026-09-02. Collection stopped 16:04:27 UTC, so the
numbers here no longer change.
The same question — is providing liquidity on these pools worth it right now? — is answered forward,
every day, at XRPL v2. Closure notice:
on the public record.
Verdict status
OUTCOME C — FAIL.
The gate is powered and v5 does not clear it.
Pre-registered dual gate — HAC t ≥ 2.0 AND ≥ 10/14 non-overlapping
phases. 14/14 phases resolved, so the gate is powered; neither leg clears — HAC t −1.46
and +0.96 against the pre-registered 2.0, and phase-stability 7% against a
required 100%. Recorded 2026-09-02 as 20260902_180000_xrpl_pooliq_verdict_20260902.
A negative result, published in full — that was the commitment.
Live gate read: —
Contract number — cross-sectional IC the gate reads (verbatim panel)
ΔIC[t1 − t0] CONTRACT · GATE PROMOTION TEST
—
—
—
The model (t1: GBM fees + E[IL]) vs the null (t0: naïve fee-APY).
A negative value means the model does not out-rank the free baseline.
ΔIC[t1 − heur]
—
—
—
Model vs the hand-tuned il_risk heuristic a screener computes for free.
Underpowered by design. With 12 pools/date required, only
8 dates qualify for cross-sectional IC; the non-overlapping phase sweep
resolves 0/14 (needs ≥ 10). Wide intervals are the expected state of a
pre-registered read before enough 14-day cohorts have closed — not a defect. Both contract
legs are negative; nothing is near promotion.
Level IC — secondary context, not a claim of skill
IC[t1] model
—
—
IC[t0] null
—
—
IC[heur]
—
—
Read the level IC with care. Within-date rank is dominated by
fee-APY, which the null (t0) already captures — so a high level IC (0.7–0.9) is
not evidence the model adds anything. The model's claim is tested only by the
paired ΔIC above, which is negative. No hit rate is computed or shown, by design.
Count reconciliation — which number is computed on which panel
Contract panel (IC computed here)
grader-verbatim · what the gate reads
cohort rows
—
resolved (14-day window closed)
—
Display partition (ledger below)
verdict-only · EXCLUDE carries no verdict
resolved (verdict-bearing)
—
pending (verdict-bearing)
—
excluded (no verdict issued)
—
total cohorts
—
—
Collection gaps — published
—
Cohort ledger — every prediction, pending and resolved
—
Pool
Open date
Entry verdict
Resolution
Realized net 14d
Path A. This ledger publishes on the pre-registered schedule
regardless of what the outcomes show. The verdict is whatever the frozen gate reads — it
cannot be re-gated, delayed, or withdrawn after the fact. Predictions were frozen before any
14-day cohort matured.