Add v5-vs-v4 held-out confirmation scorecard + value-calibration caveat (VQ-496)

README.md

@@ -44,10 +44,32 @@ fair cross-version metric.
 **The ~+300 Elo from v4→v5 comes from the teacher upgrade alone** (Stockfish →
 lc0), at identical positions and scale — the Tier-1 distillation result.
 
-> Elo caveat: the v5 gate used an 18-opening fallback suite (vs v4's broader
-> suite), so v5-vs-v4 is not pixel-identical methodology; the ~+194 lower-bound
-> margin dwarfs any opening-suite effect, but a same-suite re-eval is the airtight
-> confirmation. See each version's `metrics.json`.
+### Confirmation (held-out scorecard, v5 vs v4)
+
+Beyond the game-play Elo, v5 and v4 were scored on the **same 70,198-position
+held-out set** (seed 42, same pipeline, **vs Stockfish labels**) — a
+deterministic same-methodology check:
+
+| metric | v4 | v5 |
+|---|--:|--:|
+| policy top-1 (= agrees with Stockfish's best move) | 0.392 | **0.527** |
+| mean regret (cp) | 118 | **72** |
+| blunder rate (>300cp) | 0.130 | **0.077** |
+| value MAE | 0.213 | 0.323 |
+| value ECE | 0.076 | 0.273 |
+
+v5's **policy is decisively stronger** — and *teacher-agnostically* so (lower
+regret/blunder vs Stockfish's own cp; it even matches Stockfish's best move more
+often than the Stockfish-distilled v4). That's what drives MCTS strength and
+confirms the Elo gain.
+
+> **Honest caveat:** v5's **value** is *worse*-calibrated against Stockfish targets
+> (MAE/ECE up). This is a **teacher-scale artifact** — v5's value is trained to
+> lc0's WDL scale, not Stockfish's `tanh(cp/400)` — not a strength regression (the
+> Elo gate integrates policy + value and still shows ~+300). WDL recalibration is
+> planned future work. v5's Elo gate used an 18-opening fallback suite, so the
+> game-play number is directional; the held-out scorecard above is the
+> same-methodology confirmation.
 
 ## Files