Ghana vs. Panama - Post-Event Review
2026 World Cup, Group L, Match 19 | June 17/18, 2026 | Toronto Stadium (BMO Field), Toronto
Predicted vs. Actual | Calibration: ~78%
TL;DR
Result: Ghana 1-0 Panama (HT 0-0; Yirenkyi 90+5'). The coin-flip held (Finding 9 ✅) and broke Ghana's way — the 34% underdog side of the coin won. Scoreline 1-0 was in ALL 3 models' top 3 (the #1 calibration win of the night). But the headline story is a FINDING 13 TWIST: we flagged Ghana for sterile domination (gutted attack), but it was PANAMA who dominated possession (62%, 502 passes — a WC record) without converting (0.75 xG → 0 goals). Ghana were the EFFICIENT counter-attackers, not the sterile team. The bench-decides-it-for-Panama "Watch" variable inverted completely: Ghana's bench (Thomas-Asante assist) won it; Panama's loaded bench (Fajardo/Díaz/Carrasquilla/Godoy) all came on and combined for zero goal involvements.
- Winner: ⚠️ partial (coin-flip) — consensus PAN 36% / DRAW 30% / GHA 34%. Ghana (34%) won; per Finding 9, the underdog side of a coin-flip validates at 0.5 credit. The slight PAN lean was wrong, but the match was correctly framed as a true coin-flip.
- Scoreline: ✅✅ 1-0 GHA — in all 3 models' top 3 (mm3 12% #3, g4 19% #3, g5 12% #3). The best scoreline calibration of the Jun 17 wave.
- Who Scores: ❌ on the name, ✅ on the logic — Ghana #1 Semenyo did NOT score (❌) but was Player of the Match with the decisive run+pass for the goal (star-as-creator, Finding 7 ✅). The actual scorer (Yirenkyi) was a deep-lying midfielder — not in anyone's top 5. Panama #1 (Waterman/Fajardo) ❌.
- Ayew avoidance (Finding 16): ✅ — we AVOIDED Ayew as Ghana #1 (age-decline). He was subbed at 86', scored 0, missed a tap-in at 65'. Second validation of the age-decline discount (after Ronaldo).
- **
[...truncated 150 lines...]
ch). |
| 2. Finding 13 is DIRECTION-AGNOSTIC — apply to whoever dominates possession without xG, regardless of "favorite" status. A defensive team that sits deep and counters is EFFICIENT, not sterile. | doc-1 Finding 13 + doc-5 Step 4. Add: "Identify the projected possession-dominant team, then run the <1.0 xG screen on THEM." |
| 3. Bench strength ≠ bench impact. A "loaded bench" (Panama's Fajardo/Díaz/Carrasquilla) can produce zero G/A if the team is chasing and the block is deep. Don't over-weight bench creators against a low block. | doc-1 Finding 7 (substitution risk). Add: "bench-creator impact is conditional on game state — a chasing team's bench creators face a set defense." |
| 4. Late winner from a non-scorer is the low-event coin-flip signature. When two Tier 3 teams play, xG is low and the result hinges on one moment — often late (90+'), often from a deep midfielder or a sub. Broaden Who Scores to include "late-arriving CM/deep runner" at ~5-8%. | doc-1 Finding 7 + 9. Add a "low-event late-winner" scorer archetype. |
13. Updated Cumulative Stats (20 matches)
| Metric | Before (19, post POR-DRC) | After GHA-PAN (20) | Trend |
|---|---|---|---|
| Winner hit (coin-flip = 0.5) | ~9.5/19 (50%) | ~10/20 (50%) | — (coin-flip credit) |
| Exact scoreline top-3 | ~7/19 (37%) | ~8/20 (40%) | ↑ (1-0 GHA hit in all 3) |
| Who Scores ≥1 hit | ~12/19 (63%) | ~12/20 (60%) | ↓ (Semenyo didn't score) |
| Referee card count | ~11/19 (58%) | ~12/20 (60%) | ↑ (Nyberg 3Y exact) |
| Referee pen prediction (low band) | new | ✅ held (0 pens) | ↑ |
| Finding 12 (conservation=0) | — | ✅ confirmed | — |
| Overall | ~68-70% | ~70-72% | ↑ (rebounded) |
14. Group L after MD1
| Pos | Team | P | W | D | L | GF | GA | Pts |
|---|---|---|---|---|---|---|---|---|
| 1 | England | 1 | 1 | 0 | 0 | 4 | 2 | 3 (+2) |
| 2 | Ghana | 1 | 1 | 0 | 0 | 1 | 0 | 3 (+1) |
| 3 | Croatia | 1 | 0 | 0 | 1 | 2 | 4 | 0 (-2) |
| 4 | Panama | 1 | 0 | 0 | 1 | 0 | 1 | 0 (-1) |
Ghana's late win puts them 2nd on goal difference behind England. Panama remain pointless in WC history (0 pts, 13 goals conceded across 5 matches). MD2: England vs Ghana (the headline group match), Croatia vs Panama.
15. Pipeline Test Results (the generated-prompt question)
This was the first match through the rebuilt generated-prompt pipeline (byte-identical 17 calibration rules via build.mjs). The verdict:
- ✅ All 3 writers applied Finding 4a correctly (hard on Ghana, restrained on Panama) — including g4, which has historically over-applied it.
- ✅ g4 was NOT a bad outlier. Its 1-0 GHA prediction (19%) was the HIGHEST of the three — its more-extreme distribution actually captured the upset better than mm3/g5. The bias-correction notes did NOT make g4 worse; they removed the wrong extremes (scorer #1 inflation, favorite-bull) while preserving useful variance.
- ✅ Finding 16 (age-decline) applied unanimously — all 3 avoided Ayew as #1.
- ✅ Finishing-quality gate applied — all 3 flagged Ghana's low-xG potential.
- ⚠️ The ONE miss: Finding 13 direction. All 3 applied the sterile-domination screen to GHANA (the gutted favorite) rather than PANAMA (who became the possession-dominant team). This is a methodology gap, not a pipeline gap — the rule itself needs the direction-agnostic fix (Learning #2 above).
Bottom line: the generated-prompt pipeline works. The byte-identical rules produced a tight, well-calibrated consensus (1-0 in all top-3s) and g4's bias-correction held. The remaining miss is a content error in Finding 13, which the build makes trivial to fix in one place (_core.md) for all 4 writers.
Post-event review for TASK-47. Sources archived: sources/matches/gha-vs-pan/post-event/ (7 files: Opta, BBC, AS USA, The Athletic, Ghana FA, weather, referee). No betting odds used. FINAL stats only (Opta). First post-event review of the generated-prompt pipeline era.