World Cup 2026 β€” Predictions vs Results

Master Summary Table

Completed Matches β€” 48 reviewed (24 MD1 + 24 MD2 all groups)

MD1 (Jun 11-17) β€” 24 matches

Match Prediction Win % Actual Winner Score Hit Who Scores Acc
MEX vs RSA MEX win MEX ~fav 2-0 βœ… MEX β€” β€” ~rev
KOR vs CZE KOR win KOR ~fav 2-1 βœ… KOR β€” β€” ~rev
CAN vs BiH CAN 1-0 (35%) CAN ~50% 1-1 Draw ❌ (#3) βœ… Both mechanisms ~73%
USA vs PAR USA 2-1 (30%) USA ~70% 4-1 βœ… USA ❌ (#4) βœ… Balogun (#2) ~75%
QAT vs SUI SUI 2-0 (35%) SUI ~70% 1-1 Draw ❌ ❌ 0/2 ~56%
HAI vs SCO SCO 2-0 (30%) SCO ~78% 0-1 βœ… SCO ❌ βœ… McGinn (#3) ~78%
BRA vs MAR BRA 2-1 (28%) ~45% 1-1 Draw βœ… (#2) βœ…βœ… Vini #1 + Saibari #3 ~82%
AUS vs TUR TUR 2-0 (30%) TUR ~83% 2-0 ❌ AUS! ❌ ❌ 0/6 ~45%
GER vs CUR GER 4-0 (28%) GER ~94% 7-1 βœ… GER ❌ βœ…βœ… Havertz #1 (brace) + Musiala #2 ~72%
NED vs JPN NED 2-1 (18%) NED ~46% 2-2 Draw ⚠️ (draw #2) ❌ but Van Dijk #5 + Kubo #2 assist ~79%
CIV vs ECU 1-1 Draw (18%) ECU 35%, CIV 35% 1-0 ⚠️ CIV βœ… (#3) βœ…βœ… Amad #1 (sub!) ~78%
SWE vs TUN SWE 2-1 (18%) SWE ~60% 5-1 βœ… SWE ❌ (margin) βœ…βœ… GyΓΆkeres #1 + Isak #2 ~72%
ESP vs CPV ESP 3-0 (25%) ESP ~90% 0-0 Draw! ❌ ❌ Yamal benched ~30% (biggest MD1 miss)
BEL vs EGY BEL 2-0 (20%) BEL ~62% 1-1 Draw βœ… (#3) ⚠️ Salah assisted, Lukaku forced OG ~65%
KSA vs URU URU 1-0 (rev 49%) URU 49% 1-1 Draw βœ… (#2) ⚠️ Al-Amri (CB!) scored ~70%
IRN vs NZL IRN 1-0 (rev 53%) IRN 53% 2-2 Draw ⚠️ (draw #1-2) βœ…βœ… Just #2 + Mohebi #3 ~75%
FRA vs SEN FRA 2-1 (~17%) FRA ~56% / D ~25% 3-1 βœ… FRA ⚠️ (3-1 ~8%) βœ…βœ… MbappΓ© #1 (brace!) ~75%
IRA vs NOR NOR 2-0 (~18%) NOR ~59% / D ~25% 4-1 βœ… NOR ❌ (margin) βœ…βœ… Haaland #1 (brace!) + Hussein #1 ~78%
ARG vs ALG ARG 2-0 (~21%) ARG ~62% / D ~24% 3-0 βœ… ARG ⚠️ (3-0 ~8%) βœ…βœ…βœ… Messi #1 (HAT-TRICK!) ~80%
AUT vs JOR AUT 2-0 (~15%) AUT ~58% / D ~27% 3-1 βœ… AUT ❌ (margin) βœ… ArnautoviΔ‡ #1 (pen!) ~72%
POR vs DRC POR 2-0 (~19%) POR ~69% / D ~21% 1-1 ❌ Draw! ⚠️ (1-1 top-4) ⚠️ Ronaldo ❌ / βœ… Wissa (#3-4) ~50%
GHA vs PAN 1-1 Draw (~15%) GHA ~34% / D ~30% / PAN ~36% 1-0 GHA ⚠️ (coin-flip) βœ…βœ… 1-0 in all 3 top-3 ⚠️ Semenyo via ASSIST ~78%
ENG vs CRO ENG 2-1 (~17%) ENG ~55% / D ~26% 4-2 βœ… ENG ❌ (margin) βœ…βœ… Kane #1 (brace!) + Bellingham ~68%
UZB vs COL β›” NO PRE-MATCH β€” 1-3 COL β›” blind β›” blind β›” blind N/A

MD2 Groups A-H (Jun 18-21) β€” 16 matches

Match Prediction Win % Actual Winner Score Hit Who Scores Acc
MEX vs KOR MEX ~46% / D ~30% MEX 1-0 modal 1-0 βœ… MEX βœ…βœ… (1-0 modal!) βœ… Romo ~78%
CZE vs RSA 1-1 Draw (~15%) CZE ~39% / D ~33% 1-1 βœ… DRAW HIT βœ…βœ… (1-1 modal!) ❌ Schick / βœ… SadΓ­lek (rotated CM) ~80%
CAN vs QAT CAN ~48% CAN 2-0/2-1 modal 6-0 βœ… CAN ❌ (blowout) βœ… Larin + David ~58%
SUI vs BiH SUI win SUI 2-0 modal 4-1 βœ… SUI ❌ (margin) βœ… Amdouni ~62%
BRA vs HAI BRA ~72% BRA 3-0 modal 3-0 βœ… BRA βœ… (3-0 modal!) βœ…βœ… Cunha #1 (brace!) + Vini Jr ~88%
SCO vs MAR SCO win? SCO ~fav 0-1 ❌ MAR! ❌ βœ… Saibari #1 (2' fastest WC goal!) ~82%
USA vs AUS USA ~55% USA 2-1 modal 2-0 βœ… USA βœ… (2-0 in range) ⚠️ Burgess OG + Freeman ~75%
TUR vs PAR TUR win TUR ~fav 0-1 ❌ PAR! (upset) ❌ βœ… Galarza #1 (2'!) ~55%
GER vs CIV GER ~64% GER 2-1 modal 2-1 βœ… GER βœ…βœ… (2-1 modal!) βœ… Undav #1 (brace, 68'/90+4') ~82%
ECU vs CUR ECU ~fav ECU 2-0 modal 0-0 ❌ DRAW! ❌ β€” (Room 15 saves WC record) ~45% (biggest upset)
NED vs SWE NED ~fav NED 2-1 modal 5-1 βœ… NED ❌ (margin way off) βœ…βœ… Brobbey + Gakpo (both braces!) ~55%
TUN vs JPN JPN ~fav JPN 2-0 modal 0-4 βœ… JPN ❌ (margin) βœ… Ueda #1 (brace!) + Kamada #2 ~55%
BEL vs IRN BEL ~61% BEL 2-0/1-1 co-modal 0-0 ⚠️ DRAW (#2) ❌ (1-1 was co-modal, not 0-0) ❌ Lukaku #1 (0G, subbed) / Beiranvand MOTM ~75%
NZL vs EGY EGY ~57% EGY 1-0/2-0 co-modal 1-3 βœ… EGY ❌ (margin) ⚠️ Marmoush #1 ❌ / βœ… Salah (#2, 1G+1A) ~72%
ESP vs KSA ESP ~85% ESP 3-0 modal (Finding 22) 4-0 βœ… ESP βœ… (blowout band!) βœ… Yamal #1 (first WC goal!) + Oyarzabal #2 ~85%
URU vs CPV URU ~62% URU 1-0 modal 2-2 ⚠️ DRAW (#2, ~26%) ❌ βœ…βœ… Maxi AraΓΊjo #1 (1G+1A!) / CPV first-ever WC goals ~80%

MD2 Groups I-J (Jun 22) β€” 4 matches

Match Prediction Win % Actual Winner Score Hit Who Scores Acc
ARG vs AUT ARG 2-0 (modal) ARG ~68% / D ~23% 2-0 βœ… ARG βœ…βœ… (2-0 modal!) βœ…βœ…βœ… Messi #1 (brace! 18 WC goals, all-time top scorer) ~92% A (RAPID)
FRA vs IRA FRA 2-0 (modal) FRA ~72% / D ~21% 3-0 βœ… FRA ⚠️ (+1G margin) βœ…βœ… MbappΓ© #1 (brace! tied Klose 16) ~80% B
NOR vs SEN NOR 2-1 (modal) NOR ~60% / D ~24% 3-2 βœ… NOR ⚠️ (+1G each side) βœ…βœ… Haaland #1 (brace! 12-game streak) ~85% B
ALG vs JOR ALG 2-1 (modal) ALG ~59% / D ~26% 2-1 βœ… ALG βœ…βœ… (2-1 modal!) ⚠️ Mahrez #1 (assist only) / βœ… Gouiri alt (winner) ~88% A

MD2 Groups K-L (Jun 23) β€” 4 matches ⚑ TASK-99 independent resolution

Match Prediction Win % Actual Winner Score Hit Who Scores Acc
POR vs UZB POR 2-0 (modal) POR ~75% / D ~19% 5-0 βœ… POR ❌ (blowout, g4 F22 right) βœ…βœ… Ronaldo #1 (brace! FIRST to score at 6 WCs) ~80% B
ENG vs GHA ENG 2-0 (modal) ENG ~65% / D ~26% 0-0 ⚠️ DRAW (#2) ❌ (0-0 not modal) ❌ Kane #1 (0G, missed 86' sitter) ~45% F (BIG MISS)
CRO vs PAN CRO 2-0/1-0 (modal) CRO ~60% (unan.) / D ~26% 1-0 βœ… CRO βœ… (1-0 modal for mm3/g4!) ❌ Musa #1 (subbed HT) / βœ… Budimir sub (F19) ~68% C
COL vs DRC COL 1-0 (modal) COL ~59% / D ~29% 1-0 βœ… COL βœ…βœ… (1-0 modal!) ❌ DΓ­az #1 (0G, 2 disallowed) / βœ… MuΓ±oz alt ~82% B

Cumulative Accuracy (72 matches, group stage complete)

Component Hit Rate Grade
Winner prediction (incl. draw in top-3) ~48/72 (67%) 🟒 Good
Winner prediction (strict modal only) ~36/72 (50%) 🟑 Moderate
Exact scoreline (top 3) ~25/72 (35%) 🟑 Stable
Who Scores β‰₯1 hit ~34/50 (68%) 🟒 Good
Who Scores #1 hit ~22/50 (44%) 🟑 Sub risk + upsets
Referee red cards 100% 🟒 Perfect
Yellow card count ~50% range hit 🟑 Over-predicts
Ensemble calibration avg ~70% (42 rubric-scored) 🟒 Good

Draws: 19 of 72 group-stage matches ended in draws (26%) β€” the draw rate declined through the tournament (MD1 33% β†’ MD2 21% β†’ MD3 17%). Pipeline now correctly prices draw band 25-33% for low-block matchups.


The Draw Epidemic (Finding 1, amplified) β€” 16 draws in 47 matches (34%)

MD1 draws (8): CAN 1-1 BiH Β· QAT 1-1 SUI Β· BRA 1-1 MAR Β· NED 2-2 JPN Β· ESP 0-0 CPV Β· BEL 1-1 EGY Β· KSA 1-1 URU Β· IRN 2-2 NZL
MD2 draws (8): CZE 1-1 RSA Β· ECU 0-0 CUR Β· BEL 0-0 IRN Β· URU 2-2 CPV Β· POR 1-1 DRC (MD1 carryover) Β· ENG 0-0 GHA (K-L)

Finding 13 (sterile-domination) β€” 3 consecutive validations (POR-DRC, ESP-CPV-style, BEL-IRN): Tier 1 favorite + low block underdog + hot GK = draw. This is now structural, not noise. Draw band should be 25-30% baseline for these matchups.

Finding 22 blowout EXCLUSION β€” validated twice (ECU-CUR 0-0, URU-CPV 2-2): when underdog plays deep 4-5-1 + veteran GK, favorite can't blow them out even with 1.92 xG. The blowout exclusion works.


Yellow Card Calibration Issue (NEW β€” MD2 finding)

Range hit rate: 11/20 (55%). Avg error: 1.4 yellows. Agents over-predict:


Calibration Findings Summary (22 rules in doc-1)

Key findings validated/extended in MD2:

  1. Draws underpriced β€” 36% draw rate, now calibrated 25-33% for low-block matchups
  2. Blowout underestimate (Finding 22) β€” systematic (GER 7-1, SWE 5-1, NOR 4-1, ESP 4-0, CAN 6-0). When quality gap + elite finisher β†’ 15-20% on 4+ margin. But has LIMIT: deep 4-5-1 + veteran GK excludes blowout (ECU-CUR, URU-CPV)
  3. Tournament referee leniency β€” 20-25% discount confirmed 8+ times. But still over-predicts yellows (actual 2.4 vs predicted 3.0)
  4. Lineup news invalidates the model β€” Yamal benched (MD1), Lukaku subbed, Marmoush subbed 76'
  5. Finding 13 (sterile-domination) ⭐ β€” 3 consecutive validations. Tier 1 + low block + hot GK = draw
  6. Low-block + hot GK (Finding 6) β€” 5+ validations (Vozinha, Beach, Shobeir, Al-Owais, Beiranvand)
  7. Tournament-team escalation ⭐ β€” ESP/GER/BRA/ARG/FRA/ENG/POR/JPN escalate game-to-game (+3-5pts MD2+). Validated: ESP 4-0, GER 2-1, BRA 3-0
  8. Tournament Energy Management (Finding 12) β€” context-dependent. Defending champions REDUCE discount. Manager personality > tier
  9. Sub-scorer rule β€” continues (TrΓ©zΓ©guet 82', Undav, Amad, Barcola, Mbaye). Fresh legs vs tiring defense
  10. Finding 16 age-decline extends to GKs β€” Muslera (39yo) error cost URU vs CPV
  11. NOT tournament teams β€” BEL (0 goals in 2 matches), URU (wasteful, Bielsa leaving), TUR (5 root causes, crisis)
  12. Generational milestone rule (Finding 12 hard rule) ⭐ β€” Messi (18, all-time WC top scorer), MbappΓ© (16, tied Klose), Haaland (12-game streak). Milestone-chasing generational players = reliable #1 scorer picks. 3x validated
  13. Finding 17 resolves via HT subs ⭐ β€” sterile-domination can be unlocked mid-match by aggressive substitutions (ALG-JOR: 0.65 xG HT β†’ 1.25 xG 2H after Bentaleb+Benbouali). Managerial intervention matters
  14. Set-piece as Finding 17 equalizer ⭐ β€” when open play is sterile, set-pieces unlock low blocks (ALG-JOR: 10 corners, both goals from corners). Mahrez-tier delivery overcomes sterile domination
  15. Camp chaos "defused" β‰  resolved ⭐ β€” contract signing defuses the headline but team remains disorganized for ~1 match (SEN NOR-SEN: defensive errors persisted despite Thiaw signing)
  16. Rain (not heat) favors aerial favorite ⭐ β€” Finding 11 rain mechanism is DIFFERENT from heat. Wet pitch amplifies set-piece/aerial edge for physical teams (NOR-SEN MetLife rain β†’ Haaland volley/Pedersen)
  17. Referee assignment must come from CURRENT WC source β›” β€” stale WC2022 career-history grep returned the WRONG ref for ARG-AUT (Makkelie β†’ actual Omar). Bug caught, rule added
  18. GK changes are model-breaking (Finding 14) — Iraq (Hassan→Basil backup, at fault for goal), SEN Mendy injury 62'. Pre-match GK confirmation critical
  19. Generational milestone rule (Finding 12 hard rule) ⭐⭐ β€” Messi (18, all-time WC top scorer), MbappΓ© (16, tied Klose), Haaland (12-game streak), Ronaldo (FIRST to score at 6 WCs). Milestone-chasing generational players = reliable #1 scorer picks. 4x validated. BUT needs a finishing mechanism (weak GK/pen band/mobility swap) β€” milestone alone delivers minutes, not goals (Kane ENG-GHA 86' miss)
  20. Finding 22 blowout needs caliber split ⭐ β€” g4 systematically right on blowouts (POR-UZB 5-0, NOR 4-1, GER 7-1). Weight 4+ margin at 15-20% when: elite attack + non-veteran-GK block + (desperate underdog opening up OR tournament-team MD1-stutterβ†’MD2-statement arc). mm3/g5 too conservative on the tail
  21. GK reversal requires shot-stopping evidence, not cap count ⚠️ (ENG-GHA calibration miss) β€” Asare (domestic GHA, 13 caps) predicted as LIABILITY (lowers draw); delivered MOTM hot-GK heroics that FORCED the 0-0. Domestic-league / low-cap GKs CAN be hot-GKs at a WC. Reversal needs evidence the GK is genuinely poor, not just low-cap
  22. Finding 17 (sterile-domination) fires on favorites too ⚠️ (ENG-GHA) β€” 78.8% poss / 1.28 xG / 0 goals. The "elite finishers" safety valve does NOT guarantee F17 exclusion. Add: favorites can sterile-dominate vs deep 4-5-1 + hot-GK even with Kane/Bellingham-tier finishers

Standings After Group Stage (ALL 12 groups complete)

Group 1st 2nd 3rd 4th
A Mexico 9 βœ… (+6) South Africa 4 (-1) South Korea 3 (-1) Czechia 1 ❌ (-4)
B Switzerland 7 βœ… (+5) Canada 4 (+4) Bosnia 4 (-2) Qatar 1 ❌ (-8)
C Morocco 7 βœ… (+3) Brazil 4 (+1) Scotland 3 (-2) Haiti 0 ❌ (-7)
D USA 6 βœ… (+4) Australia 4 (+1) Paraguay 4 (-2) TΓΌrkiye 3 ❌ (-1)
E Germany 6 (+3) Ivory Coast 6 βœ… (+5) Ecuador 4 (+1) Curacao 1 ❌ (-8)
F Netherlands 7 βœ… (+8) Japan 5 (+5) Sweden 4 (-1) Tunisia 0 ❌ (-10)
G Egypt 5 (+3) Belgium 5 (+4) Iran 3 (-1) New Zealand 1 ❌ (-5)
H Spain 7 βœ… (+4) Cape Verde 3 (0) Saudi Arabia 2 (-3) Uruguay 2 ❌ (-4)
I France 9 βœ… (+9) Norway 6 βœ… (+1) Senegal 3 (-3) Iraq 0 ❌ (-7)
J Argentina 9 βœ… (+7) Austria 4 (0) Algeria 4 (-1) Jordan 0 ❌ (-5)
K Colombia 7 βœ… (+3) Portugal 5 (+4) DR Congo 4 (-1) Uzbekistan 0 ❌ (-9)
L England 7 βœ… (+4) Croatia 6 (0) Ghana 4 (0) Panama 0 ❌ (-4)

Qualified (top-2 + 8 best-3rd): 16 teams through. Eliminated: 8 teams. Round of 32 begins Jun 28.


Next Matches

Date Matches Group Status
Jun 24-27 MD3 (final group matches) All 12 groups πŸ”œ Fixtures in doc-3
Jun 28 Round of 32 begins Knockouts βœ… Match 73 done

R32 β€” Match 73: CAN 1-0 RSA (EustΓ‘quio 90+2')

R32 β€” Match 76: BRA 2-1 JPN (Casemiro 56', Martinelli 90+6' / Sano 29')

R32 β€” Match 74: PAR 1-1 GER (AET, PAR win 4-3 pens) β€” BIGGEST UPSET

R32 β€” Match 75: MAR 1-1 NED (AET, MAR win 3-2 pens) β€” Β§7K TREE EXACT

Match Prediction Win % Actual Winner Score Hit Who Scores Acc
NED vs MAR 1-1β†’ET/pens (modal) NED ~54% advance 1-1 AET, MAR 3-2 pens ❌ MAR! (was #2 option) ⚠️ 1-1 in top-3 but only 13%; modal was NED win ⚠️ Brobbey #1 blanked; Saibari #1 βœ“ (pen) + Gakpo #2 βœ“ ~45% F-D

§7K tree architecturally allowed the draw→ET→pens path but under-weighted it (13%). This is a calibration FAILURE, not a success: all 3 agents had NED as modal 90' pick, none predicted DRAW as #1. "In top-3" with 3 outcomes is structural, not predictive. WDM correctly flagged near-equals (fav_share 0.59) but writers over-adjusted to NED (54% advance) on "KO experience" that didn't materialize. MAR xG 1.47 vs NED 0.33 shows MAR were the better side.

Hot-GK calibration refinement (2nd data point): Bounou β€” 2Γ— WC shootout winner. The Γ—0.25 empirical weight over-discounts confirmed multi-tournament shootout specialists. Split proposed: Γ—0.25 for reputation, Γ—0.5-0.6 for β‰₯1 prior WC shootout win.

Ref leniency refinement: Sampaio 4.8 Y/gm career β†’ 1Y actual. CONMEBOL tournament tolerance compounds the discount multiplicatively. Proposed: 50-60% total card reduction for CONMEBOL refs at WC, not 20-25%.

R32 β€” Match 74: PAR 1-1 GER (AET, PAR win 4-3 pens) β€” BIGGEST UPSET

Match Prediction Win % Actual Winner Score Hit Who Scores Acc
GER vs PAR GER 1-0 (modal) GER ~74% advance 1-1 AET, PAR 4-3 pens ❌ PAR! ❌ consensus modal GER 1-0 wrong (mm3 had 1-1) ⚠️ Undav #1 blanked; Havertz #3 βœ“ + Enciso #2 βœ“ ~52% D

CONMEBOL reactive blueprint (NEW FINDING): GER lost to TWO CONMEBOL reactive blocks in 5 days (ECU MD3 2-1, PAR R32 pens) with the identical template: low block + hot GK + counter pace. Squad value (€775M vs €100M) cannot see this. Proposed: cap favorite advance at ~60% vs confirmed reactive-block + in-form-GK.

Hot-GK calibration revision: Gill had 2 MOTMs in 3 games; the Γ—0.25 empirical weight was too aggressive. Split proposed: Γ—0.25 for reputation-only, Γ—0.5-0.75 for β‰₯2 tournament MOTMs. Gill won the shootout (2 PK saves).

mm3 hero / g5 outlier: mm3 nailed the 1-1 modal; g5 (GER 67%/2-0/80%) was most bullish and most wrong. The Β§7K tree correctly carved PAR as a live 26% advance β€” a genuine coin-flip landed on PAR.

R32 β€” Match 76: BRA 2-1 JPN (Casemiro 56', Martinelli 90+6' / Sano 29')

Match Prediction Win % Actual Winner Score Hit Who Scores Acc
BRA vs JPN BRA 2-1 (modal) BRA ~68% advance 2-1 BRA βœ… BRA βœ…βœ… (modal exact!) ❌ Vini #1 (post) / Martinelli sub (#5, 7%) ~73% C

WDM anchor (first use): squad-value baseline BRA 43% β†’ writers adjusted to 50% β†’ actual BRA win. Anchor compressed inter-writer variance from 12pt to 5pt (honest spread). Slightly under-estimated BRA (~7pt low) but ensemble corrected. Pedigree-exclusion holds.

Β§7K knockout tree (first use): modal path (BRA 2-1 in 90') = exact outcome. ET/pens branches correctly dormant. One calibration note: 25% draw-at-90' was ~10pt too high for clear-favorite KO with non-veteran-GK block β†’ calibrate to 15-18%.

Finding 12 refinement: conservation is conditional-on-leading. Ancelotti-conservative-KO draw-tilt failed because BRA trailed from 29' β€” you can't manage a lead you don't have. Weight conservation-draw-tilt at 0% when favorite is trailing at HT.

R32 β€” Match 86: ARG 3-2 CPV (Messi 29', L.MartΓ­nez 92', Borges 111' OG / Duarte 59', Cabral 103') β€” CPV BLOCK ALMOST SHOCKED THE WORLD

Match Prediction Win % Actual Winner Score Hit Who Scores Acc
ARG vs CPV ARG 2-0 (modal) ARG ~75% advance 3-2 ARG (AET, 90' 1-1) βœ… ARG (AET) ❌ modal 2-0; 90' was 1-1 (23% band fired) βœ… Messi #1 βœ“ (29') / L.MartΓ­nez #2 βœ“ (92') / CPV #1 ❌ ~70% C

CPV came within 8 minutes of the biggest R32 upset. 1-1 at 90', 1-2 up at 103'. The 23% draw band was RIGHT to be elevated (above the 15-18% clear-favorite floor). CPV is NOT a minnow β€” they drew ESP + URU at this venue and now pushed ARG to ET.

F22 blowout exclusion FULLY validated. Predicted 4+ margins at 5-8%; actual was 3-2 AET (closest possible). The CPV block + Vozinha template prevented the blowout completely. The partial exclusion was the right call.

Vozinha confirmed-form hot-GK held up (8 saves). Γ—0.5-0.6 correct. Kept CPV in the match.

CPV scored 2 from 0.47 xG β€” extreme clinical finishing. Counter-attack efficiency at its best. ARG's xG dominance (2.26) was real but they needed ET + an OG.

F11 heat was a factor β€” ARG's press degraded in 33Β°C. The match went to ET in the heat. Tier 1 modulator reduced but didn't eliminate the penalty.

Messi milestone = 5/5+ in R32. Scored 29' (19 WC goals, extends record).

R32 β€” Match 88: EGY 1-1 AUS (Ashour 13' / Hany 55' OG; EGY win 4-2 pens) β€” COIN-FLIP + MODAL 1-1 HIT, SHOBEIR DECISIVE

Match Prediction Win % Actual Winner Score Hit Who Scores Acc
AUS vs EGY 1-1 (modal) AUS ~52% advance 1-1 AET, EGY 4-2 pens ⚠️ EGY (48% upset) βœ…βœ… MODAL 1-1 EXACT (mm3+g5 #1) ❌ both #1 missed (Salah hamstring, Irankunda blanked) / Ashour #3 EGY βœ“ / Hany OG ~72% C

Coin-flip prediction was exactly right. Modal 1-1 HIT (mm3 + g5 both #1). The 28% draw band fired. Advance bet lost (AUS 52%) β€” but the reasoning held: the match was genuinely 50/50, went to pens.

Shobeir shootout edge was THE decisive factor (g5 predicted this). g5 specifically flagged EGY's pens edge via Shobeir (MD3 pen save). EGY won 4-2 on pens. g4/mm3 under-weighted this.

Salah hamstring validated as the #1 mover. Didn't score, limited, subbed before pens. The -3 to -5pt fitness drag was correct. Without it, EGY would've been clearer favorites.

EGY out-xG'd AUS 1.32 to 0.84 β€” EGY were the better side. The coin-flip prediction slightly under-rated EGY.

R32 β€” Match 83: ESP 3-0 AUT (Oyarzabal 36', 89' brace, Porro 66') β€” BLOWOUT UNDERESTIMATE (4th validation)

Match Prediction Win % Actual Winner Score Hit Who Scores Acc
ESP vs AUT ESP 2-1/2-0 (modal) ESP ~66% advance 3-0 ESP (90') βœ… ESP ❌ modal 2-1; 3-0 blowout underestimate ⚠️ Yamal #1 ❌ (subbed 85') / Oyarzabal #2 βœ…βœ… brace (36+89) / Porro surprise (66') ~72% C

Blowout underestimate (Finding 2/22, 4th R32 validation: FRA-SWE, ESP-AUT). Predicted 2-1; actual 3-0. ESP xG 2.80 vs AUT 0.32 (0 SoT). Up-weight 3+ margins to 20-25% vs PRESSING sides (AUT gegenpress); keep exclusion + 8-10% for low-block+veteran-GK sides. Pressing-vs-blocking caliber split now has 4 datapoints.

Who Scores refinement: striker #2 > winger #1. Yamal (winger/creator) blanked; Oyarzabal (#9 finisher) braced. When #1 is a winger and #2 is a pure striker, weight the striker higher.

ESP 0-GA spine validated. AUT 0.32 xG, 0 SoT. Gegenpress fully neutralized by CubarsΓ­-Laporte-Rodri. Gegenpress β‰  CONMEBOL low-block distinction confirmed β€” pressing sides get opened up; low-block sides don't.

R32 β€” Match 85: SUI 2-0 ALG (Embolo 10', Ndoye 46') β€” POSSESSION FLIP 3rd VALIDATION

Match Prediction Win % Actual Winner Score Hit Who Scores Acc
SUI vs ALG 1-1 / SUI 2-1 (modal) SUI ~58% advance 2-0 SUI (90') βœ… SUI ⚠️ modal 1-1/2-1; 2-0 margin within 1 goal but wrong shape (clean sheet) βœ… Embolo #1 βœ“ (10') / Ndoye #3 βœ“ (46') / Mahrez #1 ALG ❌ (F16 boundary) ~80% B

Possession FLIP (Finding 8, 3rd R32 validation: MEX-ECU, USA-BIH, SUI-ALG). ALG had 56% possession; SUI won with 44% + clinical counter β€” the favorite played reactive AGAIN. Update Finding 8: possession flip is a knockout-FAVORITE pattern, not just defensive-underdog. Cap projected KO-favorite possession at ~48-52%.

Mahrez milestone β†’ 0 goals (F16 boundary). Unlike MbappΓ©/Haaland/Kane/Ronaldo (all scored), Mahrez (35) did NOT convert. Milestone rule holds for stars ≀~32 (peak), not 35+ (age-decline tier). At 35+, milestone β†’ minutes not goals.

ALG defense was the swing (as predicted). 7 GA in 3 group games β†’ 2 more from 2.45 xG. The pre-match flag was exactly right. But the coin-flip prediction (fav_share 0.63) was TOO CAUTIOUS β€” SUI won 2-0 comfortably. Learning: defensive fragility (7+ GA) should push fav_share higher than the WDM cap allows.

R32 — Match 84: POR 2-1 CRO (Periőić 53' / Ronaldo 68', G.Ramos 90') — AGE-DECLINE + HEAT COMPOUND VALIDATED

Match Prediction Win % Actual Winner Score Hit Who Scores Acc
POR vs CRO POR 1-0 / 1-1 (modal) POR ~58% advance 2-1 POR (90') βœ… POR ⚠️ modal 1-0; 2-1 margin within 1 goal but <15% βœ… Ronaldo #1 βœ“ (68' eq) / CRO #1 ❌ (PeriΕ‘iΔ‡ scored) ~80% B

WDM honesty validated (3rd time). fav_share was 0.64, not 0.72 β€” and the match WAS close (CRO led 53', POR needed a 90' winner). The €674M squad-value gap did NOT translate to dominance. R32-2 opponent-tier-weighted form was right to compress.

F11 heat + F20 age-decline compound VALIDATED (first R32). CRO's older spine (Modrić, Periőić) faded in 31°C heat — exactly as predicted. New pattern: heat degrades the OLDER team more, not necessarily the underdog. Proposed F11 refinement: add age-modulator (heat favors younger squad).

Ronaldo milestone rule = 5/5 in R32. Scored the equalizer (68'). Now scored in all 5 R32 milestone scenarios.

G.Ramos supersub winner (90'). βˆ’5 deduction (F19 unpredicted sub scorer decided the match). Late winners from subs are the recurring R32 scoring miss.

Livaković hot-GK confirmed-form credit was moot — match didn't reach pens. The ×0.5-0.6 only matters if ET happens.

R32 β€” Match 81: USA 2-0 BIH (Balogun 45', Tillman 82') β€” MODAL EXACT, POSSESSION FLIP

Match Prediction Win % Actual Winner Score Hit Who Scores Acc
USA vs BIH USA 2-0 (modal) USA ~70% advance 2-0 USA (90') βœ… USA βœ…βœ… MODAL EXACT (g4 #1 22%, mm3 #1 14%) βœ… Balogun #1 βœ“ (45') / Tillman #3 βœ“ (82') / DΕΎeko #1 BIH ❌ (F16) ~92% A

Modal scoreline USA 2-0 HIT (g4 #1 pick). The home co-host + low-block-opponent model was exactly right. USA beat PAR 4-1 MD1 (same shape); this time 2-0 but same outcome. MD1-evidence signal validated.

Possession FLIP (Finding 8, 2nd R32 validation). BIH had 52% possession; USA won with 48% + clinical counter. Same as MEX-ECU β€” home favorites with a lead can play reactive. Possession is not a favorite requirement.

F16 DΕΎeko age-decline fully validated (4th data point). 40yo, 0 WC goals, BIH 0.25 xG. Milestone β†’ minutes not goals (holds for 38+ players). BIH's attack was never a threat.

F22 blowout partial. Predicted 3+ margin 10-12%; actual 2-0 (just under). Early breach (Balogun 45') unlocked it; BIH's block limited to 2.

R32 β€” Match 82: BEL 3-2 SEN (Diarra 25', Sarr 51' / Lukaku 86', Tielemans 89', 120'+5' AET) β€” DRAW BAND FIRED, BEL CLINICAL

Match Prediction Win % Actual Winner Score Hit Who Scores Acc
BEL vs SEN BEL 2-1/1-0 (modal) BEL ~61% advance 3-2 BEL (AET, 90' 2-2) βœ… BEL (AET) ❌ modal 90' BEL win; 90' was 2-2 draw (23% band fired) ⚠️ De Bruyne #1 ❌ / Sarr #1 SEN βœ… (51') / Lukaku #5 βœ… (86') ~68% C

The 23% draw band at 90' FIRED β€” match was 2-2 at 90'. The draw mechanism (predicted: BEL sterile + SEN counter) was directionally right but for the WRONG reason: SEN out-xG'd BEL 3.54 to 1.80 β€” SEN was the better chance-creator. BEL won through clinical finishing (3 from 1.80 xG) + Courtois holding in ET. The BEL advance was clinical/lucky, not dominant.

Deviation cap (R32-1) VALIDATED. The cap held BEL at ~61% (not 70%+) despite the €337M squad-value gap. The match confirmed a Tier-1 favorite can be out-created by an AFCON underdog with a real attack (Sarr 4-in-4). The cap prevented over-rating BEL.

Sarr #1 SEN βœ… β€” the hottest scorer delivered (51'). Sarr now 4 in 4 WC games. SEN's Who Scores model was right. BEL scorers were WRONG (De Bruyne #1 missed; Lukaku supersub #5 + Tielemans surprise). BEL scoring is more distributed than the De Bruyne-centric model assumed.

F13 sterile-domination fired for the WRONG reason. Predicted BEL sterile (23 shots/0 vs IRN); actual BEL was clinical (3 from 1.80 xG). SEN was the better team on xG. The draw band was right, the mechanism was wrong.

R32 β€” Match 80: ENG 2-1 DRC (Cipenga 7' / Kane 75', 86' brace) β€” TEXTBOOK A (modal exact + Kane clutch)

Match Prediction Win % Actual Winner Score Hit Who Scores Acc
ENG vs DRC ENG 2-1 (modal) ENG ~78% advance 2-1 ENG (90') βœ… ENG βœ…βœ… MODAL EXACT (mm3/g5 #1) βœ…βœ… Kane #1 brace (75+86) / Wissa #1 ❌ (Cipenga scored) ~93% A

Textbook A β€” the genuine inverse of NED-MAR. Modal scoreline ENG 2-1 hit exactly (mm3 + g5 both had it as #1). Kane #1 pick delivered a brace (75', 86') to turn a 0-1 HT deficit into a 2-1 win. Advance bet won in 90'. All findings validated.

Lineup-checker caught the model-breaker. DRC abandoned their 5-3-2 low block for 4-3-3 β€” retired the upset template. All 3 writers correctly removed the reactive blueprint. DRC scored early (Cipenga 7', the #9 in the open shape β€” wrong scorer predicted but right scoring-direction). ENG squad quality won.

Hot-GK is a SYSTEM product (3rd learning). Mpasi (2Γ— validated hot-GK) made 5 saves but conceded 2 β€” good, not heroic, WITHOUT his low block. The R32-5 confirmed-form weight (Γ—0.5-0.6) was correct: he's a good GK, but the low block was what enabled the 8-save barrages of POR/COL. Learning: hot-GK = block + GK compound, not the GK alone.

Kane milestone rule = 4/4 in R32 (Messi, MbappΓ©, Haaland, Kane). Brace to turn a deficit β†’ clutch goals, not just minutes.

DRC scorer model needs refinement. Wissa #1 unanimous β†’ Cipenga (the 4-3-3 #9) scored. When a team shifts formation, model the NEW front-3 scorers, not the inherited #1.

R32 β€” Match 80: MEX 2-0 ECU (QuiΓ±ones 22', JimΓ©nez 31') β€” HOME FORTRESS, EARLY BLITZ

Match Prediction Win % Actual Winner Score Hit Who Scores Acc
MEX vs ECU 1-1 (modal 90') MEX ~60% advance 2-0 MEX (90') βœ… MEX ❌ modal 1-1; 2-0 was 8-9% (<15%) βœ… JimΓ©nez #1 βœ“ (31') / Plata #1 ❌ (blanked) ~79% B

Home fortress + early blitz. MEX scored both goals EARLY (22', 31') β€” NOT the predicted 60-75' altitude window. Travel fatigue (ECU 9-hr flight delay) hit from the start, not late. MEX won the first half, not the altitude phase.

Possession FLIP (Finding 8). ECU had 57% possession despite being the underdog. MEX (the favorite) played the reactive shape (43% poss + clinical counter) β€” the CONMEBOL blueprint INVERTED. A favorite can use reactive tactics too, especially at altitude with a lead.

R32-3 CONMEBOL cap was CONSERVATIVE. All 3 writers capped MEX advance at ~60%. MEX won in 90' comfortably. The cap protects against over-rating favorites, but with extreme environmental edges (Azteca altitude + home + acclimatization), the cap should flex upward.

Draw band OVER-priced (27% β†’ 0%). Altitude + rain + two-defensive-teams logic pushed the draw band high. MEX's early 2-goal lead killed the draw. Proposed: draw-band-collapse rule when home favorite scores early β€” the draw is conditional on 0-0/1-1 at ~40'.

MEX defense elite (4 clean sheets, 0 GA all tournament). Under-rated β€” MEX didn't need 3 goals, just 1-2 because ECU was never scoring.

R32 β€” Match 79: FRA 3-0 SWE (MbappΓ© 45', 74' brace, Barcola 53') β€” BLOWOUT UNDERESTIMATE

Match Prediction Win % Actual Winner Score Hit Who Scores Acc
FRA vs SWE FRA 2-1 (modal) FRA ~66% advance 3-0 FRA (90') βœ… FRA ❌ modal 2-1; 3-0 blowout underestimate βœ…βœ… MbappΓ© #1 brace (45+74) / Isak #1 ❌ (blanked) ~78% B

Blowout underestimate (Finding 2/22, re-confirmed). Predicted FRA 2-1; actual 3-0. Matches GER 7-1 CUR, NED 5-1 SWE, POR 5-0 β€” elite attack + non-veteran-GK block = margin under-predicted. Weight 15-20% on 4+ margin still under-applied.

F11 heat penalty OVER-APPLIED (headline miss). Pre-match: 87Β°F/30.5Β°C threshold crossed β†’ predicted FRA pressing degrades by 60'. Actual: FRA dominated (61% poss, 3.24 xG), SWE wilted instead. Learning: heat degrades Tier 2-3 pressing teams, NOT Tier 1 tournament teams. Proposed F11 refinement: add a tier modulator (reduce/zero heat penalty for Tier 1 favorites with elite fitness/depth).

Untested GK = SWE-negative (validated, 2nd data point). Widell ZetterstrΓΆm (debut, replaced dropped Nordfeldt) conceded 3 from 3.24 xG β€” no heroics. The NOR-CIV lesson (Nyland) held: untested GK is NOT a hot-GK. Correctly assessed pre-match.

MbappΓ© milestone rule = unstoppable (3rd validation). MbappΓ© brace β†’ now 18 WC goals (tied Messi's all-time record). Milestone motivation β†’ goals not just minutes.

R32 β€” Match 78: NOR 2-1 CIV (Nusa 39', Haaland 86' / Amad Diallo 74') β€” COIN-FLIP HONESTY HELD

Match Prediction Win % Actual Winner Score Hit Who Scores Acc
NOR vs CIV 1-1β†’ET/pens (modal) CIV ~52% advance 2-1 NOR (90') ⚠️ NOR (coin-flip 48%) ❌ modal 1-1; NOR 2-1 was g4 #3 at 13% (<15%) βœ… Haaland #1 βœ“ (86') / CIV #1 PΓ©pΓ©-DiomandΓ© ❌; Amad sub βœ“ ~62% C

Coin-flip honesty WORKED. Ensemble flagged this as 48/52 (NOR/CIV advance) β€” NOR winning sits inside the honest band, not a miss. Deviation cap (R32-1) prevented over-rating CIV on squad value. The advance bet lost (leaned CIV) but the model never pretended certainty β€” the 3pt spread was correct uncertainty signaling.

Strict rubric applied (Jun 30): no "in top-3" inflation. Modal scoreline 1-1 was wrong; NOR 2-1 at 13% is under-weighted, not a hit. Advance pick (CIV 52%) lost β†’ winner scored 13/25, not 25.

Hot-GK under-priced (3rd data point): Nyland β€” 4 saves incl. "save of the tournament" at 90'+7' (Diallo free-kick). The Γ—0.25 reputation-only weight (R32-5) can't see a GK who goes hot IN-MATCH. Nyland had no prior WC shootout pedigree but delivered confirmed hot form when it mattered. Proposed refinement: flag POTENTIAL in-match hot-GK form for any starter in a coin-flip KO (not just prior pedigree). NOR's 7-GA defensive warning was inflated by the MD3 B-team (4-1 FRA) β€” filter rotation matches from the form factor.

Amad super-sub pattern (2nd tournament validation): Amad Diallo (60' sub) scored 74' β€” the SAME Amad who scored the 90' MD1 winner off the bench vs ECU. For teams with a known super-sub, weight the bench-impact scorer higher in Who Scores. CIV's #1 split (PΓ©pΓ©/DiomandΓ©) missed because their most dangerous scorer comes off the bench.

R32 β€” Match 73: CAN 1-0 RSA (EustΓ‘quio 90+2')

Match Prediction Win % Actual Winner Score Hit Who Scores Acc
RSA vs CAN CAN 44% / RSA 32% / ET-PK 24% CAN ~54% advance 1-0 CAN βœ… CAN βœ… (modal #2, 14%) ❌ David #1 blanked; EustΓ‘quio 90+2' ~78% B

Mental-resilience note: RSA's MD1 crowd/occasion regression (2 reds vs co-host MEX at Azteca) did NOT repeat tonight β€” 0 reds, structure held 92 min. Broos's formation evolution (5-3-2 β†’ 4-2-3-1) + Sithole's disciplined return are evidence of mental adaptation. The MD1 meltdown was environment-specific (altitude + opening ceremony + debut pressure), not a structural character flaw.

MD2 K-L wave β€” CLOSED βœ… (4/4 pre + post done, TASK-99 + rubric)

Match Result Calibration Pipeline
POR-UZB 5-0 POR ~80% B βœ… FULL (last pre-TASK-99)
ENG-GHA 0-0 ~45% F (BIG MISS) βœ… FULL TASK-99
CRO-PAN 0-1 CRO ~68% C βœ… FULL TASK-99
COL-DRC 1-0 COL ~82% B βœ… FULL TASK-99

MD2 I-J wave β€” CLOSED βœ…

Match Result Calibration Pipeline
ARG-AUT 2-0 ARG ~92% A (modal exact) ⚠️ RAPID
FRA-IRA 3-0 FRA ~80% B βœ… FULL
NOR-SEN 3-2 NOR ~85% B βœ… FULL
ALG-JOR 2-1 ALG ~88% A (modal exact) βœ… FULL

Per-Agent Calibration (20 matches: 16 R32 + 4 R16)

Strict rubric (Jun 30+). Per-agent, NOT consensus. Main band = 80pts (winner_90 15 + advance 10 + scoreline 25 + findings 20 + model-breaking 10). who_scores tracked separately. Validated by experiments/validate-agent.mjs (auto: 50pts) + calibration-validator agent (qualitative: 30pts). Full detail: sources/matches/{code}/calibration-scores.md.

Match Result mm3 g4 g5
RSA-CAN CAN 1-0 71 B 78 A 50 C
BRA-JPN BRA 2-1 53 C 75 A 75 A
GER-PAR PAR pens 60 B 53 C 48 C
NED-MAR MAR pens 55 C 48 C 60 B
NOR-CIV NOR 2-1 71 B 45 D 56 C
FRA-SWE FRA 3-0 43 D 60 B 58 C
MEX-ECU MEX 2-0 70 B 49 C 46 D
ENG-DRC ENG 2-1 80 A 71 B 80 A
BEL-SEN BEL 3-2 AET 56 C 54 C 54 C
USA-BIH USA 2-0 70 B 75 A 70 B
ESP-AUT ESP 3-0 63 B 63 B 68 B
POR-CRO POR 2-1 68 B 70 B 73 A
SUI-ALG SUI 2-0 60 B 60 B 48 C
ARG-CPV ARG 3-2 AET 67 B 40 D 67 B
AUS-EGY EGY pens 56 C 49 C 58 C
COL-GHA COL 1-0 72 A 76 A 69 B
MAR-CAN MAR 3-0 49 C 59 C 49 C
FRA-PAR FRA 1-0 68 B 53 C 78 A
BRA-NOR NOR 2-1 26 F 48 C 43 D
ENG-MEX ENG 3-2 43 D 62 B 35 F

Agent averages (20 matches):

Agent Model Avg Band Strength
mm3 MiniMax-M3 60/80 (75%) B Best on coin-flips + draw-hedge. Highest findings avg. Most consistent (Οƒ=9.2).
g4 glm-5-turbo (xhigh) 59/80 (74%) B Best on clear favorites + scoreline precision. Most volatile (Οƒ=11.6).
g5 zai/glm-5.2 (xhigh) 59/80 (74%) B Best on model-breaker detection. Most A grades (4).

Key framework insights:

R16 Per-Agent Calibration (in progress)

Match Result mm3 g4 g5 Consensus
MAR-CAN MAR 3-0 49/80 C 59/80 C 49/80 C ~52% D
FRA-PAR FRA 1-0 68/80 B 53/80 C 78/80 A ~93% A
BRA-NOR NOR 2-1 26/80 F 48/80 C 43/80 D ~63% C
ENG-MEX ENG 3-2 43/80 D 62/80 B 35/80 F ~65% C
POR-ESP ESP 1-0 35/80 F 70/80 B 50/80 C ~43% F
BEL-USA BEL 4-1 25/80 F 19/80 F 23/80 F ~25% F

POR-ESP learnings: ESP 0 GA HELD vs Tier 1 (g5's opponent-inflated flag was wrong β€” elite spine CubarsΓ­-Laporte-Rodri-SimΓ³n is genuine). Ronaldo farewell β†’ 0 goals (F16 at 41, not record-chasing). Merino 90'+1' β€” 3rd R16 sub winner (F19 pattern). g4 best (leaned ESP + exact scoreline ESP 1-0).

ENG-MEX learnings: 0-GA defense broken (3 conceded β€” opponent-quality-inflated record vs weak group). ENG won with 33% possession (flip #4). Quansah (RB crisis) scored. 5-goal thriller defied altitude+block low-scoring prediction. g4 best (hedged 0-GA); g5 worst (sterile-domination bet wrong + modal 0-0).

BRA-NOR learnings: H2H curse (NOR unbeaten vs BRA ever) was under-applied (βˆ’2-3pt; should be βˆ’5pt). Haaland brace = milestone 6/6. BRA out-xG'd NOR 3:1 (2.73 vs 0.84) but profligate (4 big chances missed). Storm favoured NOR aerial edge. g4 best (storm-weighting right); mm3 worst (TOML bug + wrong modal direction).

FRA-PAR learnings: R32-3 CONMEBOL reactive blueprint held perfectly (3rd validation β€” FRA won 1-0, exactly the tight capped-favorite margin). MbappΓ© milestone = 19 WC goals (surpassed Messi). F11 heat 99Β°F applied correctly (Tier 1 muted to 1 goal). g5 nailed the modal 1-0 exact (25/25 scoreline). The cap is now the most-validated finding in the framework.

MAR-CAN learnings: R32-3 cap proved conservative (MAR won 3-0 dominantly). Bounou pens thesis mooted. Who Scores missed.


Last updated: July 7, 2026 (R16: 6/8 done). R32 complete. R16 remaining: ARG-EGY. QF bracket: FRA-MAR, NOR-ENG, ESP-vs-ARG/EGY, BEL-vs-COL/SUI.*