PREDICTION RATIONALE — South Korea vs Czechia, WC2026 Group A (Estadio Akron, 2026-06-12 02:00 UTC)

1. DATA. Last 10 internationals scraped from Sofascore for Korea
   (record WDLWWWLLWW), last 10 for Czechia (record WDDLWWWWWW).
   Korea: 1.50 GF / 1.20 GA per match, xG 1.4 / xGA 1.17,
   possession 54.5%, pass acc 86.0%.
   Czechia: 2.70 GF / 1.30 GA, xG 1.6 / xGA 0.89,
   possession 49.0%, pass acc 75.5%.
   Caveat: raw means are badly distorted by schedule — Czechia's 2.70 GF/game
   includes 6-0 Gibraltar, 1-0 San Marino and 3-1 Guatemala, while they LOST
   1-2 to the Faroe Islands and were out-xG'd 0.46-1.96 by Denmark; Korea's
   sample includes Brazil (0-5) and Cote d'Ivoire (0-4) alongside 5-0 vs
   Trinidad. Strength-of-schedule correction is therefore essential.

2. RATINGS (opponent-strength adjusted). Each match performance
   (0.6*xG + 0.4*goals) is multiplied by an opponent-strength tier for attack
   (scoring vs Brazil counts ~3.4x scoring vs San Marino) and divided by it
   for defense (conceding twice to the Faroes is punished). Tier table logged
   in raw JSON. Result:
   Korea atk 1.28, def 1.00; Czechia atk 1.57, def 1.17.
   lambda_raw(KOR) = atk_KOR * def_CZE/1.30 * 1.06 = 1.22
   lambda_raw(CZE) = atk_CZE * def_KOR/1.30 * 0.93 = 1.13
   85% weight on model, 15% shrink to baseline -> lambda KOR 1.23, CZE 1.15.

3. CONTEXT ADJUSTMENTS. Stadium elevation 1675 m. Korea ran an altitude camp
   in Utah and has been resident in Guadalajara since June 5 with a 10-minute
   commute (quasi-home conditions, large diaspora support expected): +6%.
   Czechia trains at sea level in Mansfield TX and flies in for the match;
   altitude effects bite hardest after 60', their physical pressing style is
   aerobically expensive, mitigated partly by a set-piece-heavy attack: -7%.
   Weather at kickoff (20.7C, 80% RH, 63% rain probability) is mild;
   possible evening storm adds variance but no directional edge.

4. SCORELINE MODEL. Karlis-Ntzoufras bivariate Poisson with covariance
   lambda3 = 0.12 (typical positive goal correlation in internationals),
   truncated at 8 goals and renormalized. 1X2: KOR 37.3% /
   draw 29.6% / CZE 33.2%. Over 2.5 42.2%, BTTS 49.6%.
   WC openers between evenly-matched sides skew cagey; shrinkage handles this.

5. EVENTS.
   - Red card: referee Amin Omar carries 0.20 reds/match over 115 games, but
     much of that sample is CAF/Egyptian league football where reds run higher;
     blended 50/50 with the recent World Cup group-stage base rate (~0.13) ->
     Poisson rate 0.165 -> P(any red) = 15.2%. His 16.9 fouls/match and 3.86
     yellows/match suggest he lets little go; Czechia's physical duel-heavy
     style and Korea's transition pace raise card volume, supporting the
     career-leaning blend.
   - Corners: expected total 9.4 from both sides' for/against averages;
     Czech wingback crossing game inflates their corner generation. Poisson on
     the total (overdispersion acknowledged -> O/U 8.5 probabilities are
     slightly overconfident toward the mean).
     player shares = positional prior + 0.7 * scraped goals in last 10
     internationals, normalized over the probable XI (Al Jazeera). Son Heung-min
     (talisman, penalties) and Patrik Schick (focal striker, elite header,
     set-piece-fed) head the list.

6. LIMITATIONS. 10-match samples are small; no explicit strength-of-schedule
   model; corners treated as Poisson; lineups are probable not confirmed;
   altitude/fatigue multipliers are expert priors, not fitted parameters.