Model cards & calibration

These cards report the actual out-of-sample backtest metrics Golavo emits, one card per competition. They are generated from the schema-validated eval_summary*.json artifacts by scripts/build_model_cards.py — never hand-edited — so the numbers here match what CI validates. Log loss is primary. No model is a declared champion; forward evidence (the calibration record) is kept separate from these historical folds.

How to read a card

Each card lists every deterministic candidate evaluated on that competition against the climatological baseline — the five seated families everywhere, plus any club-league candidate on trial in the domestic cards. Skill is 1 - model log loss / baseline log loss; its 95% interval is a seeded, fold-stratified bootstrap over held-out matches. Metrics are out-of-sample on strictly chronological folds. League strengths are not comparable across competitions — each league is modeled independently from its own pack.

Men’s senior full internationals

• Scope: Men’s senior full internationals (forward seal→score surface plus these historical test folds).
• Source snapshot: martj42/international_results ddd7249ac0c2, retrieved 2026-07-10 (CC0-1.0).
• Folds: WC2022, EURO2024, WC2026 — strictly chronological; fitting and decay selection use only rows before each fold’s cutoff.

Competition report cards (positive skill means lower log loss than climatology):

FIFA World Cup report card (2022-11-20 to 2026-07-19):

Model	Matches / folds	Log loss	Skill vs baseline (95% CI)	ECE	Fold rank
climatological (baseline)	161 / 2	1.0636	+0.0% (+0.0% to +0.0%)	0.0310	5.0 (5–5)
Elo ordinal-logit	161 / 2	0.9494	+10.7% (+5.6% to +15.4%)	0.1559	1.0 (1–1)
independent Poisson	161 / 2	1.0027	+5.7% (+0.7% to +10.6%)	0.0972	3.0 (3–3)
time-decayed Dixon-Coles	161 / 2	1.0000	+6.0% (+0.6% to +10.9%)	0.0920	2.0 (2–2)
bivariate Poisson	161 / 2	1.0027	+5.7% (+0.5% to +10.5%)	0.0972	4.0 (4–4)

Skill intervals use 2,000 seeded, fold-stratified match-bootstrap samples.

UEFA Euro report card (2024-06-14 to 2024-07-14):

Model	Matches / folds	Log loss	Skill vs baseline (95% CI)	ECE	Fold rank
climatological (baseline)	51 / 1	1.1422	+0.0% (+0.0% to +0.0%)	0.1377	5.0 (5–5)
Elo ordinal-logit	51 / 1	1.0300	+9.8% (+1.3% to +17.4%)	0.0625	4.0 (4–4)
independent Poisson	51 / 1	1.0228	+10.5% (+3.4% to +17.2%)	0.0887	2.0 (2–2)
time-decayed Dixon-Coles	51 / 1	0.9973	+12.7% (+4.5% to +19.8%)	0.0890	1.0 (1–1)
bivariate Poisson	51 / 1	1.0228	+10.5% (+3.6% to +17.7%)	0.0887	3.0 (3–3)

Skill intervals use 2,000 seeded, fold-stratified match-bootstrap samples.

Log loss by fold (primary metric; lower is better; bold = best in fold):

Model	WC2022	EURO2024	WC2026
climatological (baseline)	1.0742	1.1422	1.0565
Elo ordinal-logit	1.0157	1.0300	0.9055
independent Poisson	1.0677	1.0228	0.9598
time-decayed Dixon-Coles	1.0650	0.9973	0.9571
bivariate Poisson	1.0677	1.0228	0.9598

Every candidate beats the climatological baseline on log loss on every fold; the best model varies by fold and none is crowned a champion.

Calibration — most recent fold (WC2026):

Model	Brier	ECE	RPS
climatological (baseline)	0.6371	0.0158	0.2252
Elo ordinal-logit	0.5318	0.1603	0.1725
independent Poisson	0.5717	0.0876	0.1936
time-decayed Dixon-Coles	0.5706	0.1029	0.1925
bivariate Poisson	0.5717	0.0876	0.1936

Reliability — Elo ordinal-logit on WC2026 (Wilson 95% intervals; empty bins omitted):

Confidence bin	n	Empirical	Wilson 95%
0.3–0.4	17	0.529	[0.31, 0.74]
0.4–0.5	34	0.588	[0.42, 0.74]
0.5–0.6	28	0.786	[0.60, 0.90]
0.6–0.7	15	0.733	[0.48, 0.89]
0.7–0.8	3	0.667	[0.21, 0.94]

English Premier League

• Scope: English Premier League (historical, completed seasons only — not live).
• Source snapshot: openfootball a5dd38b3bcbe, retrieved 2026-07-11 (CC0-1.0).
• Folds: EPL2023-24, EPL2024-25, EPL2025-26 — strictly chronological; fitting and decay selection use only rows before each fold’s cutoff.

Competition report cards (positive skill means lower log loss than climatology):

English Premier League report card (2023-08-01 to 2026-06-30):

Model	Matches / folds	Log loss	Skill vs baseline (95% CI)	ECE	Fold rank
climatological (baseline)	1140 / 3	1.0732	+0.0% (+0.0% to +0.0%)	0.0252	6.0 (6–6)
Elo ordinal-logit	1140 / 3	1.0065	+6.2% (+4.6% to +7.7%)	0.0579	3.7 (1–5)
independent Poisson	1140 / 3	1.0105	+5.8% (+4.0% to +7.6%)	0.0543	2.3 (1–3)
time-decayed Dixon-Coles	1140 / 3	1.0112	+5.8% (+4.0% to +7.5%)	0.0575	3.3 (1–5)
bivariate Poisson	1140 / 3	1.0105	+5.8% (+4.0% to +7.7%)	0.0543	3.3 (2–4)
Dixon-Coles with per-club home advantage and rest days	1140 / 3	1.0115	+5.7% (+3.9% to +7.6%)	0.0679	2.3 (2–3)

Skill intervals use 2,000 seeded, fold-stratified match-bootstrap samples.

Log loss by fold (primary metric; lower is better; bold = best in fold):

Model	EPL2023-24	EPL2024-25	EPL2025-26
climatological (baseline)	1.0551	1.0822	1.0823
Elo ordinal-logit	0.9634	1.0075	1.0487
independent Poisson	0.9494	1.0451	1.0370
time-decayed Dixon-Coles	0.9587	1.0452	1.0298
bivariate Poisson	0.9494	1.0451	1.0370
Dixon-Coles with per-club home advantage and rest days	0.9582	1.0428	1.0336

Every candidate beats the climatological baseline on log loss on every fold; the best model varies by fold and none is crowned a champion.

Calibration — most recent fold (EPL2025-26):

Model	Brier	ECE	RPS
climatological (baseline)	0.6551	0.0224	0.2276
Elo ordinal-logit	0.6322	0.0390	0.2164
independent Poisson	0.6236	0.0438	0.2120
time-decayed Dixon-Coles	0.6183	0.0591	0.2100
bivariate Poisson	0.6236	0.0438	0.2120
Dixon-Coles with per-club home advantage and rest days	0.6215	0.0806	0.2117

Reliability — time-decayed Dixon-Coles on EPL2025-26 (Wilson 95% intervals; empty bins omitted):

Confidence bin	n	Empirical	Wilson 95%
0.3–0.4	71	0.493	[0.38, 0.61]
0.4–0.5	162	0.364	[0.29, 0.44]
0.5–0.6	90	0.544	[0.44, 0.64]
0.6–0.7	48	0.667	[0.53, 0.78]
0.7–0.8	9	0.778	[0.45, 0.94]

La Liga

• Scope: La Liga (historical, completed seasons only — not live).
• Source snapshot: openfootball a5dd38b3bcbe, retrieved 2026-07-11 (CC0-1.0).
• Folds: LALIGA2022-23, LALIGA2023-24, LALIGA2025-26 — strictly chronological; fitting and decay selection use only rows before each fold’s cutoff.

Competition report cards (positive skill means lower log loss than climatology):

La Liga report card (2022-08-01 to 2026-06-30):

Model	Matches / folds	Log loss	Skill vs baseline (95% CI)	ECE	Fold rank
climatological (baseline)	1140 / 3	1.0590	+0.0% (+0.0% to +0.0%)	0.0241	6.0 (6–6)
Elo ordinal-logit	1140 / 3	0.9994	+5.6% (+4.1% to +7.2%)	0.0663	5.0 (5–5)
independent Poisson	1140 / 3	0.9830	+7.2% (+5.3% to +9.0%)	0.0457	2.0 (1–3)
time-decayed Dixon-Coles	1140 / 3	0.9840	+7.1% (+5.1% to +9.0%)	0.0533	2.3 (1–3)
bivariate Poisson	1140 / 3	0.9830	+7.2% (+5.3% to +9.0%)	0.0457	2.3 (1–4)
Dixon-Coles with per-club home advantage and rest days	1140 / 3	0.9876	+6.7% (+4.7% to +8.7%)	0.0394	3.3 (2–4)

Skill intervals use 2,000 seeded, fold-stratified match-bootstrap samples.

Log loss by fold (primary metric; lower is better; bold = best in fold):

Model	LALIGA2022-23	LALIGA2023-24	LALIGA2025-26
climatological (baseline)	1.0519	1.0767	1.0484
Elo ordinal-logit	1.0006	1.0030	0.9946
independent Poisson	0.9863	0.9743	0.9883
time-decayed Dixon-Coles	0.9936	0.9694	0.9891
bivariate Poisson	0.9863	0.9743	0.9883
Dixon-Coles with per-club home advantage and rest days	0.9986	0.9725	0.9917

Every candidate beats the climatological baseline on log loss on every fold; the best model varies by fold and none is crowned a champion.

Calibration — most recent fold (LALIGA2025-26):

Model	Brier	ECE	RPS
climatological (baseline)	0.6314	0.0309	0.2232
Elo ordinal-logit	0.5915	0.0711	0.2049
independent Poisson	0.5870	0.0600	0.2032
time-decayed Dixon-Coles	0.5874	0.0578	0.2033
bivariate Poisson	0.5870	0.0600	0.2032
Dixon-Coles with per-club home advantage and rest days	0.5892	0.0557	0.2040

Reliability — bivariate Poisson on LALIGA2025-26 (Wilson 95% intervals; empty bins omitted):

Confidence bin	n	Empirical	Wilson 95%
0.3–0.4	72	0.333	[0.24, 0.45]
0.4–0.5	143	0.483	[0.40, 0.56]
0.5–0.6	95	0.484	[0.39, 0.58]
0.6–0.7	40	0.775	[0.62, 0.88]
0.7–0.8	27	0.852	[0.68, 0.94]
0.8–0.9	3	1.000	[0.44, 1.00]

Bundesliga

• Scope: Bundesliga (historical, completed seasons only — not live).
• Source snapshot: openfootball a5dd38b3bcbe, retrieved 2026-07-11 (CC0-1.0).
• Folds: BUNDESLIGA2023-24, BUNDESLIGA2024-25, BUNDESLIGA2025-26 — strictly chronological; fitting and decay selection use only rows before each fold’s cutoff.

Competition report cards (positive skill means lower log loss than climatology):

Bundesliga report card (2023-08-01 to 2026-06-30):

Model	Matches / folds	Log loss	Skill vs baseline (95% CI)	ECE	Fold rank
climatological (baseline)	918 / 3	1.0797	+0.0% (+0.0% to +0.0%)	0.0302	6.0 (6–6)
Elo ordinal-logit	918 / 3	1.0144	+6.0% (+4.5% to +7.7%)	0.0405	3.0 (1–5)
independent Poisson	918 / 3	1.0085	+6.6% (+4.6% to +8.6%)	0.0666	2.7 (1–4)
time-decayed Dixon-Coles	918 / 3	1.0104	+6.4% (+4.4% to +8.4%)	0.0625	3.0 (2–4)
bivariate Poisson	918 / 3	1.0085	+6.6% (+4.7% to +8.6%)	0.0666	3.7 (2–5)
Dixon-Coles with per-club home advantage and rest days	918 / 3	1.0104	+6.4% (+4.4% to +8.4%)	0.0698	2.7 (1–5)

Skill intervals use 2,000 seeded, fold-stratified match-bootstrap samples.

Log loss by fold (primary metric; lower is better; bold = best in fold):

Model	BUNDESLIGA2023-24	BUNDESLIGA2024-25	BUNDESLIGA2025-26
climatological (baseline)	1.0754	1.0930	1.0706
Elo ordinal-logit	1.0264	1.0237	0.9931
independent Poisson	1.0197	1.0353	0.9704
time-decayed Dixon-Coles	1.0287	1.0338	0.9687
bivariate Poisson	1.0197	1.0353	0.9704
Dixon-Coles with per-club home advantage and rest days	1.0369	1.0322	0.9621

Every candidate beats the climatological baseline on log loss on every fold; the best model varies by fold and none is crowned a champion.

Calibration — most recent fold (BUNDESLIGA2025-26):

Model	Brier	ECE	RPS
climatological (baseline)	0.6478	0.0095	0.2314
Elo ordinal-logit	0.5916	0.0767	0.2039
independent Poisson	0.5740	0.0628	0.1957
time-decayed Dixon-Coles	0.5734	0.0593	0.1956
bivariate Poisson	0.5740	0.0628	0.1957
Dixon-Coles with per-club home advantage and rest days	0.5697	0.0476	0.1938

Reliability — Dixon-Coles with per-club home advantage and rest days on BUNDESLIGA2025-26 (Wilson 95% intervals; empty bins omitted):

Confidence bin	n	Empirical	Wilson 95%
0.3–0.4	60	0.400	[0.29, 0.53]
0.4–0.5	108	0.426	[0.34, 0.52]
0.5–0.6	69	0.551	[0.43, 0.66]
0.6–0.7	45	0.800	[0.66, 0.89]
0.7–0.8	16	0.938	[0.72, 0.99]
0.8–0.9	8	0.750	[0.41, 0.93]

Serie A

• Scope: Serie A (historical, completed seasons only — not live).
• Source snapshot: openfootball a5dd38b3bcbe, retrieved 2026-07-11 (CC0-1.0).
• Folds: SERIEA2022-23, SERIEA2023-24, SERIEA2025-26 — strictly chronological; fitting and decay selection use only rows before each fold’s cutoff.

Competition report cards (positive skill means lower log loss than climatology):

Serie A report card (2022-08-01 to 2026-06-30):

Model	Matches / folds	Log loss	Skill vs baseline (95% CI)	ECE	Fold rank
climatological (baseline)	1140 / 3	1.0855	+0.0% (+0.0% to +0.0%)	0.0214	6.0 (6–6)
Elo ordinal-logit	1140 / 3	1.0108	+6.9% (+5.4% to +8.3%)	0.0439	2.7 (1–5)
independent Poisson	1140 / 3	1.0085	+7.1% (+5.2% to +9.0%)	0.0398	3.0 (3–3)
time-decayed Dixon-Coles	1140 / 3	1.0057	+7.3% (+5.4% to +9.2%)	0.0398	1.7 (1–2)
bivariate Poisson	1140 / 3	1.0085	+7.1% (+5.2% to +9.0%)	0.0398	4.0 (4–4)
Dixon-Coles with per-club home advantage and rest days	1140 / 3	1.0071	+7.2% (+5.2% to +9.1%)	0.0460	3.7 (1–5)

Skill intervals use 2,000 seeded, fold-stratified match-bootstrap samples.

Log loss by fold (primary metric; lower is better; bold = best in fold):

Model	SERIEA2022-23	SERIEA2023-24	SERIEA2025-26
climatological (baseline)	1.0792	1.0880	1.0892
Elo ordinal-logit	1.0040	1.0182	1.0102
independent Poisson	1.0120	1.0029	1.0106
time-decayed Dixon-Coles	1.0091	0.9985	1.0096
bivariate Poisson	1.0120	1.0029	1.0106
Dixon-Coles with per-club home advantage and rest days	1.0136	0.9967	1.0109

Every candidate beats the climatological baseline on log loss on every fold; the best model varies by fold and none is crowned a champion.

Calibration — most recent fold (SERIEA2025-26):

Model	Brier	ECE	RPS
climatological (baseline)	0.6609	0.0397	0.2341
Elo ordinal-logit	0.6042	0.0401	0.2063
independent Poisson	0.6048	0.0386	0.2064
time-decayed Dixon-Coles	0.6042	0.0193	0.2063
bivariate Poisson	0.6048	0.0386	0.2064
Dixon-Coles with per-club home advantage and rest days	0.6048	0.0358	0.2067

Reliability — time-decayed Dixon-Coles on SERIEA2025-26 (Wilson 95% intervals; empty bins omitted):

Confidence bin	n	Empirical	Wilson 95%
0.3–0.4	69	0.304	[0.21, 0.42]
0.4–0.5	134	0.448	[0.37, 0.53]
0.5–0.6	104	0.548	[0.45, 0.64]
0.6–0.7	57	0.649	[0.52, 0.76]
0.7–0.8	16	0.875	[0.64, 0.97]

Ligue 1

• Scope: Ligue 1 (historical, completed seasons only — not live).
• Source snapshot: openfootball a5dd38b3bcbe, retrieved 2026-07-11 (CC0-1.0).
• Folds: LIGUE1-2022-23, LIGUE1-2023-24, LIGUE1-2024-25 — strictly chronological; fitting and decay selection use only rows before each fold’s cutoff.

Competition report cards (positive skill means lower log loss than climatology):

Ligue 1 report card (2022-08-01 to 2025-06-30):

Model	Matches / folds	Log loss	Skill vs baseline (95% CI)	ECE	Fold rank
climatological (baseline)	992 / 3	1.0734	+0.0% (+0.0% to +0.0%)	0.0298	6.0 (6–6)
Elo ordinal-logit	992 / 3	1.0198	+5.0% (+3.4% to +6.5%)	0.0485	3.7 (1–5)
independent Poisson	992 / 3	1.0123	+5.7% (+3.7% to +7.6%)	0.0543	1.3 (1–2)
time-decayed Dixon-Coles	992 / 3	1.0148	+5.5% (+3.6% to +7.4%)	0.0437	3.3 (3–4)
bivariate Poisson	992 / 3	1.0123	+5.7% (+3.9% to +7.6%)	0.0543	2.3 (2–3)
Dixon-Coles with per-club home advantage and rest days	992 / 3	1.0174	+5.2% (+3.2% to +7.2%)	0.0561	4.3 (4–5)

Skill intervals use 2,000 seeded, fold-stratified match-bootstrap samples.

Log loss by fold (primary metric; lower is better; bold = best in fold):

Model	LIGUE1-2022-23	LIGUE1-2023-24	LIGUE1-2024-25
climatological (baseline)	1.0747	1.0917	1.0535
Elo ordinal-logit	1.0190	1.0445	0.9960
independent Poisson	1.0216	1.0337	0.9794
time-decayed Dixon-Coles	1.0229	1.0349	0.9844
bivariate Poisson	1.0216	1.0337	0.9794
Dixon-Coles with per-club home advantage and rest days	1.0230	1.0397	0.9883

Every candidate beats the climatological baseline on log loss on every fold; the best model varies by fold and none is crowned a champion.

Calibration — most recent fold (LIGUE1-2024-25):

Model	Brier	ECE	RPS
climatological (baseline)	0.6363	0.0302	0.2358
Elo ordinal-logit	0.5938	0.0671	0.2147
independent Poisson	0.5820	0.0591	0.2099
time-decayed Dixon-Coles	0.5853	0.0626	0.2103
bivariate Poisson	0.5820	0.0591	0.2099
Dixon-Coles with per-club home advantage and rest days	0.5877	0.0660	0.2114

Reliability — independent Poisson on LIGUE1-2024-25 (Wilson 95% intervals; empty bins omitted):

Confidence bin	n	Empirical	Wilson 95%
0.3–0.4	40	0.425	[0.29, 0.58]
0.4–0.5	143	0.483	[0.40, 0.56]
0.5–0.6	76	0.645	[0.53, 0.74]
0.6–0.7	35	0.714	[0.55, 0.84]
0.7–0.8	10	0.700	[0.40, 0.89]
0.8–0.9	2	1.000	[0.34, 1.00]

Promotion criteria for challengers

A black-box challenger (e.g. gradient boosting on engineered features, including Dixon-Coles outputs) may be considered only after: (1) at least two full forward seasons of evaluation, (2) better RPS and log loss (paired bootstrap, p < 0.05), (3) no calibration regression, and (4) a feature-attribution audit. Until then it stays a lab exhibit, not a shipped model.

Full method, leakage controls, and references: Prediction methodology.