Reliability & Calibration

Model Reliability

One model set is trained per league group (at most two leagues each), so each is judged separately. A group that beats its baseline is worth betting; one that does not, is not — regardless of how the others do.

Out-of-sample by league group (the numbers that count)

Central

Trained on 4744 matches before 2024-11-08; scored on 1187 after.

Market	Test matches	Accuracy	Baseline	Brier	Verdict
Match result	1187	51.0%	45.1% (always H)	0.6088	beats baseline
Over / Under 2.5 goals	1187	53.8%	53.4% (always Y)	0.4953	beats baseline
Both teams to score	1187	54.5%	54.5% (always Y)	0.4964	no better

China

Trained on 3026 matches before 2025-04-15; scored on 757 after.

Market	Test matches	Accuracy	Baseline	Brier	Verdict
Match result	757	48.3%	45.3% (always H)	0.6236	beats baseline
Over / Under 2.5 goals	757	57.2%	58.0% (always Y)	0.4837	no better
Both teams to score	757	61.6%	61.6% (always Y)	0.4795	no better

Denmark

Trained on 2812 matches before 2024-08-18; scored on 704 after.

Market	Test matches	Accuracy	Baseline	Brier	Verdict
Match result	704	44.9%	47.0% (always H)	0.6315	no better
Over / Under 2.5 goals	704	57.0%	58.5% (always Y)	0.4901	no better
Both teams to score	704	58.7%	58.9% (always Y)	0.4883	no better

England

Trained on 8511 matches before 2025-04-05; scored on 2000 after.

Market	Test matches	Accuracy	Baseline	Brier	Verdict
Match result	2000	44.0%	43.2% (always H)	0.6418	beats baseline
Over / Under 2.5 goals	2000	51.4%	52.8% (always Y)	0.5041	no better
Both teams to score	2000	50.3%	54.3% (always Y)	0.5013	no better

Finland

Trained on 2304 matches before 2024-05-05; scored on 576 after.

Market	Test matches	Accuracy	Baseline	Brier	Verdict
Match result	576	48.6%	45.0% (always H)	0.6192	beats baseline
Over / Under 2.5 goals	576	43.6%	56.4% (always Y)	0.5096	no better
Both teams to score	576	44.6%	55.4% (always Y)	0.5008	no better

France

Trained on 4589 matches before 2025-02-07; scored on 1148 after.

Market	Test matches	Accuracy	Baseline	Brier	Verdict
Match result	1148	48.0%	44.2% (always H)	0.6217	beats baseline
Over / Under 2.5 goals	1148	52.5%	51.5% (always Y)	0.4961	beats baseline
Both teams to score	1148	52.8%	52.7% (always Y)	0.499	beats baseline

Germany

Trained on 4812 matches before 2025-02-01; scored on 1204 after.

Market	Test matches	Accuracy	Baseline	Brier	Verdict
Match result	1204	43.8%	41.0% (always H)	0.6384	beats baseline
Over / Under 2.5 goals	1204	59.6%	59.6% (always Y)	0.4816	no better
Both teams to score	1204	58.6%	58.6% (always Y)	0.4854	no better

Iberia East

Trained on 3534 matches before 2025-05-10; scored on 884 after.

Market	Test matches	Accuracy	Baseline	Brier	Verdict
Match result	884	48.1%	40.8% (always H)	0.6012	beats baseline
Over / Under 2.5 goals	884	52.7%	50.0% (always N)	0.4975	beats baseline
Both teams to score	884	51.0%	51.9% (always N)	0.4997	no better

Ireland

Trained on 2175 matches before 2023-09-25; scored on 544 after.

Market	Test matches	Accuracy	Baseline	Brier	Verdict
Match result	544	43.8%	40.6% (always H)	0.644	beats baseline
Over / Under 2.5 goals	544	54.6%	54.6% (always N)	0.4975	no better
Both teams to score	544	47.6%	52.4% (always Y)	0.5063	no better

Italy

Trained on 5280 matches before 2025-02-23; scored on 1320 after.

Market	Test matches	Accuracy	Baseline	Brier	Verdict
Match result	1320	47.4%	41.9% (always H)	0.6222	beats baseline
Over / Under 2.5 goals	1320	52.4%	53.6% (always N)	0.4986	no better
Both teams to score	1320	50.8%	50.8% (always Y)	0.501	no better

Japan

Trained on 3750 matches before 2023-11-25; scored on 938 after.

Market	Test matches	Accuracy	Baseline	Brier	Verdict
Match result	938	44.7%	41.7% (always H)	0.6391	beats baseline
Over / Under 2.5 goals	938	52.0%	51.5% (always N)	0.4993	beats baseline
Both teams to score	938	51.4%	51.4% (always Y)	0.5001	no better

Low Countries

Trained on 3833 matches before 2025-04-27; scored on 959 after.

Market	Test matches	Accuracy	Baseline	Brier	Verdict
Match result	959	48.3%	41.3% (always H)	0.6159	beats baseline
Over / Under 2.5 goals	959	53.8%	53.8% (always Y)	0.4987	no better
Both teams to score	959	55.7%	55.7% (always Y)	0.4947	no better

Mexico

Trained on 3817 matches before 2024-01-18; scored on 955 after.

Market	Test matches	Accuracy	Baseline	Brier	Verdict
Match result	955	47.4%	47.0% (always H)	0.6242	beats baseline
Over / Under 2.5 goals	955	47.3%	54.7% (always Y)	0.5038	no better
Both teams to score	955	56.5%	56.5% (always Y)	0.4919	no better

Norway

Trained on 3909 matches before 2024-04-21; scored on 978 after.

Market	Test matches	Accuracy	Baseline	Brier	Verdict
Match result	978	47.3%	47.0% (always H)	0.6132	beats baseline
Over / Under 2.5 goals	978	60.0%	60.0% (always Y)	0.4824	no better
Both teams to score	978	55.7%	55.7% (always Y)	0.4939	no better

Poland

Trained on 3373 matches before 2024-03-09; scored on 844 after.

Market	Test matches	Accuracy	Baseline	Brier	Verdict
Match result	844	45.7%	45.0% (always H)	0.6392	beats baseline
Over / Under 2.5 goals	844	49.6%	51.9% (always Y)	0.5012	no better
Both teams to score	844	55.0%	57.3% (always Y)	0.496	no better

Romania

Trained on 3446 matches before 2024-02-29; scored on 862 after.

Market	Test matches	Accuracy	Baseline	Brier	Verdict
Match result	862	45.2%	42.9% (always H)	0.6373	beats baseline
Over / Under 2.5 goals	862	57.9%	54.6% (always N)	0.4908	beats baseline
Both teams to score	862	49.8%	50.7% (always Y)	0.5013	no better

Russia

Trained on 3077 matches before 2023-09-16; scored on 770 after.

Market	Test matches	Accuracy	Baseline	Brier	Verdict
Match result	770	51.9%	43.5% (always H)	0.6073	beats baseline
Over / Under 2.5 goals	770	52.6%	52.6% (always N)	0.5045	no better
Both teams to score	770	47.4%	52.7% (always Y)	0.5082	no better

Scotland

Trained on 1596 matches before 2025-08-03; scored on 399 after.

Market	Test matches	Accuracy	Baseline	Brier	Verdict
Match result	399	47.4%	42.9% (always H)	0.632	beats baseline
Over / Under 2.5 goals	399	49.4%	50.6% (always N)	0.503	no better
Both teams to score	399	49.4%	50.6% (always N)	0.5009	no better

South America

Trained on 11364 matches before 2024-09-14; scored on 2000 after.

Market	Test matches	Accuracy	Baseline	Brier	Verdict
Match result	2000	45.9%	45.4% (always H)	0.631	beats baseline
Over / Under 2.5 goals	2000	60.3%	60.3% (always N)	0.4775	no better
Both teams to score	2000	54.0%	54.0% (always N)	0.4962	no better

Spain

Trained on 6088 matches before 2025-04-13; scored on 1523 after.

Market	Test matches	Accuracy	Baseline	Brier	Verdict
Match result	1523	49.6%	47.6% (always H)	0.6149	beats baseline
Over / Under 2.5 goals	1523	49.6%	50.4% (always Y)	0.5066	no better
Both teams to score	1523	44.9%	55.5% (always Y)	0.5035	no better

Sweden

Trained on 3229 matches before 2024-07-14; scored on 808 after.

Market	Test matches	Accuracy	Baseline	Brier	Verdict
Match result	808	43.2%	41.1% (always H)	0.6416	beats baseline
Over / Under 2.5 goals	808	50.0%	50.2% (always N)	0.5064	no better
Both teams to score	808	52.8%	52.4% (always Y)	0.4997	beats baseline

Switzerland

Trained on 2179 matches before 2024-04-13; scored on 545 after.

Market	Test matches	Accuracy	Baseline	Brier	Verdict
Match result	545	43.7%	43.3% (always H)	0.6367	beats baseline
Over / Under 2.5 goals	545	61.1%	61.1% (always Y)	0.4775	no better
Both teams to score	545	63.5%	63.5% (always Y)	0.4697	no better

Usa

Trained on 6138 matches before 2024-07-18; scored on 1535 after.

Market	Test matches	Accuracy	Baseline	Brier	Verdict
Match result	1535	45.3%	45.3% (always H)	0.6457	no better
Over / Under 2.5 goals	1535	59.7%	61.2% (always Y)	0.4843	no better
Both teams to score	1535	62.3%	62.1% (always Y)	0.478	beats baseline

In-sample (not a measure of skill)

Below, each group's live models are scored on the matches they were trained on. With team-identity features a model can memorise individual fixtures, which is why these read far higher. These figures do not indicate the model works — they are kept only to spot training failures. Judge by the out-of-sample section above.

Central (in-sample)

Market	Matches	Accuracy	Brier
Match result	800	59.6%	0.5429
Over / Under 2.5 goals	800	58.5%	0.4635
Both teams to score	800	51.7%	0.5185

China (in-sample)

Market	Matches	Accuracy	Brier
Match result	800	66.5%	0.472
Over / Under 2.5 goals	800	58.8%	0.4523
Both teams to score	800	59.0%	0.5215

Denmark (in-sample)

Market	Matches	Accuracy	Brier
Match result	800	64.5%	0.5293
Over / Under 2.5 goals	800	48.6%	0.548
Both teams to score	800	55.7%	0.5382

England (in-sample)

Market	Matches	Accuracy	Brier
Match result	800	52.1%	0.593
Over / Under 2.5 goals	800	73.4%	0.4648
Both teams to score	800	52.9%	0.4918

Finland (in-sample)

Market	Matches	Accuracy	Brier
Match result	800	55.1%	0.5655
Over / Under 2.5 goals	800	17.0%	0.5343
Both teams to score	800	22.5%	0.5243

France (in-sample)

Market	Matches	Accuracy	Brier
Match result	800	58.8%	0.5459
Over / Under 2.5 goals	800	78.1%	0.4618
Both teams to score	800	57.8%	0.4776

Germany (in-sample)

Market	Matches	Accuracy	Brier
Match result	800	60.3%	0.5463
Over / Under 2.5 goals	800	59.0%	0.4448
Both teams to score	800	58.0%	0.4726

Iberia East (in-sample)

Market	Matches	Accuracy	Brier
Match result	800	69.8%	0.4517
Over / Under 2.5 goals	800	77.6%	0.4428
Both teams to score	800	53.0%	0.4962

Ireland (in-sample)

Market	Matches	Accuracy	Brier
Match result	800	61.8%	0.54
Over / Under 2.5 goals	800	43.2%	0.5268
Both teams to score	800	52.2%	0.5112

Italy (in-sample)

Market	Matches	Accuracy	Brier
Match result	800	64.0%	0.5229
Over / Under 2.5 goals	800	49.4%	0.4965
Both teams to score	800	57.4%	0.4797

Japan (in-sample)

Market	Matches	Accuracy	Brier
Match result	800	58.1%	0.5748
Over / Under 2.5 goals	800	65.9%	0.4851
Both teams to score	800	52.5%	0.4938

Low Countries (in-sample)

Market	Matches	Accuracy	Brier
Match result	800	63.1%	0.5094
Over / Under 2.5 goals	800	58.5%	0.4636
Both teams to score	800	58.1%	0.4913

Mexico (in-sample)

Market	Matches	Accuracy	Brier
Match result	800	56.2%	0.577
Over / Under 2.5 goals	800	51.7%	0.4977
Both teams to score	800	52.9%	0.4948

Norway (in-sample)

Market	Matches	Accuracy	Brier
Match result	800	56.5%	0.5601
Over / Under 2.5 goals	800	59.3%	0.4898
Both teams to score	800	59.0%	0.4724

Poland (in-sample)

Market	Matches	Accuracy	Brier
Match result	800	57.6%	0.5665
Over / Under 2.5 goals	800	26.0%	0.5064
Both teams to score	800	51.2%	0.5005

Romania (in-sample)

Market	Matches	Accuracy	Brier
Match result	800	55.0%	0.5681
Over / Under 2.5 goals	800	61.3%	0.4518
Both teams to score	800	52.9%	0.4915

Russia (in-sample)

Market	Matches	Accuracy	Brier
Match result	800	57.4%	0.5468
Over / Under 2.5 goals	800	54.1%	0.4977
Both teams to score	800	51.5%	0.5002

Scotland (in-sample)

Market	Matches	Accuracy	Brier
Match result	800	72.9%	0.4803
Over / Under 2.5 goals	800	49.2%	0.5096
Both teams to score	800	88.4%	0.4866

South America (in-sample)

Market	Matches	Accuracy	Brier
Match result	800	50.4%	0.5972
Over / Under 2.5 goals	800	58.9%	0.4747
Both teams to score	800	54.0%	0.488

Spain (in-sample)

Market	Matches	Accuracy	Brier
Match result	800	56.6%	0.5513
Over / Under 2.5 goals	800	60.3%	0.4642
Both teams to score	800	73.9%	0.4943

Sweden (in-sample)

Market	Matches	Accuracy	Brier
Match result	800	53.7%	0.5734
Over / Under 2.5 goals	800	51.2%	0.4845
Both teams to score	800	53.7%	0.5106

Switzerland (in-sample)

Market	Matches	Accuracy	Brier
Match result	800	61.3%	0.5475
Over / Under 2.5 goals	800	58.1%	0.5085
Both teams to score	800	58.4%	0.5292

Usa (in-sample)

Market	Matches	Accuracy	Brier
Match result	800	50.5%	0.6033
Over / Under 2.5 goals	800	54.7%	0.5261
Both teams to score	800	57.3%	0.5246