Is accuracy betting advice?

No. The page reports historical model performance only; football outcomes remain uncertain.

Public data

AI Football Prediction Quality Report

Name: Football Predictor football prediction dataset
Creator: Football Predictor
License: https://football.edccc.cn/en/about

Public tracking for evaluated prediction samples, hit rate, league cohorts, and model version performance.

Samples

245

Hits

117

Accuracy

47.8%

Public picks

43.7%

League

League	Samples	Accuracy	Draw bias
西甲	53	58.5%	-5.7%
意甲	50	38%	-4%
英超	46	52.2%	-13%
德甲	35	42.9%	-20%
世界杯	32	50%	-18.7%
法甲	24	41.7%	-16.7%
欧冠	5	40%	-40%

Model version

Model version	Samples	Hits	Accuracy
selective-v3	139	60	43.2%
legacy	99	52	52.5%
draw-calibrated-v2	7	5	71.4%

Confidence cohort

Confidence cohort	Samples	Hits	Accuracy
1-4	9	4	44.4%
5-7	204	95	46.6%
8-10	32	18	56.3%

Pick policy

Pick policy	Samples	Hits	Accuracy
Main pick	0	0	0%
Solid lean	71	31	43.7%
Watch list	12	5	41.7%
Avoid / thin data	162	81	50%

Prediction type

Prediction type	Samples	Hits	Accuracy
HOME WIN	137	76	55.5%
DRAW	40	12	30%
AWAY WIN	68	29	42.6%

Data grade

Data grade	Samples	Hits	Accuracy
high	44	17	38.6%
medium	68	31	45.6%
low	34	17	50%
unknown	99	52	52.5%

Odds signal

Odds signal	Samples	Hits	Accuracy
With odds	107	49	45.8%
No odds	39	16	41%
Unknown odds	99	52	52.5%

Probability margin

Probability margin	Samples	Hits	Accuracy
0-4pp	47	18	38.3%
5-9pp	37	21	56.8%
10pp+	161	78	48.4%

Quality notes

主推/稳健预测尚未明显优于全量预测，建议继续收紧主推门槛：提高概率领先差和赔率完整度要求。

英超平局预测偏少，建议提高低比分/实力接近场景的平局下限。

DRAW 类型命中率偏低，建议检查该结果类型的概率校准。

Methodology

This report only counts latest-round AI predictions for fixtures that have finished and been verified. Accuracy is not betting advice and does not guarantee future outcomes.

Public Dataset and Methodology

Football Predictor publishes public prediction quality samples, field definitions, methodology notes, and responsible-use documentation for review and citation.

View public GitHub repository