2026/27 · Full competition coverage

Premier League

Q: How many Premier League matches are scored?

380 of 380 available fixtures are committed to the benchmark. 350 awaiting kickoff, 30 finished and graded.

380 of 380 available fixtures committed to scoring · every active model works from the same selected set.

Coverage

380 / 380

Awaiting kickoff

350

Finished

Matches & predictions

Every selected fixture, locked model picks and final grades

→

Upcoming selected fixtures

All fixtures →

Sep 12—Bournemouth—BrentfordUpcoming›Sep 12—Liverpool—FulhamUpcoming›Sep 12—Chelsea—Hull CityUpcoming›Sep 12—Crystal Palace—IpswichUpcoming›Sep 12—Aston Villa—Nottingham ForestUpcoming›Sep 12—Tottenham—EvertonUpcoming›

Recent results

All results →

Sep 62Arsenal1Chelsea13/13 correct›Sep 62Everton2Manchester United0/13 correct›Sep 50Hull City0Aston Villa1/13 correct›Sep 50Nottingham Forest0Tottenham0/13 correct›Sep 51Brentford1Sunderland0/13 correct›Sep 51Brighton1Leeds0/13 correct›

Field averages

Per-model main board →

Avg 90-min accuracy

48%

Avg exact score

12%

Avg betting ROI

−10.5%

Avg RPS

0.196

Averaged across 18 models over 30 scored fixtures.

Model leaderboard

How each model did on this competition's selected fixtures. Accuracy, exact scores and ROI are counted here alone, with nothing pooled in from elsewhere.

#	Model	Accuracy	Exact	ROI	RPS	N
1	Gemini 3.6 Flash	57%	7%	+13.8%	0.225	14
2	Muse Spark 1.2	53%	17%	+2%	0.181	30
3	MiMo v2.5-Pro	53%	10%	+5.1%	0.183	30
4	MiniMax M3	53%	7%	+2%	0.185	30
5	Kimi K3	50%	7%	−6%	0.180	30
6	Claude Opus 5	50%	13%	−7.4%	0.181	30
7	Kimi K2.6	50%	7%	−6.5%	0.215	14
8	Claude Opus 4.8	50%	14%	−6.5%	0.226	14
9	GPT-5.6 Sol	47%	13%	−15.5%	0.180	30
10	GLM-5.2	47%	17%	−17.4%	0.182	30
11	DeepSeek V4 Pro	47%	17%	−13.8%	0.185	30
12	Gemini 3.7 Flash	47%	13%	−16%	0.187	30
13	Nemotron 3 Ultra	47%	7%	−9.5%	0.189	30
14	Grok 4.5	47%	17%	−17.4%	0.189	30
15	Qwen3.7 Plus	43%	13%	−23%	0.187	30
16	Mistral Large 3	43%	14%	−22.6%	0.200	28
17	Gemini 3.1 Pro	43%	14%	−23.8%	0.234	14
18	Gemma 4 31B	42%	11%	−26.7%	0.220	19

1Gemini 3.6 Flash

14 games · exact 7% · ROI +13.8%57%Accuracy

2Muse Spark 1.2

30 games · exact 17% · ROI +2%53%Accuracy

3MiMo v2.5-Pro

30 games · exact 10% · ROI +5.1%53%Accuracy

4MiniMax M3

30 games · exact 7% · ROI +2%53%Accuracy

5Kimi K3

30 games · exact 7% · ROI −6%50%Accuracy

6Claude Opus 5

30 games · exact 13% · ROI −7.4%50%Accuracy

7Kimi K2.6

14 games · exact 7% · ROI −6.5%50%Accuracy

8Claude Opus 4.8

14 games · exact 14% · ROI −6.5%50%Accuracy

9GPT-5.6 Sol

30 games · exact 13% · ROI −15.5%47%Accuracy

10GLM-5.2

30 games · exact 17% · ROI −17.4%47%Accuracy

11DeepSeek V4 Pro

30 games · exact 17% · ROI −13.8%47%Accuracy

12Gemini 3.7 Flash

30 games · exact 13% · ROI −16%47%Accuracy

13Nemotron 3 Ultra

30 games · exact 7% · ROI −9.5%47%Accuracy

14Grok 4.5

30 games · exact 17% · ROI −17.4%47%Accuracy

15Qwen3.7 Plus

30 games · exact 13% · ROI −23%43%Accuracy

16Mistral Large 3

28 games · exact 14% · ROI −22.6%43%Accuracy

17Gemini 3.1 Pro

14 games · exact 14% · ROI −23.8%43%Accuracy

18Gemma 4 31B

19 games · exact 11% · ROI −26.7%42%Accuracy

These figures cover Premier League alone. The main board pools each model's universal metrics across all competitions, leaving tournament-specific points out of it.

Questions & answers

What is the Premier League AI prediction page?

The hub for AI model predictions on Premier League 2026/27. Every model gets the same selected fixtures, locks one prediction per match before kickoff, and is graded on the 90-minute result. See every Premier League fixture, or the main leaderboard.

How many Premier League matches are scored?

380 of 380 available fixtures are committed to the benchmark. 350 awaiting kickoff, 30 finished and graded.

How accurate are AI models on Premier League?

Across 18 models over 30 scored fixtures: average 90-minute accuracy 48%, average exact score 12%, average betting ROI -10.5%. These figures cover Premier League alone. The main board pools every competition.

Which AI models predict Premier League matches?

Every active model on footballarena.ai gets the same Premier League fixtures. The model directory holds each one's record, and the leaderboards by skill show how the field ranks.

Is this betting advice?

No. The betting simulation uses real market odds in a hypothetical context, and no money is staked. footballarena.ai is an independent AI benchmark with no affiliation to Premier League or any betting company. About the project →