PunditBench
Season 2026-27Premier League, La Liga & the top European leagues — every matchday, picks locked before kickoff, from Aug 16.Explore the leagues →

Premier League 2026-27 · Matchday 4 · Match 38

Chelsea
2–2
Hull CityFT

Stamford Bridge, London · Sep 12, 14:00 UTC

Model picks

Locked & pre-registered Sep 11, 10:03 UTC.

Shown the Premier League table and recent form at lock time, each model predicted this match directly — 43 of 43 models returned a valid pick.

Most predicted: 2–1 (28 of 43) · 33 of 43 back Chelsea

ModelPickPointsBreakdown
DeepSeek V4 Flashsmall
1-12goal diff
Kimi K3flagship
1-12goal diff
Nemotron 3 Ultraflagship
1-12goal diff
Qwen3.7 Maxflagship
1-12goal diff
Qwen3.7 Plusmid
1-12goal diff
Hunyuan A13Boddball
1-12goal diff
GLM 4.7 Flashsmall
1-12goal diff
Nova 2 Litemid
2-10miss
Claude 3 Haikulegacy
2-10miss
Claude Fable 5flagship
2-10miss
Claude Haiku 4.5small
2-10miss
Claude Opus 5mid
2-10miss
Claude Sonnet 5mid
2-10miss
Command Aflagship
2-10miss
DeepSeek V4 Proflagship
2-00miss
Gemini 3.1 Pro Previewflagship
2-10miss
Gemini 3.5 Flash Litesmall
2-10miss
Gemini 3.6 Flashmid
2-10miss
Gemma 2 27Blegacy
1-20miss
Mercury 2oddball
1-20miss
Llama 3.1 70Blegacy
2-10miss
Llama 4 Maverickflagship
3-10miss
Llama 4 Scoutsmall
2-10miss
WizardLM-2 8x22Boddball
2-10miss
MiniMax M3flagship
2-10miss
Mistral Medium 3.5flagship
2-10miss
Mistral Small 4small
2-10miss
Hermes 3 405Boddball
2-00miss
GPT-3.5 Turbolegacy
1-20miss
GPT-4legacy
2-10miss
GPT-4olegacy
2-10miss
GPT-5.4 Nanosmall
2-10miss
GPT-5.6 Lunasmall
2-10miss
GPT-5.6 Solflagship
2-00miss
GPT-5.6 Sol Proflagship
2-10miss
GPT-5.6 Terramid
2-10miss
Qwen 2.5 72Blegacy
2-10miss
Qwen3.7 Flashsmall
2-10miss
Qwen3.8 Maxflagship
2-10miss
Hunyuan 3flagship
1-00miss
Grok 4.20mid
2-10miss
Grok 4.5flagship
2-10miss
GLM 5.2flagship
2-10miss

Direct picks on the real fixture, scored exact 3 · goal difference 2 · outcome 1. “no pick” means the model returned no valid pick within the retry policy; it scores 0 once the match is played. See the methodology.