Eredivisie Week 21 AI Model Audit
Cogito v2 70B led Eredivisie predictions with 3.00 avg points/match, followed by Llama 3.2 3B Turbo (2.89) and GPT-OSS 20B (2.67). Models achieved 30.49% correct tendency overall. PSV Eindhoven vs Feyenoord (3-0) was a significant upset, with a consensus draw prediction.
Cogito v2 70B led Eredivisie predictions with 3.00 avg points/match, followed by Llama 3.2 3B Turbo (2.89) and GPT-OSS 20B (2.67). Models achieved 30.49% correct tendency overall. PSV Eindhoven vs Feyenoord (3-0) was a significant upset, with a consensus draw prediction.
Top 10 Models
| # | Model | Matches | Total Points | Avg Pts/Match | Tendency % | Exact % |
|---|---|---|---|---|---|---|
| 1 | Cogito v2 70B (Deep Cogito) | 9 | 27 | 3.00 | 55.6% | 33.3% |
| 2 | Llama 3.2 3B Turbo (Meta) | 9 | 26 | 2.89 | 55.6% | 11.1% |
| 3 | GPT-OSS 20B (OpenAI) | 3 | 8 | 2.67 | 66.7% | 33.3% |
| 4 | Qwen 2.5 72B Turbo (Alibaba) | 9 | 23 | 2.56 | 44.4% | 22.2% |
| 5 | Llama 3.3 70B Turbo (Meta) | 9 | 22 | 2.44 | 44.4% | 22.2% |
| 6 | Kimi K2 Instruct (Moonshot) | 8 | 19 | 2.38 | 50.0% | 12.5% |
| 7 | Gemma 3n E4B (Google) | 8 | 19 | 2.38 | 50.0% | 12.5% |
| 8 | Marin 8B Instruct (Marin Community) | 9 | 21 | 2.33 | 44.4% | 11.1% |
| 9 | DeepSeek R1 (Reasoning) | 9 | 20 | 2.22 | 44.4% | 11.1% |
| 10 | Rnj-1 Instruct (Essential AI) | 9 | 17 | 1.89 | 44.4% | 11.1% |
Match-by-Match Audit
- FC Volendam vs GO Ahead Eagles: 20.0% correct tendency (1-1 result)
- Heerenveen vs Utrecht: 44.4% correct tendency (1-1 result)
- Heracles vs Fortuna Sittard: 7.4% correct tendency (2-1 result)
- PSV Eindhoven vs Feyenoord: 37.0% correct tendency (3-0 result)
- Excelsior vs Ajax: 25.9% correct tendency (2-2 result)
- Sparta Rotterdam vs Groningen: 19.2% correct tendency (2-0 result)
- PEC Zwolle vs Telstar: 16.7% correct tendency (4-1 result)
- AZ Alkmaar vs NEC Nijmegen: 63.0% correct tendency (1-3 result)
- NAC Breda vs Twente: 40.7% correct tendency (2-2 result)
Biggest Consensus Misses
- FC Volendam vs GO Ahead Eagles (1-1) | Consensus: A (76.0%)
- Excelsior vs Ajax (2-2) | Consensus: A (74.1%)
- NAC Breda vs Twente (2-2) | Consensus: A (59.3%)
- Heracles vs Fortuna Sittard (2-1) | Consensus: A (55.6%)
- PSV Eindhoven vs Feyenoord (3-0) | Consensus: D (55.6%)
Methodology
Models were evaluated based on their match-level predictions. Points were awarded as follows: 8 points for exact score, 3 points for correct tendency but wrong score, 0 points otherwise. Average points per match and percentage statistics were calculated based on these scores.
Generation cost: $0.0021
Tokens: 4,434 input + 1,118 output
Frequently Asked Questions
What is this article about?
You might also like
Eredivisie Week 24 AI Model Audit: Llama 4 Scout Leads
Llama 4 Scout (OpenRouter) performed best with 4.00 avg points/match, followed by Trinity Large Preview (3.22) and Llama 3.3 70B Instruct (3.00). Overall accuracy was 52.05% correct tendency. The biggest upset was Twente's 2-1 win over Groningen, which fooled 84.2% of models predicting a draw.
Feb 23, 2026
Eredivisie Round 23 AI Model Accuracy: Devstral Small Leads
Devstral Small (OpenRouter) led Eredivisie predictions this week with 3.43 points per match, followed by Llama 3.1 8B (OpenRouter) at 3.40 and Llama 3.3 70B Instruct (OpenRouter) at 3.25. Models achieved 36.86% correct tendency overall, though FC Volendam's 2-1 win over PSV Eindhoven caught most models off guard.
Feb 16, 2026
Eredivisie AI Predictions Audit - Regular Season 22
Gemma 3n E4B led with 3.11 avg points/match, followed by Mistral 7B v0.3 and Kimi K2 Thinking (3.00). Models achieved 50.62% correct tendency overall. Twente vs Heerenveen (5-0) was a major upset, with only 12% correct tendency.