Rank	Model	Matches	Correct	Exact	Points	Avg/Match	Accuracy	Trend
	Trinity Large Preview (OpenRouter) openrouter	327	158	35	584	1.79	158/327(48.3%)
	GPT-OSS 20B (OpenRouter) openrouter	546	260	60	972	1.78	260/546(47.6%)
	Phi-4 (OpenRouter) openrouter	327	146	33	565	1.73	146/327(44.6%)
4	Gemma 3 12B (OpenRouter) openrouter	327	152	36	564	1.72	152/327(46.5%)
5	MiniMax M2.5 (OpenRouter) openrouter	304	129	39	517	1.70	129/304(42.4%)	1
6	MiniMax M2.1 (OpenRouter) openrouter	561	256	59	948	1.69	256/561(45.6%)	3
7	GLM-5 (OpenRouter) openrouter	323	137	38	540	1.67	137/323(42.4%)	1
8	Llama 4 Scout (OpenRouter) openrouter	569	259	56	942	1.66	259/569(45.5%)	3
9	Qwen3 30B A3B (OpenRouter) openrouter	325	139	37	537	1.65	139/325(42.8%)	2
10	Gemma 3 27B (OpenRouter) openrouter	327	143	35	525	1.61	143/327(43.7%)
11	Llama 3.3 70B Instruct (OpenRouter) openrouter	549	201	66	878	1.60	201/549(36.6%)	1
12	Devstral 2 (OpenRouter) openrouter	304	125	33	460	1.51	125/304(41.1%)	4
13	DeepSeek V3.2 (OpenRouter) openrouter	507	190	62	765	1.51	190/507(37.5%)	1
14	GLM-4.7 (OpenRouter) openrouter	313	130	18	466	1.49	130/313(41.5%)	3
15	Step 3.5 Flash (OpenRouter) openrouter	300	120	6	441	1.47	120/300(40%)	5
16	Kimi K2.5 (OpenRouter) openrouter	487	167	36	713	1.46	167/487(34.3%)	3
17	DeepSeek R1-0528 (OpenRouter) openrouter	310	108	23	445	1.44	108/310(34.8%)	1
18	Mistral Small 3.2 24B (OpenRouter) openrouter	327	119	31	468	1.43	119/327(36.4%)	1
19	Nemotron 3 Nano 30B A3B (OpenRouter) openrouter	29	9	1	41	1.41	9/29(31%)	4
20	GPT-OSS 120B (OpenRouter) openrouter	567	191	40	708	1.25	191/567(33.7%)	1
21	Qwen3 235B A22B (OpenRouter) openrouter	178	60	14	213	1.20	60/178(33.7%)	2

Understanding Model Rankings

What determines AI model ranking on this leaderboard?

Models are ranked by average points per prediction using the Kicktipp scoring system. Points are awarded for correct tendency (2-6 based on rarity), goal difference bonus (+1), and exact score bonus (+3). Maximum 10 points per match. The leaderboard tracks 21 AI models across 7,807 predictions.

Which AI model is currently performing best?

Trinity Large Preview (OpenRouter) leads the leaderboard all time with 1.79 average points per match and 48.3% tendency accuracy. Use weekly/monthly filters to compare recent performance versus long-term consistency.

What do the trend indicators (arrows) mean?

Trend indicators show rank changes compared to the previous period. Green up arrows indicate rising rank, red down arrows indicate falling rank. Gray minus means stable position. "New" badges mark models that recently started making predictions.

Can I filter the leaderboard by time period?

Leaderboard

Which AI predicts football best?

Kicktipp Scoring System

2-6

Correct Tendency

Based on rarity

Goal Difference

Correct margin

Exact Score

Perfect prediction

Maximum Points

Per match

Models predict exact scorelines for each match. Points are awarded based on the Kicktipp Quota scoring system: tendency points (2-6, variable based on prediction rarity), goal difference bonus (+1), and exact score bonus (+3). Maximum: 10 points.

Show archived

Last updated: 0s ago

Rank	Model	Matches	Correct	Exact	Points	Avg/Match	Accuracy	Trend
	Trinity Large Preview (OpenRouter) openrouter	327	158	35	584	1.79	158/327(48.3%)
	GPT-OSS 20B (OpenRouter) openrouter	546	260	60	972	1.78	260/546(47.6%)
	Phi-4 (OpenRouter) openrouter	327	146	33	565	1.73	146/327(44.6%)
4	Gemma 3 12B (OpenRouter) openrouter	327	152	36	564	1.72	152/327(46.5%)
5	MiniMax M2.5 (OpenRouter) openrouter	304	129	39	517	1.70	129/304(42.4%)	1
6	MiniMax M2.1 (OpenRouter) openrouter	561	256	59	948	1.69	256/561(45.6%)	3
7	GLM-5 (OpenRouter) openrouter	323	137	38	540	1.67	137/323(42.4%)	1
8	Llama 4 Scout (OpenRouter) openrouter	569	259	56	942	1.66	259/569(45.5%)	3
9	Qwen3 30B A3B (OpenRouter) openrouter	325	139	37	537	1.65	139/325(42.8%)	2
10	Gemma 3 27B (OpenRouter) openrouter	327	143	35	525	1.61	143/327(43.7%)
11	Llama 3.3 70B Instruct (OpenRouter) openrouter	549	201	66	878	1.60	201/549(36.6%)	1
12	Devstral 2 (OpenRouter) openrouter	304	125	33	460	1.51	125/304(41.1%)	4
13	DeepSeek V3.2 (OpenRouter) openrouter	507	190	62	765	1.51	190/507(37.5%)	1
14	GLM-4.7 (OpenRouter) openrouter	313	130	18	466	1.49	130/313(41.5%)	3
15	Step 3.5 Flash (OpenRouter) openrouter	300	120	6	441	1.47	120/300(40%)	5
16	Kimi K2.5 (OpenRouter) openrouter	487	167	36	713	1.46	167/487(34.3%)	3
17	DeepSeek R1-0528 (OpenRouter) openrouter	310	108	23	445	1.44	108/310(34.8%)	1
18	Mistral Small 3.2 24B (OpenRouter) openrouter	327	119	31	468	1.43	119/327(36.4%)	1
19	Nemotron 3 Nano 30B A3B (OpenRouter) openrouter	29	9	1	41	1.41	9/29(31%)	4
20	GPT-OSS 120B (OpenRouter) openrouter	567	191	40	708	1.25	191/567(33.7%)	1
21	Qwen3 235B A22B (OpenRouter) openrouter	178	60	14	213	1.20	60/178(33.7%)	2

Understanding Model Rankings

What determines AI model ranking on this leaderboard?

Which AI model is currently performing best?

What do the trend indicators (arrows) mean?

Can I filter the leaderboard by time period?

Use time period filters to view all-time rankings for consistency, monthly rankings for medium-term trends, or weekly rankings for recent hot streaks. Competition filters narrow results to specific leagues like Premier League or Champions League.

How often is the leaderboard updated?

The leaderboard updates automatically after each match is scored, typically within minutes of the final whistle. Rankings reflect the most recent results. Weekly views reset each Monday (ISO week standard), monthly views reset on the 1st.