🤖 AI benchmark: hit-rate of 7 models

Prematch Live (in-play)

Seven external AI models (Hermes contour) independently analyze the same matches — predicting the outcome (1X2), total (Over/Under), both teams to score (BTTS) and the exact score. Here we honestly compare their predictions against the real result after the final whistle and combine everything into a single accuracy rating. An informational and analytical snapshot, not betting advice.

⚠️ Data is still accumulating — counting starts from 09.07.2026, so all models are compared on the same events (early test predictions are excluded). The sample is still small and not representative. Right now the snapshot holds 409 match(es), 1515 settled AI predictions (Basketball). The figures below are N, not «a percentage you can trust»: the more matches are played out, the more reliable the snapshot becomes. We show it transparently from day one, not only once the sample becomes «convenient».

All sports Football Tennis Basketball Volleyball Hockey Table tennis

Leaderboard · Basketball

Model	N (settled)	1X2	Total points	Exact score	Composite accuracy
Claude	348	63.5%(221/348)	52.5%(136/259)	0.0%(0/263)	41.0%(357/870)
Kimi	195	63.6%(124/195)	51.5%(84/163)	0.0%(0/163)	39.9%(208/521)
Google AI	438	64.1%(280/437)	47.8%(196/410)	0.0%(0/413)	37.8%(476/1260)
ChatGPT	76	60.5%(46/76)	42.4%(25/59)	0.0%(0/59)	36.6%(71/194)
GLM 5.2	224	63.8%(143/224)	42.5%(94/221)	0.9%(2/222)	35.8%(239/667)
Qwen	197	64.0%(126/197)	35.8%(63/176)	0.6%(1/176)	34.6%(190/549)
DeepSeek	37	51.4%(19/37)	38.2%(13/34)	0.0%(0/37)	29.6%(32/108)

grey — sample <5, not representative; «—» — the model has not made a settled prediction yet.

Composite accuracy — the share of correct predictions across all shown markets together: (sum of correct picks) ÷ (sum of all settled picks) across the markets 1X2 + Total points + Exact score. Each market-pick weighs equally. This is hit-rate, not profitability — for money/ROI by model see /ai-agent. Total: a push (score exactly on the line) is excluded from the denominator. «Exact score» — the full final score was guessed correctly (H and A matched); predictions with no recognized score do not count toward the denominator.

Composite model rating · all markets · Basketball

Bar height = the model's composite accuracy across all applicable markets on the current sample. Sorted from best to worst.

41.0% (357/870)

Opus 4.8

39.9% (208/521)

Kimi 2.6

37.8% (476/1260)

Gemini 3.5 Flash

36.6% (71/194)

GPT 5.5

35.8% (239/667)

GLM 5.2

34.6% (190/549)

Qwen 3.7 Plus

29.6% (32/108)

DeepSeek V4 Pro

Bars are AI models by version; grey/dimmed — sample <5, not representative. The snapshot is informational, not betting advice.

Accuracy by market · Basketball

Where each model is strong: one mini-bar per applicable market, with the percentage and (hits/sample).

Claude Composite 41.0%

1X2

63.5% (221/348)

Total points

52.5% (136/259)

Exact score

0.0% (0/263)

Kimi Composite 39.9%

1X2

63.6% (124/195)

Total points

51.5% (84/163)

Exact score

0.0% (0/163)

Google AI Composite 37.8%

1X2

64.1% (280/437)

Total points

47.8% (196/410)

Exact score

0.0% (0/413)

ChatGPT Composite 36.6%

1X2

60.5% (46/76)

Total points

42.4% (25/59)

Exact score

0.0% (0/59)

GLM 5.2 Composite 35.8%

1X2

63.8% (143/224)

Total points

42.5% (94/221)

Exact score

0.9% (2/222)

Qwen Composite 34.6%

1X2

64.0% (126/197)

Total points

35.8% (63/176)

Exact score

0.6% (1/176)

DeepSeek Composite 29.6%

1X2

51.4% (19/37)

Total points

38.2% (13/34)

Exact score

0.0% (0/37)

The model's favorite by 1X2 = the max of P1/X/P2 in its probabilities; for sports without a draw (tennis, volleyball, etc.) the «X» option doesn't participate. grey — sample <5, not representative. Not betting advice.