Skip to main content
Methodology Reference

What the Confidence Score Actually Means

The confidence score tells you how much to trust a match insight. Here's what it looks like in practice, across thousands of real matches.

35–42%
Low confidence
Below coin-flip. The engine lacks reliable data.
47–60%
Medium confidence
Above random, but with notable uncertainty.
69–82%
High confidence
Significantly more reliable. Worth paying attention to.

Performance by Confidence Score

Based on completed matches where the engine's favoured outcome was tracked against the actual result. Based on 22,078 matches. Snapshot: February 28, 2026.

1 / 7Low
34.9%
371 correct out of 1,064
2 / 7Low
41.5%
1,533 correct out of 3,692
3 / 7Medium
46.6%
1,686 correct out of 3,615
4 / 7Medium
52.2%
1,759 correct out of 3,372
5 / 7Medium
59.7%
2,391 correct out of 4,006
6 / 7High
68.6%
2,086 correct out of 3,043
7 / 7High
82%
2,694 correct out of 3,286

Based on 22,078 matches (snapshot: Feb 28, 2026). 856 "Too Close to Call" matches excluded.

What This Means for You

When the engine rates a match as High confidence, the favoured outcome materialised in 69–82% of matches in our historical data. That's a meaningful signal in a sport where even strong favourites lose regularly.

At Medium confidence, accuracy ranges from 47–60% depending on the score — around or above coin-flip, but with notable uncertainty. The analysis is useful context, not a strong signal.

Low confidence is an honest signal too: accuracy drops to 35–42%, below coin-flip. The engine doesn't have enough reliable data to form a stable view. Read the reasoning, but treat it as background context rather than a reliable indicator.

Note: Confidence measures how reliable the analysis is, not which team will win. A balanced 50/50 match can have High confidence if the data says it's genuinely even. A heavy favourite can have Low confidence if signals conflict. The score reflects data quality and signal consistency — not outcome certainty.

How the Score Is Built

The 1–7 score is the sum of four dimensions, each rated 0–2, clamped to a practical range.

1
Data Completeness
Is the data fresh and complete?
2
Signal Agreement
Do the 29 signals point the same way?
3
Context Stability
Normal schedule, no major disruptions?
4
Team Volatility
Are recent results consistent?
One hard cap: even with a perfect score, if no outcome reaches 45% probability the engine considers the match a genuine coin flip and caps confidence at Medium. A confident "this is too close to call" is still honest analysis.

See it live on every match

Every match insight shows its confidence score and the specific reasons behind it.