How BetDNA Uses AI

When you import a bet slip, a vision model reads the image and extracts structured fields — bookmaker, sport, selection, odds, stake, returns, date and bet type — each with its own confidence score. Anything below our confidence threshold is flagged for you to confirm, so a low-quality screenshot never quietly becomes bad data.

Once bets are stored, BetDNA calculates behavioural metrics across four families: staking behaviour, emotional patterns, betting style and performance. These are ordinary statistics computed from your own records — average stake, stake variance, odds distribution, timing, streak behaviour, chasing patterns and consistency over time.

Insights combine your metrics with the behaviour events recorded on your account — moments like a first profitable month, a discipline improvement, or a run of chasing losses. The system identifies the patterns that changed most, ranks them by significance, and writes them up in clear language.

• Only your own data is used to produce your insights.
• Insights describe what has already happened — they are never forward-looking predictions.
• Comparisons are made against your own past behaviour, not against other users.
• Where the data is thin, the insight says so instead of overstating.

An insight you cannot verify is just an opinion. Every coaching insight includes two expandable layers: Why am I seeing this?, which explains the reasoning in one or two sentences, and Show me the evidence, which reveals the supporting statistics, the trend over time and the bets involved.

If you disagree with an insight, the evidence lets you see exactly where it came from — and correcting the underlying data corrects the insight.

BetDNA publishes no single composite score. Each Behavioural Marker carries its own confidence level, set by how many months of your history are covered and how consistent the underlying evidence is.

• Confidence has four levels: Tentative, Emerging Pattern, Established Pattern and Highly Consistent.
• Confidence is earned through covered months of history, never by placing more bets.
• Where a marker cannot yet be measured reliably, BetDNA says so instead of estimating.
• A statement is never more confident than the weakest input behind it.
• Confidence decays if your history stops being covered and recent behaviour becomes unknown.

Behaviour is only visible across repetition. A single losing week says almost nothing; the same pattern appearing across six months is a genuine trait. As your history grows, BetDNA can separate variance from behaviour — which is the difference between luck and discipline, and the whole point of the platform.

AI extraction is very good but not perfect. You can edit any field on any bet at any time, and corrections take effect immediately across your metrics, profile and insights. If you believe an insight is wrong, contact us with the reference shown on the insight and we will investigate it.

Because AI can occasionally misread a slip or describe a pattern imprecisely, insights are intended to assist your understanding rather than to make decisions for you, and they are not professional advice of any kind. The corresponding clauses in our Terms & Conditions — Nature of the Service, AI Limitations and Accuracy of Analysis — say exactly the same thing in contractual language.