If you've used Sovarion for more than five minutes, you've noticed the big number on every forecast card: the confidence score. But what does "62" actually mean? And why is the maximum 82, not 100?
The cap at 82 is intentional. Financial markets are inherently uncertain. No model, no matter how sophisticated, should claim 100% confidence in a 30-minute price prediction. We cap the score at 82 because anything higher would be dishonest. When you see 82, it means "this is the strongest signal our system can produce" — not "this is guaranteed to happen."
How it's calculated. The confidence score is a weighted composite of 6 independent quality dimensions:
- . Signal Quality (weight: 35%) — How strong is the evidence? This measures agreement across technical indicators, volume confirmation, and pattern reliability.
2. Context Fit (weight: 20%) — Does the signal match the current market regime? A bullish breakout signal in a trending market gets higher context fit than the same signal in a choppy range.
3. Fragility (weight: 15%, inverted) — How easily could this signal be invalidated? High fragility means one news headline could flip the thesis. This is subtracted, not added.
4. Risk Load (weight: 15%, inverted) — How much risk is the market pricing in right now? Near earnings, near FOMC meetings, during geopolitical crises — risk load goes up and confidence comes down.
5. Pattern Reliability (weight: 10%) — Has this specific pattern worked before in similar conditions? The Brain tracks historical analogs and weights their outcomes.
6. Critic Penalty (weight: 5%, subtracted) — The adversarial critic agent may have flagged weaknesses. If it assigned a penalty, confidence drops.
What the numbers mean in practice:
- 70-82: Strong signal. The Brain sees clear evidence, low fragility, favorable regime. These are rare — maybe 5-10% of all forecasts.
- 50-69: Moderate signal. Directional lean is present but there are counter-arguments. Most actionable forecasts fall here.
- 30-49: Weak signal. Some evidence exists but it's not convincing. Proceed with caution.
- Below 30: No edge. The Brain is essentially saying "I don't know." These are often classified as "no_edge" and shouldn't be traded.
A practical tip: Don't just look at the confidence number. Look at the combination of confidence + fragility. A forecast with confidence 55 and low fragility (stable) is often better than one with confidence 65 and high fragility (could reverse quickly).
The confidence score doesn't tell you "will this stock go up?" It tells you "how much should I trust this particular signal?" That's a subtle but important difference.