Here's something that confuses first-time Sovarion users: the same technical pattern can get wildly different confidence scores depending on when it appears. A bullish breakout above resistance might get confidence 65 on Tuesday and confidence 35 on Thursday. Same pattern, same asset, different score. Why?
The answer is market regimes.
What is a market regime? It's the broader market condition that defines the rules of the game. The Brain classifies every moment into one of 9 regimes:
- . trend_up — Clear upward trend. ADX above 25, price above key moving averages. Trend-following signals work well here.
- . trend_down — Clear downward trend. Same mechanics, opposite direction.
- . range — Sideways market. ADX below 20. Mean-reversion signals work better than breakouts.
- . high_volatility_range — Sideways but with big swings. Dangerous for both trend and range strategies.
- . event_driven — A high-impact event (earnings, FOMC, CPI) is imminent. All bets are off until the event resolves.
- . risk_on — VIX is low, money flowing into growth stocks. Bullish signals get a boost.
- . risk_off — VIX is elevated, money flowing to safety. Bearish signals get a boost.
- . unstable_breakout — The market is at a critical level and could break either way. High reward but also high risk.
- . low_liquidity_noise — Volume is too low for reliable signals. Weekend hours, pre-market, holidays.
Why regimes matter for your trading:
The same "bullish MACD crossover" means very different things in different regimes:
- In trend_up: Confirms the existing trend. Confidence boost. Probably actionable.
- In range: Could be a fakeout. The MACD crossed because price bounced off the range floor, not because a trend started. Confidence penalty.
- In event_driven: Completely unreliable. The FOMC decision in 3 hours could nuke any technical signal. Massive confidence penalty.
- In low_liquidity_noise: The crossover happened on negligible volume. Statistically meaningless. Confidence near zero.
How the Brain uses regimes:
Every forecast includes the regime classification and a "regime confidence" percentage. When you see "trend_up, 80% confidence," it means the Brain is fairly certain we're in an uptrend. When you see "range, 55% confidence," the Brain is less sure — maybe we're transitioning between regimes.
The regime feeds into every subsequent module: the hypothesis agent generates different hypotheses for trending vs ranging markets. The scoring layer adjusts weights based on regime. The critic agent knows to be more skeptical during regime transitions.
A practical tip: When the regime confidence is below 60%, treat the forecast with extra caution. The Brain is telling you "I'm not sure what market we're in right now" — which means the signals it generates are inherently less reliable.
One more thing: For 7-day and 30-day analysis, we skip the "low_liquidity_noise" regime entirely. Weekend volume is irrelevant for a monthly investment thesis. This is a deliberate design choice that improves the quality of longer-horizon forecasts.