🔮 How Predictive CandleScope Works

The Model

A time-series foundation model trained on a large corpus of financial and general time-series data. It takes the last 512 hourly candles as context and outputs probabilistic price and volume paths for the next trading session.

Monte Carlo Sampling

Instead of a single prediction, the model runs 30 independent samples with stochastic decoding (temperature 1.2, top-p 0.95). This produces a distribution of possible futures rather than one deterministic line.

What You See on the Chart

• Candles Historical 1-hour close prices (last 12 days)
• Orange line Mean forecast — average across all 30 sampled paths
• Shaded band 10th–90th percentile range of all sampled paths
• Red dashed Forecast boundary — where history ends and prediction begins
• White line Live intraday price (market hours only)

Stat Cards Explained

• Upside Probability — % of the 30 sampled paths whose final price is above the last known close. Values above 60% are bullish green, below 35% are bearish red.
• Volatility Amplification — Ratio of predicted volatility to recent historical volatility, scaled to a 0–100% range. High values signal the model expects an unusually active session.
• Aggregate Prediction — Average predicted price change (%) across all tracked tickers, giving a broad market sentiment signal.

Important Notes

• Forecasts are updated once per day before market open and are not real-time predictions.
• During market hours, live intraday prices are overlaid so you can compare the forecast against actual movement.
• This tool is for informational purposes only and does not constitute financial advice.