Core Pages

Forecasting

6 min read

What it is

Forecasting runs Bayesian probability + Monte Carlo simulation (1,000 paths) on real Binance OHLCV to produce probability-of-up-vs-down estimates for the next 1h / 4h / 24h. Picks any symbol — no bot config required. Output includes a confidence band (5th / 50th / 95th percentile), expected change, expected range, market regime classification, and a plain-English insight.

How to think about it

Forecasts do not predict the future with certainty — no model can. They give you a probability distribution over possible outcomes. Use them to set realistic TP / SL targets and to size positions in proportion to confidence, not to bet on point predictions.

Step-by-step

  1. Pick a symbol + horizon (top of page)

    Quick-pick from BTC/USDT, ETH/USDT, SOL/USDT, etc. — or type any Binance symbol. Toggle 1h / 4h / 24h to switch the forecast window. The whole page (probabilities, expected change, Monte Carlo) recomputes for the selected horizon.

  2. Price Probability ring (left card)

    The big circular ring shows P(Up) — the percentage chance price closes HIGHER than now at the end of the selected horizon. Green portion = up probability. Red portion = down probability. Together they almost sum to 100%; the tiny remainder is "no change."

  3. Forecast Summary (right card)

    Three badges + 2 stat cards + auto-generated insight. The badges tell you the current regime; the stats give expected change + breakout probability; the insight summarizes everything in plain English.

  4. Risk Level badge

    Low / Medium / High / Extreme based on current volatility (ATR) + trend strength. Extreme pulses to draw attention. Adjust position size DOWN when risk is High/Extreme even if the directional probability looks favorable.

  5. Market Regime badge

    Trend-Up / Trend-Down / Sideways / High-Volatility / Accumulation / Distribution. Tells you what type of market the engine thinks you're in. Trend regimes favor momentum strategies; Sideways favors mean-reversion (Grid Bot); High-Vol favors wider stops or sitting out.

  6. Directional Bias indicator

    Bullish ↑ / Bearish ↓ / Neutral → based on the weight of evidence across the indicator stack. This is the engine's lean — slightly different from the verdict you'd see on the AI Signals page because Forecasting weights probabilities, not just indicator counts.

  7. Expected Change stat

    The MEAN expected price move over the horizon (e.g. +2.3% over 24h). This is the center of the probability distribution, NOT a guaranteed move. The Range below it shows the typical ±band — actual moves outside this range happen ~30% of the time.

  8. Breakout Probability

    Probability that price breaks out of its recent range during the forecast window. High breakout probability + High volatility = volatile expansion likely. Low breakout probability + Sideways regime = expect chop, range strategies fit.

  9. Three Forecast Cards (1h / 4h / 24h)

    Side-by-side comparison of all three horizons. Each card shows confidence (high/medium/low/pending), Up %, Down %, Expected %. Use these to spot when the engine has conflicting views — e.g. bullish 1h but bearish 24h often means a near-term bounce inside a longer downtrend. Clicking a card switches the active horizon.

  10. Monte Carlo Simulation chart

    1,000 simulated price paths (using current price + estimated volatility). Three colored lines show the percentile bands: red dashed = 5th (worst case), green solid = 50th (median), purple dashed = 95th (best case). 8 faint sample paths show what individual simulations look like. Wide spread between p5 and p95 = high uncertainty.

  11. Worst / Median / Best price points

    The three small cards below the chart give you concrete target prices at the END of the forecast horizon. Use Worst Case to set a defensive stop-loss. Use Median for realistic TP. Use Best Case to size your "moon" upside scenario — but plan around the median.

Tips & pitfalls

  • Forecasts are most reliable over SHORT windows. 1h forecasts are accurate ~65-70% of the time; 24h drops to ~55%. Beyond 24h the model loses statistical power — that's why there's no 7d or 30d horizon.
  • High Risk + wide Monte Carlo bands means HIGH UNCERTAINTY, not high opportunity. Many beginners size up when they see "Bullish 60%" without noticing the band is ±15%. That ±15% is bigger than the +5% expected move.
  • When the 3 timeframe cards disagree (e.g. 1h Bullish, 4h Neutral, 24h Bearish), the longer horizon usually wins — short-term bounces inside a downtrend are common.
  • P(Up) of 55-60% on a 24h horizon is NOT a strong signal. It means the model is barely above coin-flip. Trade only on 70%+ unless you have other confluence.
  • The plain-English insight text is generated from the same data you see in the cards. If it says "high uncertainty," believe it — that's often more informative than the headline probability.
  • Monte Carlo bands widening over time is normal — uncertainty compounds. A tight band at hour 1 and a wide band at hour 24 reflects the math, not a model bug.
  • See platform-wide forecast accuracy (Wilson-bounded) at [Forecast Accuracy](/forecast-accuracy) before trusting any individual call.