How does the ML Prediction widget generate forecasts?

Nydar's ML Prediction widget uses XGBoost gradient boosting models that extract 100+ features from price, volume, and technical indicators, then select the top 15 by predictive importance. Each prediction includes a confidence score from 0-100% and the key factors driving the forecast, refreshing every 60 seconds with fresh market data.

Should I trade based solely on ML predictions?

No. Nydar's ML predictions are one tool among many and should be combined with the Signals and Levels widget, your own chart analysis, and proper risk management. Even predictions above 70% confidence are right roughly 7 out of 10 times — meaning 3 out of 10 will be wrong. No model predicts the market with certainty.

What do the confidence scores mean?

Nydar's confidence scores reflect how strongly the model's 15 selected features align with the predicted direction. Scores of 50-60% indicate weak agreement and low conviction. Scores of 70-80% show strong indicator alignment and are more actionable. Scores above 80% are rare — appearing on roughly 5-10% of predictions — and represent the highest-conviction forecasts.

Back to Help Center

ML Prediction

Nydar's ML prediction widget uses XGBoost models that extract 100+ features and select the top 15 by importance to generate directional forecasts with confidence scores

What Is Nydar's ML Prediction Widget?

Nydar's ML Prediction widget uses XGBoost models that extract 100+ features and select the top 15 by predictive importance to forecast short-term price direction. Each prediction includes a bullish or bearish call, a confidence score from 0-100%, and the key indicators driving the forecast. Predictions update every 60 seconds with fresh market data. The gradient boosting framework is detailed in Tianqi Chen and Carlos Guestrin's XGBoost: A Scalable Tree Boosting System (Proceedings of KDD, 2016), which demonstrated how regularised gradient boosting achieves state-of-the-art performance on structured data, and Marcos Lopez de Prado's Advances in Financial Machine Learning (Wiley, 2018), which covers feature importance, cross-validation pitfalls, and bet sizing for ML-driven trading strategies.

See the AI signals feature page for a full overview of Nydar's machine learning capabilities, or read the AI trading signals guide for the technical details.

How Does Nydar's ML Model Generate Predictions?

Nydar's model extracts 100+ features — including RSI, MACD, Bollinger Bands, moving averages, volume ratios, macro-economic indicators, and momentum oscillators — then selects the top 15 by importance and feeds them into an XGBoost gradient boosting classifier. Nydar outputs the probability of upward versus downward movement in the next period, displayed as a direction and confidence percentage, refreshing every 60 seconds.

Input Features

Nydar's model analyzes: RSI, MACD, moving averages, Bollinger Bands, volume, price momentum, volatility, and other technical indicators.

XGBoost Model

Nydar uses gradient boosting to find patterns in the features that historically preceded price movements.

Output

Nydar outputs the probability of upward vs downward movement in the next period, displayed as direction and confidence percentage.

What Do Nydar's Confidence Scores Mean?

Nydar's confidence scores reflect how strongly the model's features align with the predicted direction. A score of 50-60% means weak agreement with low conviction. Scores of 70-80% indicate strong indicator alignment and are more actionable. Scores above 80% are rare and represent Nydar's highest-conviction predictions.

Bullish Prediction

Nydar's model predicts price is more likely to increase. Higher confidence = stronger signal.

Bearish Prediction

Nydar's model predicts price is more likely to decrease. Higher confidence = stronger signal.

Confidence Score

50-60%: Weak signal, low confidence
60-70%: Moderate signal
70-80%: Strong signal
80%+: Very strong signal (rare)

Which Indicators Drive Nydar's Predictions?

Nydar's XGBoost model extracts 100+ features and selects the top 15, with momentum indicators like RSI and MACD typically ranking highest in feature importance. Volatility measures (ATR, Bollinger Band width) and moving average relationships follow. Understanding which features drive a prediction helps you judge whether it aligns with current conditions.

Momentum Indicators

High Importance

RSI, MACD, and rate of change are typically the strongest predictors. Nydar's model detects momentum shifts before they become visually obvious.

Volatility Measures

Medium-High

Bollinger Band width, ATR, and standard deviation help Nydar's model distinguish between trending and ranging conditions.

Moving Averages

Medium

Price relative to 20/50/200 MAs, plus MA crossover states, provide trend context for the prediction.

Volume

Moderate

Relative volume, OBV trend, and volume-price divergences help confirm whether moves have conviction.

When Should I Trust or Ignore Nydar's Prediction?

Trust Nydar's predictions more when confidence is above 70%, the prediction aligns with the broader trend, and no major news events are pending. Discount or ignore Nydar's predictions below 60% confidence, during earnings announcements, in extreme market conditions, or when they contradict strong technical levels.

Trust More When

• Confidence is 70%+ (stronger signal)
• Prediction aligns with the broader trend
• No major news events pending
• Volume is normal or above average
• Multiple timeframes agree

Ignore or Discount When

• Confidence is below 60%
• Major earnings or Fed announcements pending
• Market is in extreme fear/greed conditions
• Low volume (e.g., holidays, after-hours)
• Prediction contradicts strong technical levels

Critical Warnings

No prediction is guaranteed: Nydar's ML models can and do get it wrong. Never trade blindly based on predictions.
Past performance ≠ future results: Nydar's model is trained on historical data. Markets change.
Black swan events: Nydar's models cannot predict news, hacks, regulations, or other unexpected events.
Use as one input: Combine Nydar's predictions with your own analysis. ML is a tool, not a crystal ball.

Pro Tips

High confidence + technical confluence: Best setups are when Nydar's ML prediction aligns with chart patterns and Nydar's indicator signals.
Avoid trading weak signals: 55% confidence is barely better than a coin flip. Wait for Nydar's stronger signals.
Track accuracy: Note Nydar's prediction outcomes in your journal to assess real-world model performance.
Timeframe matters: Nydar's predictions are for the selected timeframe. A bullish 1h prediction says nothing about the daily trend.

How ML Prediction Compares to Other Nydar Widgets

Feature	ML Prediction	Signals & Levels	Daily Picks	Indicators	Backtester
Approach	XGBoost ML model	Rule-based convergence	Multi-factor ranking	Raw values	Historical simulation
Features analysed	100+ (top 15 selected)	13 patterns + 5 EMA pairs	Multi-indicator score	105 indicators	Strategy-dependent
Confidence score	0-100% probability	Weak / Moderate / Strong	Ranked list	No	Win rate %
Update frequency	Every 60 seconds	Per candle close	Daily	Real-time	On-demand
Explainability	Feature importance shown	Signal type shown	Score breakdown	Full values visible	Trade-by-trade log
Best paired with	Signals + Indicators	Prediction + Chart	Signals + Prediction	Signals + Prediction	Prediction + Signals