Stock & Cryptocurrency Market Prediction
Processing historical market data for time series analysis.
Optimized accuracy of predictions and continuously improved machine learning models.
Purchase monthly membership to get access to all symbols with up to a quarter of predictions:
In retrospect
Comparison between prediction a period ago and actual prices that happened in the mean time. Select different periods from dropdown above to update this graph.
We are constantly working to improve accuracy of predictions.
30-Day Prediction Accuracy
How Accuracy is Calculated
The accuracy score reflects how well our model's predictions for the past 30 days matched the actual market movements. This calculation is performed daily for all symbols, and the chart above shows the average accuracy for each day in the 30-day forecast period.
The calculation process for each day's prediction is as follows:
- Directional Correctness: First, we check if the prediction correctly guessed the direction of the price change (up, down, or stable). To avoid penalizing for minor fluctuations, we consider any change within 0.05% of the previous day's price as "neutral". If either the predicted change or the actual change is neutral, or if they both move in the same direction, the prediction is considered directionally correct. If the direction is wrong, the accuracy for that day is 0%.
- Magnitude of Error: If the direction is correct, we then measure the accuracy based on the magnitude of the price change. We calculate the difference between the predicted price change and the actual price change (the absolute error).
- Normalized Accuracy: This absolute error is then normalized by dividing it by the previous day's price. This gives us a percentage error. The final accuracy is calculated as
100% - normalized percentage error. The accuracy is capped at 0%, so it cannot be negative.
Membership access to symbols
AAPL AMZN MSFT GOOG GOOGL BABA FB BRK.A BRK.B TSM V WMT TSLA JNJ PG NVDA MA BTC ETH
Join us and get
Focus on ML models for time series analysis and historical market data.
Continuous research of machine learning techniques. Optimization for prediction accuracy.
Regular updates. Daily inclusion of new predictions and data sources.
