BitBank
Docs > Accuracy

Live Crypto Charts Prediction Accuracy

Our algorithms use a proprietary evidence based machine learning approach to prediction and are constantly improving and adapting to market conditions

Found evidence is judged as relevant using AI and Machine Learning to find relevant evidence in a way that maximizes expected returns upon trading in a live market. Please flag any irrelevant evidence or any decrease in forecasting accuracy to let us know how we can improve

We find our predictions are more accurate closer up at 5 minutes and less and less accurate as you go out so often the 5 minute forecast is the most accurate and as market conditions change the 2 hour forecast can quickly become outdated so keep upto date with the latest forecasts, It helps to take a few forecasts into consideration to reduce the chances of relying on noise instead of signal and to keep updated with the newest forecast by watching the charts

Some statistics about our historical prediction accuracy: These where made by calculating for every featureset comparing the predicted weighted trade average with what the actual trade average is that we observe. This is one of the tools we can use to know if our forecasts are doing a good job. Over 24 hours (27-28th Nov 10pm UTC) we find:

  • The projected weighted trade average over the next 5 minutes was 0.0243% over, std deviation σ: 0.8792%
  • The projected weighted trade average over the next 30 minutes was 0.0576% over, std deviation σ: 1.3385%
  • The projected weighted trade average over the next 60 minutes was 0.1213% over, std deviation σ: 1.9863%
  • The projected weighted trade average over the next 120 minutes was 0.1919% over, std deviation σ: 2.5071%