QuantForexTrading - Quantitative Forecast

Please sign up to gain access to the weekly forecast, for most the recent forecasts.

Please sign up to gain access to the daily forecast, for most the recent forecasts.

Combining Multiple Predictions

Each signal / prediction produced by the platform is derived from an individual machine learning model trained on historical data that best matches current market conditions. Every signal derived is first rigorously tested for out of sample performance and given an accuracy score. Only signals that meet a minimum level of historical accuracy are displayed. Each signal is grouped according to an accuracy threshold, allowing you to make further judgement regarding market movement at different confidence intervals.

The platform periodically produces a large number of signals from many different machine learning models tailored specifically for recent market news. Machine learning models are grouped into six broad categories and trained using a different modelling methodology, referred to as M1, M3, M6, M9, M12, M15 modelling methodology. By combining many different predictions made from many different models trained on different historical data, the platform is able to deliver a strong aggregated view of future market movement based on current economic conditions.

Such an approach always outperforms any one single model with many features that generalizes poorly and does not effectively model internal decision making process regarding currency valuation used by large institutional investors and banks.

M1 Methodology Description

The M1 Models predicts the likelihood of hitting either a positive or negative barrier (price level) first. This prediction is made for 20 different price levels in 5 pip increments. All such predictions are accompanied by an accuracy score and used to assign a net price movement at different confidence intervals. Every prediction made looks at news releases over the last four days and corresponding market reaction and only includes a prediction if it meets a given historical accuracy threshold.

M3 Methodology Description

The M3 Models forecasts an expected price movement in either the positive or negative direction over a given time horizon. The forecast is made 20 times corrsponding to 1 to 20 days look ahead window into the future. All such forecasts are accompanied by an accuracy score and used to assign a net price movement at different confidence intervals. Every prediction made looks at news releases over the last four days and corresponding market reaction and only includes a prediction if it meets a given historical accuracy threshold.

M6 Methodology Description

The M6 Models forecasts an expected price movement in either the positive or negative direction over a given time horizon. The M6 models differs in the features included in the model and uses a larger number of features to forecast price movement. The features look at the historical distribution of expected vs actual economic data across all major currencies. It also looks at the impact of given news items to produce a more complex model.

M9 Methodology Description

The M9 Models predicts the likelihood of hitting either a positive or negative barrier (price level) first. The M9 models differs in the features included in the model and uses a larger number of features to forecast price movement. The features look at the historical distribution of expected vs actual economic data across all major currencies. It also looks at the impact of given news items to produce a more complex model.

M12 Methodology Description

The M12 Models forecasts an expectes price movement in either the positive or negative direction over a given time horizon. The M12 models differs in the features included in the model and focuses on a change in market direction or behaviour in relation to a cluster of news items previously released.

M15 Methodology Description

The M15 Models predicts the likelihood of hitting either a positive or negative barrier (price level) first. The M15 models differs in the features included in the model and focuses on a change in market direction or behaviour in relation to a cluster of news items previously released.