TRADING SIGNALS

OUR SERVICE

We deliver a weekly report that identifies the current prevailing regime in all 4 asset classes. In the weekly report, we also update our trend scanners which identify situations of critical overbought or oversold.

We deliver a monthly report, which is presented to the client during dedicated meetings. In the monthly report, the client finds trading signals on the expected direction of all asset classes for the following month.

All results are backtested extensively and a year and a half of live track record validates our results.

Our trading signals are extrapolated from: Options, Technical Indicators, A.I. predictive models, and Crash Alpha models.

MUCH MORE THAN MACHINE LEARNING

Forecasting financial time series is much more than putting some technical indicators inside a neural network.

Forecasting is effective only when ALL these components are properly developed:

  1. Predictors
  2. Data-Mining
  3. Predictive Models
  4. Backtesting
abstract-3629844_1280-min

OUR SCIENTIFIC APPROACH

PREDICTORS

One of the most neglected rules of financial forecasting is “GARBAGE IN-GARBAGE OUT”.

We dedicate more time to building a reliable set of powerful predictors than anything else.

We divide our long list of financial predictors into 5 main subsets:

  • Option Based
  • Spread Based
  • Technical
  • Economical
  • Microstructural

DATA MINING

Usually, there are hundreds of candidate predictors for a given predictive model.

Consequently, data-mining procedures must be employed to select a more manageable number of predictors:

  • Clustering
  • Feature Selection
  • Anomaly Detection
  • Dimensionality Reduction

Data mining is not complete without an econometric analysis of the chosen predictors and the target variables

PREDICTIVE MODELS

We employ a broad range of predictive models, from simple linear support vector machines to more complex deep reinforcement learning. In the range of models we propose, we include Alpha Crash models and models that summarize the informative content of 40+ technical indicators. No predictive model is operational before passing through a rigorous backtesting procedure to avoid over-fitting.

The final results stem from the weighted average of the forecasts coming from a variety of models which are largely complementary in their predictive capabilities.