QanSystem is not a Crystal Ball…
…but it is as close as it gets!
QanSystem provides short-term distributional forecasts of financial market behaviour and generates trading advice for selected markets (e.g. the German stock market index, DAX) using a purely data-driven approach which exploits trends, hidden patterns and other market features that are completely ignored by model-dependent methods. On the basis of this forecast algorithm the trades suggested and executed are entirely reflective of the client’s utility and risk profile.
Comparing QanSystem with ‘perfect foresight’, the forecast quality is almost as good as it conceivably can be. Over the past 15 years QanSystem correctly predicted the short-term direction of the DAX in more than 64% of all cases.
To take advantage of QanSystem’s forecast quality, the trading signals themselves are translated into trading positions using futures, options, etc. Given that the entire distribution function is forecasted every 15 minutes, the investment process not only generates prospective risk / return ratios and opening signals but also provides closing signals in a timely manner. That is, QanSystem has an inherent risk assessment and position control process. For open positions, this makes it possible to let profits run and have losses cut short, possibly after just 15 minutes.
Comparing Predictive Methodologies
The normal distribution (red line) is obviously symmetric and very smooth, implying that normal distributions don’t have much scope to capture many facets of what is happening in the real world. As the forecasted normal distribution centred at the current DAX level contains only one parameter, namely volatility, this is all that can be captured.
The more sophisticated QanSystem probability distribution (green line) is not symmetrical, providing a far greater insight into the future market position. In this example we see a much greater downside risk than suggested by the normal distribution.
As a consequence of this the QanSystem hit rate for predicted market direction is usually over 60% and close to the conceivable optimum of about 64.5% (based on the DAX and a forecast horizon of 5 trading days), compared with a hit rate of 50% for the normal distribution, making the comparative advantage of QanSystem clearly evident.
QanSystem’s Predictive Power
The main feature of QanSystem is the 5-trading-day distributional forecast generated every 15 minutes. To verify its predictive power, statistical tests of the forecast quality are being performed and charted using p-p plots of the forecasted cumulative distribution function vs reality (see figure below). In the ideal case that all forecasts were perfect, one would obtain a straight line from (0, 0) to (1, 1), as depicted in blue. Any departure from this line is an indication that the forecast distribution functions are not perfect. In such situations the confidence in the distribution forecast decreases, down to a point where the forecast has to be rejected. Depending on the market and the time period considered this may occur in 5 to 20 per cent of cases. However, usually QanSystem closely matches the “perfect result” curve, an indicator that the trading recommendations are of great quality. By contrast, not only is it that conventional methods typically don’t produce distribution forecasts, even if they did, they would have to be rejected almost all of the time.
Although the DAX has been chosen as the first market traded, QanSystem works equally well for other markets. There are just two important requirements:
- Sufficient market liquidity (as is the case for most major indices)
- At least three years of historical data at 15 minute intervals