Knowledge of the historical and current values of financial time series (such as stock index and technical indicator values, stock prices, changes in yield, etc.) is not enough for making effective investment decisions. The ability to transform this knowledge into reliable forecast of future values is much more important. All the efforts of analysts are directed to receive future outlook whatever method they use – fundamental or technical, quantitative or something else.
The ability to make reliable forecasts is determined by the trader's experience, intuition, and intelligence. However even the trader with all aforesaid qualities cannot always provide highly accurate forecasts perfectly.
What would you prefer: a price forecast coinciding with real value in three cases out of ten and in the other seven cases missing by 50% or forecast of all ten values with relative error 5% at most?
We prefer the second one.
What is necessary for good forecast? We suppose one of the most important components is a reliable forecast generation mechanism being never tired.
e-MasterTrade has developed such mechanism using the latest achievements of modern mathematics.
It is necessary to fine-tune forecasting models for each type of financial series (indices, stock prices or technical indicators) since ones have their own movement peculiarities.
What kind of forecast do you need?
What kind of stocks are you going to build forecast for? Our unique analytical tool can help to answer this question.
StockPuzzle can help to select and analyze the most suitable stocks for you and pick out the most typical stocks form each reviled cluster.
How to evaluate the quality of forecast ? |
It is very important to have a possibility to make evaluation of forecast quality without bias. This evaluation allows to take into account possible error risk while making investment decisions.
We have specially developed G criterion to evaluate the accuracy of forecast. It takes into account both forecast dispersion against true values and possible error in trend direction. G values lie in an interval from 0 to 100. The peak value is reached when the real and forecasted time series are completely identical (i.e. when values coincide). The minimal value represents the worst kind of forecast. Fortunately it never can be achieved.
This criterion illustrates not only the accuracy of forecast but the reliability as well. Thus the forecast can be considered reliable if G criterion approaches the peak value of 100 because the strong deviation is improbable and the direction of price change is most likely correct. There are no such guarantees if the forecast is not reliable. Though the real value may sometimes coincide with the forecasted one there are no grounds for such concurrences.
The certain criteria of forecast quality we have managed to develop using the theory of errors is the following:
- G >= 76.9 - high-quality forecast, there is rather slight error risk, the color of forecast plots is green;
- 66.7 =< G < 76.9 - good-quality forecast, the risk amount is acceptable, the color of forecast plots is violet;
- G < 66.7 - there are no guarantee the forecast is reliable, the error risk is significant, the color of forecast plots is red.
G criterion provides the evaluation of forecast reliability immediately when the forecast is ready. It helps to put into practice the effective risk management and develop the successful strategies of stock trading.
In conclusion we would like to note that the reliability of financial time series forecast often depends on the certain period of time. It does not mean that the forecasting mechanism "doesn’t work". The reason is that any kind of regularities in the world of finance has a short-term character.
The mankind has never given up attempts to describe and model the stock market more completely and adequately since it appeared. The science is not standing still and brand new methods of modern physics and mathematics are moving from exclusively academic area to applied financial analysis. e-MasterTrade has the special page About analytical tools that contains some interesting information about such latest innovative approaches. There you can find out how to predict future trend direction by time series, get to know about new risk measures for series analysis and use some other analytical tools.
|
Main
