WHAT IS FORECASTING ACCURACY?

The FORECASTING ACCURACY is obtained by information in the WINS column. The WINS column has the following information:

At the bottom of the WINS column, a percentage WINS number is given. This percentage is calculated for each issue for which the current position (Long or Short) is more than one day old. The % WINS gives the FORECASTING ACCURACY achieved on the most recent date for which price data is given in the advisory. The percentage WINS is the same as the FORECASTING ACCURACY.

A FORECASTING ACCURACY of 50% or less would not be very impressive; throwing darts should average out to about 50% FORECASTING ACCURACY. But if the FORECASTING ACCURACY averages, over time, to a number higher than 55%, that is a good batting average.

If the FORECASTING ACCURACY averages, over time, to a number higher than 60%, that is a very good batting average, and under these conditions, the advisories and forecasts should probably be very worthwhile and useful, in that you can consider these forecasting inputs when making decisions regarding your trading and investments.

What is a price-direction forecasting ratio advantage? In current market conditions, a client should expect the FORECASTING ACCURACY to be consistently 60% or better. 60% is 1.5 times the magnitude of 40%, giving the client (trader) an expectation of having a price-direction forecasting ratio advantage of 150% when the FORECASTING ACCURACY is 60%. When the FORECASTING ACCURACY is 70%, the client (trader) has a price-direction forecasting ratio advantage expectation of 233% (twice as likely to make a profit as opposed to a loss).

"Letting the winners run" will tend to improve the FORECASTING ACCURACY. This is a MAIN point and should be a primary goal of any trading system. I believe this is the ONLY way to improve FORECASTING ACCURACY, and (at least for this forecasting system) it should be assumed that a high FORECASTING ACCURACY is reported ONLY when letting the profits increase has been the method for achieving a high FORECASTING ACCURACY.

To achieve a high FORECASTING ACCURACY, we use an ADAPTIVE forecasting system, which responds and ADAPTS to changing market conditions. This is necessary in rapidly-changing financial markets. We use forecasting theory developed for engineering applications to analyze trends, coupled with learn-and-optimize techniques from the field of artificial intelligence and a filtering method to separate the orderly price moves from the random price moves.

ALTERNATE WAYS TO CALCULATE FORECASTING ACCURACY --

FORECASTING ACCURACY = 100 x (wins)/(total issues)

But, this has been found to be somewhat misleading, as explained below. A little history is perhaps helpful here. During the first year or so that I provided forecasts to clients, I did indeed compute the forecasting accuracy as follows:

FORECASTING ACCURACY = 100 x (wins)/(total issues)

An important thing about that "old FORECASTING ACCURACY" is that it was more concervative (lower FORECASTING ACCURACY). But, there was a very bad thing about it, as explained here. The "old FORECASTING ACCURACY" calculation was actually misleading to a trader at precicely the time when the forecasts are more valuable. The forecasts are more valuable immediately following the action of the forecasting system to ADAPT to major changes in the market. This forecasting system is a 1-dimentional system; the ONLY way it can adapt is to change position. At the time when major changes occur in the market (major tops and major bottoms), the forecasting system, when working best, must adapt rapidly when major market condition changes occur rapidly.

The following example is given to illustrate how and why the "old FORECASTING ACCURACY" calculation was actually misleading to a trader at precicely the time when the forecasts are more valuable: Let's say there are 50 issues in the portfolio. On a day when market conditions reverse dramatically, let's say that 40 of the issues reverse position, and that this turns out (later, upon observation) to be the RIGHT thing for the forecasting system to do. On this day, let's say there are 8 wins and 2 losses. Furthur, assume that these 40 position changes were NOT necessarily because the 40 positions were losses. (This forecasting system does not "reverse on losses" as a standard way of operating; it can and does reverse position on both losses and wins. In fact, in "good markets" with good tends, there will be times when ALL the position reversals will be for WINNING positions. This is the most desirable result possible, because it compounds profits without any whipsaw trades.)

But, on the day being analyzed, the FORECASTING ACCURACY, calculated the old way would have been:

FORECASTING ACCURACY = 100*8/50 = 16%

If FORECASTING ACCURACY is being used by the trader as an indication of how much RELIANCE the trader should be placing on the forecasts, the trader would (mistakenly) be drawn to the conclusion that he/she could not rely on the forecasts because of the low FORECASTING ACCURACY.

Key point: The reason the FORECASTING ACCURACY is low (using the old FORECASTING ACCURACY formula) is that the number of reversals is high; not because the forecasts were wrong. Usually, a day on which many positions are reversed is followed by days when the forecasts are MORE ACCURATE because the the forecasting system has ADAPTED to the changing market conditions.

Now, for the day in question, the FORECASTING ACCURACY, when calculated the new way (as now being posted) would be

FORECASTING ACCURACY = 100 * 8/10 = 80%

and this does NOT mislead the trader into the conclusion that the forecasts are (or have been) mostly wrong.

Any user of ANY forecasts (from this source or any other sources) needs some indication of FORECASTING ACCURACY that can be used (without confusion) for determining how much reliance can be placed in the current forecast.

I continue the quest for a better way to calculate and report FORECASTING ACCURACY. Readers are requested to please study this situation carefully and give me your critical analysis and comments. And, please remember that my goal is to provide something that is easy to understand and easy to use for traders and investors.

FOR MORE INFORMATION --

To obtain a copy, send a check ($25.00) to:

	  PRICE FORECASTS
          4300 Castle Ridge Court
          Charlotte, NC 28269