Details of the Predictions page

Description

Numerical Stock Price Prediction, like Numerical Weather Prediction, uses historical measurements to detect trends and cyclical behavior and to estimate possible future conditions. These are different for each stock, just as the weather forecast is different for each city. In either case, the forecast may not be valid, that is, the numbers in the forecast may not match the numbers actually observed.

Validation compares a forecast made earlier with an observation on the date at which the forecast claims to be valid. To test a forecast, you need to know the claimed precision of the parts of forecast. The most important part is the date valid. You can't test a forecast that has no date. For example, if an analyst says that he has a 'target' temperature of 75 degrees for Dallas, but doesn't say what day he is talking about, he is wasting your time. Here the date range is the five trading days nearest the date valid.

Similarly, the analyst must supply (or imply) a range for the second part of the forecast, the variable that is being forecast. For variables with a normal distribution, this usually is stated as a target plus or minus one standard deviation, say, 75 degrees +/- 5 degrees. This means that, on the average, for more than two thirds of such forecasts, the observed temperature at Dallas on the date valid should lie between 70 and 80 degrees inclusive.

Sadly, weather is not normally distributed, and neither are stock prices. However, the logarithm of a stock's price is more nearly normal, so one can specify an asymmetric range, effectively a target 'multiplied by or divided by' some number. This is the meaning of the three prices quoted: the outer two bracket a probable range around the inner price. For example, $ 29.40 | 33.40 | 37.94 means that, on the average, for more than two thirds of such forecasts, the price lies nearer to $33.40 more often than it is less than $29.40 or greater than $37.94.

We have verified about 25,000 old forecasts to measure the Skill for each stock, the number of good forecasts (actual price within price range, within date range) divided by the sum of good and bad forecasts, times 100%. This appears on the Predictions page after the forecasts that have not yet reached their valid date. We collected about ten forecasts for each of about 2,500 stocks to test their statistical properties, as described in the 2010 DiligentInvestor Blu-ray video animation. An example of the prediction process is shown for a single stock in a clip from that Blu-ray in the Prediction Chart 3.2*. Its typical result is summarized in the Verification Chart 3.22

The question as to whether an old Skill applies to a new forecast is harder to answer. That's the reason for the Predictions page, so you can look at this question yourself. Please type in the Ticker Symbol for any stock, and if it is in the dataset, make note of the date valid and the price range. You should not have to wait more than a few months to collect useful results.

Derivation

You may have seen the forecasts available on the Screener page. These raw forecasts appear whenever cyclic behavior is strong enough to dominate a trend, and go away when the cyclic behavior weakens. The curated forecasts shown on the Predictions page start with these purely cyclical forecasts, but treat them as a departure from the trend. When the trend changes, the curated forecast changes also, to maintain the departure and improve its accuracy. The curated forecasts continue to be updated, even if the initial cyclical behavior disappears for a while. If cyclical behavior resumes, another raw forecast may appear with nearly the same time valid. In this case, the curated forecast merges with the raw forecast, and the result has a smaller price range, increasing its utility.

Disclaimer

These Numerical Stock Price Prediction tools are new to the individual investor community, so there has been little opportunity for public comment. Therefore, you should check any forecast dates and prices from this page against other methods. We believe that the dates and prices for each stock accurately reflects the numerical calculations designed into the software. However, DiligentInvestor cannot guarantee correctness, utility, fitness or performance. We have no affiliation with any investment firm and we are not registered with or certified by any public authority. We do not provide professional investment advice.

Use of this page is free. DiligentInvestor accepts no liability of any kind for any problems that may arise from its use. We do not recommend the sale or purchase of any stock. We make no warranty that these calculations bear on future performance of your stocks. You can lose money when trading stocks. If you do not have experience with investing in the market, you should not use these data. All risk is yours alone, without remedy.

*The 2010 Blu-ray disc is full HDTV 1920x1080i at 59.94 fields per second. Your display needs to have sufficient resolution to see all of this content. This 56-second MPEG-2 clip is therefore quite large, 167 MBytes, so transmission times may be very long. The sound track is empty.

© Copyright 2007-2010 DiligentInvestor