Engineering Professor (EngProf6)

Mathematical Model of Stocks and Indices

Some Comments About the Model (Short-Term Forecasting):

Introduction:

I am a professor of engineering at a research-intensive university. I have developed a mathematical model which I’ve back tested and am now testing in the field (i.e. live with real players and all the emotions that come with this). The model is based on price movements. It has no inputs to account for emotional decisions, earnings, announcements and the like. The inputs are simple and they require essentially no knowledge of the company (similar to reading a chart).

My analysis differs from much of the conventional technical (or trend) analysis in that I have created dimensionless numbers that, as far as I know, are unique. The chronological progression of these numbers in time is used to generate the results.

What the Model Does:

The model only forecasts the direction of a cycle. Forecasting the magnitude is out of the question (at least for me). When a reversal in a cycle happens (for example, the trend switches from DOWN to UP), the model picks this up and then projects what the remaining length of this new cycle may be (e.g. 15 days). Each day, the model is used to re-evaluate all positions. The normal progression of a cycle would entail a reduction of one day in the length of the cycle as each day passes. If the reduction is more than one, then the model is accelerating towards the perceived reversal point. If the reduction is zero or, in fact, there is an increase, then the model is decelerating the velocity towards reversal.

What I have found in my research is that in many cases, second-guessing the forecasts of the model can be ‘detrimental to one’s financial health’. Naturally, one needs to build up confidence in a model before one makes crucial decisions.

Readily available on the web are results from other models such as ‘americanbulls’, ‘stockpickreport’, ‘tradetrek’ etc.. They all aim to forecast short-term direction. You can do your own comparisons. In the end, I think you will find that my model is, on average, a bit more accurate.

The ‘Count Down’:

One distinguishing factor about the forecasts is the ‘count down’. Each day you have an idea of when the reversal will take place. Moreover, you also get, as the reversal approaches, a threshold value that needs to be crossed. For example, in the case of a reversal to the UP, the threshold value represents the price above which the stock will need to close the next day.

The ‘Fake-Out’:

One event that sometimes occurs is the proverbial ‘fake-out’. For example, the model forecasts a reversal and then within a short period of time (say 1 to 3 days) something happens to upset the cycle. The model picks this up and starts to reform the reversal, and, in so doing, it creates a ‘double-bottom’ (or ‘double-top’) as explained in the next section. I announce such events on the web page, however, I do not suggest that you, the investor, try to follow these second order oscillations. What I do suggest is that when it happens one can use this ‘second opportunity’ to buy (or sell) more to average out the price.

Not 1 but 2 Models:

While I refer to the model as one, it is in fact, 2 models. One model is used for detecting Entry points (i.e. UP cycles) while the other is used to find Exit points (i.e. DOWN cycles). The models are similar, however, there are differences. It would appear that we react differently when we Enter an UP cycle versus when we Exit. One major difference the models point out is that people tend to buy more easily than they sell. They are more easily swayed to buy than to sell. What this means is that it is easier to start an UP (i.e. BUY) cycle than it is to end an UP cycle (i.e. start a DOWN cycle). If this is true, then it implies that the model should be more sensitive when it’s deciding on the start of an UP cycle. Because of this, the model can be ‘faked-out’ more easily. That’s the basic reason why you will notice some ‘double-bottoms’ but few ‘double-tops’.

So now you have some idea of how it’s done. The 2 models are computer programs that use chronological price data to infer the embedded cycles. The results from the models can also be used to compare stocks to market averages in an effort to determine whether a given stock correlates with a given average. Such an analysis goes beyond what is normally found by computing a linear regression of one to the other.

Easy Comparisons to Markets:

My models focus on actual cycles and, in particular, the portions of short term cycles that reside in the ‘future’. For instance, we might find that company XYZ is in an UP cycle that will switch in 10 days. We may also note that the SP500 is also in an UP cycle that will terminate in 10 days. Such an observation can be interpreted to mean that in the current context, XYZ and the SP500 are correlating at least in direction. What is important to note is the distinction between comparing past historical trends between the 2 entities and what can be done with the model. When you use the model to compare a stock with an index, for instance, you are using the current term cycle and its extrapolation into the near future to make the comparison. In conventional comparison techniques, the past is used to establish correlations. Naturally, one can also extend this logic and the use of the model to compare stocks to each other or markets to each other.

The End:

I hope this has helped to improve your understanding of what it is that I am doing. Stay tuned for more in due time. Your comments are always welcome.