Overview

Turning points engine

The Turning Points add-on is based on the concept of recognizing local peaks and valleys where a price time series changes its direction from up to down (peak) and from down to up (valley).

A turning point (TP) peak is defined as the point where the high is higher than or equal to any other highs in the neighborhood of the point.

A turning point valley is defined as the point where the low is lower than or equal to any other lows in the neighborhood of the point.

The neighborhood of a point (Nhood) ranges from the point -Nhood bars to the point + Nhood bars, inclusive of the endpoints. Therefore, the total number of bars included in a turning point analysis is 2*Nhood+1.

The above definitions are illustrated by an example graph below.

On the example graph the neighborhood parameter is Nhood=5. That means that it takes 5 bars to the left and 5 bars to the right of each bar to establish that the central bar is either a turning point peak, a turning point valley, or neither of them.

Due to the fact that Nhood bars are required to the right of a central bar, the most recent turning point on the chart is always located at least Nhood bars away from the most recent bar on the chart. In other words, no turning points can be established within Nhood bars of the most recent bar on the chart.

The Turning Points add-on algorithm does not just blindly search for the first consecutive valley after a peak (or a first consecutive peak after a valley). It tries to optimize on turning points and locates such peaks and valleys that a human expert would choose by looking at the chart.

There is no requirement in this add-on for a peak to be higher than the previous valley, and for a valley to be lower than previous peak. Our experimentations in the lab have shown that if such a condition were applied, then potentially no turning points could be established on vast portions of data. These data segments would then be lost to analysis.

TPplot and statistics indicators

After the turning points engine has established the location of each turning point on the chart, a number of indicators can be built on top of them.

The TPplot indicator connects consecutive TPs with straight lines. The result is a zigzag line connecting consecutive TPs. This indicator enables you to visualize peak and valley locations on the chart. TheTurning Point Example #1.cht chart shows how TPplot is used.

Important note – Do not use the TPplot indicator in a trading system. That is, do not use it in a trading strategy or prediction. The reason is that TPplot looks ahead in time in order to accurately plot the true turning points. None of the other indicators in this add-on look ahead in time and therefore may be used freely in any trading system.

Most Turning Points indicators provide various statistics applied to a particular measure of TPs. The following measures between peaks and valleys are used in the add-on: horizontal distances (in bars), vertical distances (change in value), vertical distances (percent change in value), and the slope between two TPs or a TP and the current bar (expressed in degrees).

The following statistics are applied to these measures: mean (average), standard deviation, and the median value.

The statistical indicators are built using a moving window across data. For statistical indicators to work well, the size of the moving window must be large enough to provide a sufficient number of peaks and valleys in it. One rule of thumb is to use a window size that is at least ten times the neighborhood parameter. Generally, however, the larger the window the better. The effectiveness of trading systems using turning points indicators will be influenced by your skill in choosing the most appropriate window and neighborhood sizes, or your skill in using the NeuroShell optimizer to do so.

Peak Probability and Valley Probability indicators

The most interesting statistics are the mean and standard deviation of the percent change in value from peak to valley or from valley to peak. The mean value shows the average of the percent changes. The standard deviation shows how tightly all percent changes cluster around the mean value. Once these two characteristics of the percent change are known in the lookback window, it is possible to predict the probability that the most recent bar is at a new peak or at a new valley.

The way this prediction is done is based upon the statistical properties of the mean and standard deviation in a normal distribution:

If the percent change in value from the most recent valley to the most recent bar equals the mean of all of the peak-to-valley percent changes in the window, then the probability that the most recent bar is at a peak is 0.5.

If percent change in value from the most recent valley to the most recent bar exceeds the mean of all of the peak-to-valley percent changes in the window by 2 standard deviations, then the probability that the most recent bar is at a peak is 0.977.

Levels less than 2 standard deviations above the mean will produce probabilities between .05 and .977.

The same logic applies to computation of probability that the most recent bar is at a valley. The Peak probability (% change) and Valley probability (% change) indicators give you the probability that the current bar is at a peak or at a valley, based upon the methods just discussed.

In addition to probabilities based on percent change, we have indicators to provide probabilities based on horizontal change (bars). These are Peak probability (bars) and Valley probability (bars). These work exactly like the probability indicators discussed above, except that probabilities are established by the mean and standard deviation of the number of bars between previous peaks and valleys.

Obviously, if the probability becomes high that the price is close to a valley, it is time to buy. If the probability becomes high that the price is close to a peak, it is time to sell. The Peak and Valley Probability indicators can thus be used in a trading strategy, or even simply viewed as objective indicators to confirm your own subjective feelings.

Important disclaimer - Both Peak probability and Valley probability indicators assume that the distribution of percent change values is normal. This may or may not be the case. Even if a stock or other issue shows regular cyclic tendencies, a major news event can disrupt the cycle. As with all statistics and market predictions, use of our indicators will not result in predictions with high scientific precision or accuracy.

Support, Resistance, and Fibonacci retracement indicators

Turning points form certain support and resistance levels on a price stream. In this add-on a support level is defined as the value of a previous turning point valley. A resistance level is defined as the value of a previous turning point peak. The Support and Resistance indicators can display not only the most recent turning point level, but the level of any turning point in previous history as well.

When support and resistance levels from a previous peak-to-valley or valley-to-peak segment are known, it is possible to draw a Fibonacci retracement level. The Fibonacci retracement level is a certain price level between a previous peak and valley on the price time series. One Fibonacci theory is that the price may stay for a while around a Fibonacci retracement level before taking a larger jump or drop.

Using the other indicators

We have provided many other indicators in this package that produce various statistics that are more basic then the Peak and Valley Probability indicators. For example, they produce many statistics like mean, median, and standard deviation of the three measures of vertical and horizontal change, as well as slope. These are provided as basic building pieces for our users to construct complex strategies for buying, selling, and prediction.

For example, you might feel that for your issues, the Peak and Valley probability indicators might not be as effective as predictions based upon the slope of bars since a peak or valley. For that, you can use the Peak to Valley (PV) or Valley to Peak (VP) indicators to compute the mean and standard deviation based on slope (PVmeanslope, VPmeanslope, PVsdslope and VPsdslope). You might then look at the current slope since a turning point (TPslope) and make judgments about future peaks and valleys based on how far above or below the mean the current slope is.

If you don’t care to work as hard as we described in the previous paragraph, you can simply find combinations of these other indicators that make good predictions in the Prediction Wizard or our other add-ons like Cluster Indicators, Adaptive TurboProp 2, Adaptive Net Indicators, Neural Indicators, and the Fuzzy Pattern Recognizer. Of course, the optimizer will come in handy there. (When using the optimizer, be sure to set the parameter ranges that you believe are appropriate – do not depend on our default ranges!) There are so many possibilities that there is plenty of room for research, and in our testing we have only scratched the surface of this tool.