Omicron Research Institute

QuanTek Econometrics Software

QuanTek Main Graph

One of the most important features of QuanTek is its Main Graph, with five different scales. The scales are labeled by the number of pixels per data point, and these are scale 1, 2, 4, 8, 16. On each scale the entire 2048-day set of data is displayed, in multiple views spanning the whole range, but these are views of different sections of the same graph. The graphs show logarithmic prices rather than actual prices, so equal intervals on the vertical scale correspond to equal percentage changes in value. Since the data are corrected for splits & dividends, you can get a picture of the true value of the security as it has evolved over the past 2048 days (8 years). This is also the type of data preparation needed for the Adaptive (Wavelet) Filter to operate and make a meaningful Price Projection.

These scrolling graphs also have the property that you can move to any part of the graph by scrolling and switching views. (The blue toolbar arrows scale the graph up or down, while the magenta toolbar arrows switch between different views along the horizontal axis.) When you scroll the horizontal scroll bar, the vertical scroll bar moves automatically to keep the center line of the graph centered in the display. You can also adjust the vertical scroll bars yourself, if necessary. So by scrolling and switching views, you can move to any part of the 2048-day graph, which forms a single graph view of the past 2048 days of corrected prices.

Note: The scale on the right shows the prices for stocks, so long as they lie in a reasonable range. However, if the price range is too high or too low, or it is an index or mutual fund, then the price scale is replaced by a symbolic scale with the current price indicated by dollar signs. The price for all securites at any point is also indicated by tooltips. If you move the mouse pointer anywhere in the graph, it displays the date and price corresponding to that point. So you can find the price at any point in the graph just by moving the mouse pointer to that point, no matter what the range of the price.

Scale 4 (2019-01-04)

We begin with Scale 4, since this scale is displayed initially when a graph opens. Here is a graph of Amazon:

AMZN_Amazon (Scale 4)

On this scale you see the 512-day (acausal) Savitzky-Golay Smoothing Curve (yellow curve in center). The "acausal" means that at any point, the smoothing mixes up both past and future data -- the smoothing is over the whole graph at once. Hence as time progresses, the smoothing curve at past points can change. But this means that there is also no time delay in the smoothing curve. This smoothing curve is also projected ahead N days, where N is the Time Horizon setting for the calculations (here, 32 days). (The present time is denoted by "ZERO".) So it provides yet another kind of Price Projection, abeit a simple one.

Surrounding the S-G Smooting Curve are the Bollinger Bands. These are spaced from the S-G Curve by 1x and 2x the average absolute deviation of the prices from the 512-day S-G Smooting Curve. This is the average trading range to be expected normally, in the absence of any exogenous events. These curves also serve as overbought/oversold indicators. In the above graph, you can see that the price in September appears overbought, ranging above its outer Bollinger Band, and then at the end of December it undergoes the selling climax and becomes oversold, ranging below its outer Bollinger Band. Presumably, due to the return to the mean mechanism, the price will tend to return back to the central S-G Smooting Curve.

In blue on the right is the Price Projection, due to the Default Adaptive Filter. (You can also calculate a second filter of your choice and toggle the display back and forth between the two for comparison.) The calculates the expected N-day return and displays it as a straight line, starting from the average price of the latest trading day (average of open, high, low, close prices). This expected return is also displayed in the header information. The Default Aptive Filter used is a Least-Mean Square filter, which is a low-pass filter. Hence any features shorter in duration than N days (in this case, 32 days) does not have much effect on the Price Projection. Generally, but not always, the Price Projection is an indicator of the long-term trend. This can be seen above since the Price Projection is approximately parallel to the projection from the S-G Smooting Curve. The fluctuations shorter than N days are averaged out and only the long-term trend remains. When presented as above, the Price Projection is close to what the Random Walk model would predict. (Except the Random Walk model does not make clear how the long-term trend is to be computed -- it takes the long-term trend as fixed a-priori.)

The blue bars on the Price Projection are error bars. Just as the Price Projection represents the expected N-day return, the error bars represent the expected N-day range of prices. This is obtained by measuring the average absolute deviation (1-day range or high minus low) of the log prices, and then multiplying this by the square root of N, for N future days. (This is only an estimate. There is also a direct measurement of the average N-day range available on the toolbar, which usually turns out to be smaller than this estimate, so it is a conservative estimate).

Also notice the green arrows which are Buy Points. There are also red arrows above the graph which are Sell Points. These correspond to the Relative Price (band-pass) display which is a type of oscillator. So these arrows represent N-day inflection points in the Relative Price graph. But the Buy Points are only shown when the Price Projection is positive, and the Sell Points are only shown when the Price Projection is negative. These Buy/Sell Points are useful for swing trading. They also depend on the setting of the Range control in the Trading & Portfolio Parameters dialog.

The graph along the bottom is the logarithmic volume, relative to its average value. The bars range from the average (log) value to a factor of 10 above and below the average value of the log volume. (Since the logarithmic values are plotted, multiplying and dividing the volume by 10 corresponds to the same distance above and below the average log value.) The volume has been extended to future values using the Standard LP filter, mainly for aesthetic reasons. (The number in the lower right corner is the size of the data set. However, only 2048 days are used.)

Scale 8 (2019-01-04)

Here is the graph of Amazon on Scale 8. Both horizontal and vertical axes are doubled in size from Scale 4, preserving the aspect ratio. In other works, the slope of all graphs is preserved, enabling direct comparison of N-day returns over any interval:

AMZN_Amazon (Scale 8)

Notice how all the slopes are preserved between the two graphs. But all the features are twice as large. You can compare the risk of different securities by comparing the width of the Bollinger Bands, but you must do this with all securities on the same scale.

The main feature that distinguishes this graph is that the daily prices are represented by candlesticks. These display the price range from low to high by a vertical line. The open and close prices are represented by the upper and lower edges of the rectangles. If the close is higher than the open, the rectangle is light blue, and if the close is lower than the open, the rectangle is dark blue. This provides a way to display the open prices, which are not displayed in the more usual method as in Scale 4.

In Scale 4 the Buy/Sell Points were displayed as green/red arrows. Scale 8 does not display these, but instead displays Buy/Sell Signals as green/red rectangles. The difference is that the Buy/Sell Signals are derived from the Relative Price (low-pass) display which is a type of overbought/oversold indicator. It corresponds closely to the price level relative to the Bollinger Bands on the graph. There are no green/red rectangles on this particular graph, but in other areas of the graph they may be seen just below/above the range of prices. This is because they indicate the optimum buy/sell levels for GTC buy/sell orders. The level at which these are set depends again on the Range control in the Trading & Portfolio Parameters dialog. Thus, despite appearances, the prices in the above graph, on a 32-day average, are actually within a normal trading range and hence do not trigger the buy/sell signals. On a shorter time scale, or with a different setting of the Range control, the buy/sell signals might be displayed at the price extremes of the above graph. (They can also be displayed in the future Price Projection by means of the future projection of the Relative Price indicator, via the Standard LP filter.)

Scale 16 (2019-01-04)

Here is the graph of Amazon on Scale 16. Both horizontal and vertical axes are doubled in size from Scale 8, preserving the aspect ratio. In other works, the slope of all graphs is preserved, enabling direct comparison of N-day returns over any interval:

AMZN_1190104

As in Scale 8, the daily prices are represented by candlesticks. These display the price range from low to high by a vertical line. The open and close prices are represented by the upper and lower edges of the rectangles. If the close is higher than the open, the rectangle is light blue, and if the close is lower than the open, the rectangle is dark blue. This provides a way to display the open prices, which are not displayed in the more usual method as in Scale 4.

This graph is very big and I thought it would be a good place to plot the actual historical Price Projections. In other words, the Price Projection for all days in the past is plotted on this graph for each corresponding day. For a given day in the past, the positive price projections are marked by a little green rectangle at the close price, and the negative price projections are marked by a little red rectangle at the close price. Then, for the N-day Price Projection, the projected price for each given day is shown by a double-triangle, green for a positive Price Projection and red for a negative Price Projection. These N-day Price Projections for each day are called the N-day Price Target for that day.

This can also be viewed as a representation of the Long/Short Signals, which are just the estimated returns from the Price Projections. The green/red rectangles mark the days for which the Long/Short Signals are triggered. These are also plotted as the N-day Projected Future Returns in one of the splitter windows. Whether these Long/Short Signals are triggered depends on the setting of the Threshold control in the Trading & Portfolio Parameters dialog. So the Long/Short Signals represent intervals in which the optimal position should be long/short, according to the Price Projection. Note that, once again, the expected return of the N-day Price Projection represents an N-day smoothed optimal position. The short-term price fluctuations in the graph do not affect the slope of the Price Projection very much, and hence the optimal position is also not affected very much by these short-term price swings.

Scale 2 (2019-01-04)

Going down now in scale from Scale 4, we have a graph of Amazon on Scale 2. Once again, the aspect ratio is preserved:

AMZN_Amazon (Scale 2)

Just as in Scale 4, the Buy/Sell Points are displayed as green/red arrows on Scale 2. However, on this scale, instead of displaying a 512-day Savitzky-Golay Smoothing Curve, the 2048-day (Robust) Trend Line is displayed. The slope of this line, denoting the average 2048-day return, is displayed in the header instead of the slope of the Price Projection. This is another, probably better, display of overbought/oversold conditions. In fact, according to this display, the recent sell-off and selling climax shows up merely as a correction from a very overbought condition back to the normal trend-line. Then the Bollinger Bands are drawn with respect to this long-term trend-line instead of the SG Smoothing Curve.

Another feature of this graph is the 200-day Simple Moving Average (in gray). This is such a popular indicator that we decided to include it. However, the 200-day MA is computed with respect to the long-term trend-line, not some other way. This means that, going to longer and longer term MAs, they will converge to the long-term trend-line. Here might be a good place to also point out that you can plot a wide variety of exponentially weighted moving averages, in whatever color you want. These are also computed with respect to the long-term trend-line. For completeness, you can also plot horizongal lines at any price level, in any color. (Note: If there are fewer than 2048 days of data, then that number of data days is plotted on the graph, and the trend line is over that number of days.)

Scale 1 (2019-01-04)

Finally, going down from Scale 2, we reach Scale 1. This is a panoramic view of the price action, with one pixel per data day. If your screen is 2048 pixels wide, you can see the entire price action at once going back 8 years:

AMZN_Amazon (Scale 1)

This scale has all the same features of Scale 2, except that there are no Buy/Sell Points shown. But the 200-day Simple MA is shown on this scale. From this bird's-eye view it appears that Amazon was very overbought and the sell-off and selling climax of the past few months merely brought it back to its normal trend-line. This kind of panoramic view is very useful to make such deductions, to avoid the confusion and hysteria that occurs when everybody is just looking at the short-term graphs (on a linear scale, without any perspective).

Go back to QuanTek Features