Method for categorizing pivot points in technical analysis
A method and product for identifying and categorizing pivot points in a time series, such as a time series of financial data (e.g. stock prices) or medical data (e.g. ECG, EEG). The method permits categorization of pivot points in a time series according to their spatial importance and temporal occurrence. The time series has a predetermined end point, predetermined maximum and minimum box sizes, and a predetermined box size increment. The method consists of determining a breakout direction for the time series. Then, for each incremental box size from the maximum to the minimum, pivot points are identified. Commencing at the end point and working backwards through the time series, any point where a reversal of more than the current box size occurs is identified. For each such identified point, the previous extreme, its associated lag from the end point, and the box size at which it is first identified is then recorded.
1. A method for generating a set of categorized pivot points from a time series of financial data for pattern recognition analysis for a user, the time series having time steps identified by a lag from an initial time step, the method comprising:
determining, by a computer, a breakout direction for the initial time step in the time series comprising a series of time steps covering a time period, wherein each time step is associated with a time, and wherein the initial time step is a most recent time step in the time period of the time series and a final time step is an oldest time step in the time period of the time series;
determining a maximum box size for the time series, wherein the maximum box size lacks intrinsic noise in the time series, wherein the intrinsic noise comprises a false pivot point that fails to represent a true change in direction in the time series;
determining a minimum box size for the time series, wherein the minimum box size is based, in part, on intrinsic noise in the time series;
determining a box size increment for the time series;
after determining the breakout direction for the initial time step, the maximum box size, the minimum box size, and the box size increment, decreasing, by the computer, a box size by the box size increment from the maximum box size to the minimum box size to generate a series of decreasing box sizes, and for each box size in the series of decreasing box sizes, incrementing a time step from the initial time step to the final time step, and for each incremental time step, by the computer
identifying a subsequent extreme value and subsequent end value for a subsequently considered time step based on the breakout direction and any previously identified reversals;
determining a current extreme value for each incremental time step by comparing a previous extreme value with a subsequent extreme value and: replacing the previous extreme value with the subsequent extreme value if the subsequent extreme value corresponds to the breakout direction;
maintaining the current extreme value if the subsequent extreme value does not correspond to the breakout direction;
determining whether a reversal in the time series has occurred, wherein a reversal has occurred when the subsequent extreme value deviates from the breakout direction and the magnitude of the difference between the current extreme value and the subsequent end value is greater than or equal to a current box size of the series of decreasing box sizes, and:
switching the breakout direction if a reversal in the time series is identified; and
maintaining the breakout direction if a reversal in the time series is not identified;
for each identified reversal, recording, by the computer, the current extreme value associated with the reversal, the associated lag from the initial time step, and the current box size associated with the reversal;
determining that the largest box size at which the associated reversal is not associated with intrinsic noise and is identified within a desired resolution for pattern recognition analysis;
categorizing, by the computer, the pivot points by the largest box size at which the associated reversal is identified and the associated lag from the initial time step;
generating, by the computer using the pivot points by the largest box size at which the associated reversal is identified, the set of categorized pivot points for use as a filter in point and figure charting or as an input to a neural net;
determining, by the computer, a trading pattern by analyzing the set of categorized pivot points starting from the initial time step and ending with the final time step; and
presenting, by the computer, the trading pattern to the user.
2. The method according to claim 1 , wherein a perceived trend determines breakout direction.
3. The method according to claim 1 , wherein a pattern of interest determines the breakout direction.
4. The method according to claim 1 , further comprising a step of normalizing the time series.
5. The method according to claim 1 , further comprising a step applying a mirror image technique to the time series.
6. The method according to claim 1 , wherein the time series is a series of financial data and the financial data comprises price-bar information comprising an end point, the end point being the most recent price-bar information for the time series.
7. A computer-readable medium storing computer-executable instructions that when executed in a computer programming language cause a machine to perform the steps of:
determining a breakout direction for an initial time step in a time series of financial data comprising a series of time steps covering a time period, wherein each time step is a value associated with a time and wherein the initial time step is a most recent time step in the time period of the time series and a final time step is an oldest time step in the time period of the time series;
determining a maximum box size for the time series;
determining a minimum box size for the time series, wherein the minimum box size is based, in part, on intrinsic noise in the time series, wherein the intrinsic noise comprises a false pivot point that fails to represent a true change in direction in the time series;
determining a box size increment for the time series;
after determining the breakout direction for the initial time step, the maximum box size, the minimum box size, and the box size increment, decreasing a box size by the box size increment from the maximum box size to the minimum box size to generate a series of decreasing box sizes, and for each box size in the series of decreasing box sizes, incrementing a time step from the initial time step to the final time step, and, for each incremental time step;
identifying a subsequent extreme value and subsequent end value for a subsequently considered time step based on the breakout direction and any previously identified reversals;
determining a current extreme value for each incremental time step by comparing a previous extreme value with a subsequent extreme value and:
replacing the previous extreme value with the subsequent extreme value if the subsequent extreme value corresponds to the breakout direction;
maintaining the current extreme value if the subsequent extreme value does not correspond to the breakout direction;
determining whether a reversal in the time series has occurred, wherein a reversal has occurred when the subsequent extreme value deviates from the breakout direction and the magnitude of the difference between the current extreme value and the subsequent end value is greater than or equal to a current box size of the series of decreasing box sizes, and:
switching the breakout direction if a reversal in the time series is identified; and
maintaining the breakout direction if a reversal in the time series is not identified;
for each identified reversal, recording the current extreme value, the associated lag from the initial time step, and the current box size associated with the reversal;
determining that the largest box size at which the associated reversal is not associated with intrinsic noise and is identified within a desired resolution for pattern recognition analysis;
categorizing the pivot points by the largest box size at which the associated reversal is identified and the associated lag from the initial time step;
generating, using the pivot points by the largest box size at which the associated reversal is identified, the set of categorized pivot points for use as a filter in point and figure charting or as an input to a neural net;
determining, by the computer, a trading pattern by analyzing the set of categorized pivot points starting from the initial time step and ending with the final time step; and
presenting the trading pattern to the user.
8. The computer-readable medium according to claim 7 , wherein a perceived trend determines the breakout direction.
9. The computer-readable medium according to claim 7 , wherein a pattern of interest determines the breakout direction.
10. The computer-readable medium according to claim 7 , further comprising means for normalizing the time series.
11. The computer-readable medium according to claim 7 , further comprising means for applying a mirror image technique to the time series.
12. The computer-readable medium according to claim 7 , wherein the time series is a series of financial data and the financial data comprises price-bar information comprising an end point, wherein the end point is the most recent price-bar information for the time series.