Real Data set Results
Data set 4 had some interesting results, as the martingale algorithm had captured more anomalies than the yahoo marked points. The last point was not captured as it needed the next martingale following the data series to see the change.
The test martingale shows the various changes noticed in the data set as each point was added.
Data set 7 shows the martingales ability to adapt quickly to the change happening. The right side has not been marked as the assumption is that the strangeness of the point is not strange if it had occurred previously.
Data set 8 shows the ability to capture a variety of different height anomalies, that varied throughout the data set. Note that this is much easier to do when the values of the domain are very large.
Dataset 9 was a simple example to illustrate that the variances between anomalies can be quite different.
Synthetic Data set Results
Clearly the Martingale Framework is unable to cope with data that moves in waves
This is the actual sample that should have shown in the above processed data set