Investigate off-by-one error

[ehmatthes]

In preparing slides for a recent talk, I realized the algorithm for finding critical points is not implemented correctly. It's not examining all the preceding points that could make a point critical. The current implementation is valid, because that implementation is what's been evaluated. But, I should clean up that algorithm and then re-run the analysis.

• Look at code for finding critical points. Update to look backwards from current point, not forwards from 5 hours ago.
• Re-run analysis to determine optimal critical factors.
• Re-evaluate confusion matrix.
• Update real-time tool to incorporate these changes.
• Communicate this change.