Stable Schedule Ordering + Tooltip Fixes

Stable Schedule Ordering

The auto-scheduler now schedules each epic in a consistent order. This makes interpreting the data much more easy.

Previous Behavior

For each run of the monte-carlo simulation, the autoscheduler would probabilistically select the sizes of each epics and then run the scheduler algorithm.

This could result in WILD differences between runs in when an epic took place. This made the results harder to understand.

For example, lets say you have some epics and their dependencies like:

A.1
  \ A.2
B.1
  \ B.2
C.1
  \ C.2

This means A.1 blocks A.2. If all of these epics are about the same size and uncertainty, the autoscheduler shows something like this:

This is because 1/3 of the time, it runs A.1 first, 1/3 of the time it runs B.1 first. This causes the start and end dates to be very spread out.

New Behavior

The new behavior is to determine the scheduling order once, and repeat that scheduling order, but with "random/probablistic" timing for each epic.

This means that if, on the first pass, it's determined that A.1 is scheduled first, it will always schedule A.1 first, even if A.1 is not on the critical path.

The first pass uses epic timing with 80% of their probabilistic timing.

The result is something MUCH more what folks would sort of expect ...

Here it shows A.1, then A.2, then B.1, etc in a nice order .... but the "error bars" extend out because of the uncertainty in what happens before ...

Tooltips

Tooltips show above the bar if you're at the bottom of the viewport.