Event timestamps on logs occur in batches of roughly 45ms. As a result, for most GCD speeds the time difference between two consecutive GCDs will never actually match the
tooltip GCD value - for example, a MNK with 1.93s tooltip GCD speed will generally have logged time differences of roughly 1.915s or 1.96s in between GCDs.
However, the ratios at which these "wrong" values occur give us pretty decent insight in the underlying real GCD speed at play, i.e. a player at 1.93s should see more GCDs that
are 1.915s apart and fewer that are 1.96s than a player at 1.94s tooltip GCD speed.
The exact method used is the following:
While your tooltip GCD value ingame only has two decimal places, your actual GCD speed is the result of not only that tooltip value but also how much time you're losing while
waiting for your client to render the next frame once your GCD comes off cooldown. As a result, given a sufficiently big difference in GCD speed losses from FPS, the effective
GCD speed of a player with 1.93s tooltip GCD can actually be slower than the effective GCD speed of a player at 1.94s tooltip GCD speed. This makes it rather pointless to try
to determine the tooltip GCD speed value and for a lot of jobs this tool should be sufficiently accurate to get a decent estimate of your effective GCD speed down to three
decimal places and could be used to make sure that measures you're taking to achieve good FPS alignment for your GCDs are actually effective.
Generally, the more GCDs using your real GCD speed can be found the better the estimate will be. For instance, Black Mage could be a bit off because they operate at a
number of differenct GCD speeds thorought the fight so there'll be fewer GCD samples taken into account than for jobs like MNK, DRG, tanks, etc. which just keep pressing
GCDs at the same speed all day. Stuff like Ninja mudras and Dancer steps shouldn't influence the results at all since they're never gonna be your most common actions.
For questions message Aya#0847