The Real Problem with Parsers

[JackFross]

This is a discussion for another thread, tbh, but out of curiosity, how many data points are you getting where you find those min/max values? You only seem to record those two rather than the full range, so I'm just curious how many times you're using these longer cooldown skills like Sneak Attack.

And the fact that it's vaguely NOT perfectly linear doesn't change the fact that, for all intents and purposes beyond *precise* calculation, it's linear "enough" to use ratios between potencies as direct correspondences to ratios between damage outputs. You're splitting hairs to prove my discussion of optimization wrong when none of your data does anything to disrupt the things I discussed in any notable way.


I don't deny that it's wonky, but the variances in the linear trend of the potencies don't matter *enough* to make the practice of using potencies as approximations of damage ratios ineffective/incorrect to any reasonable extent.

(so optimal rotations can still be reasoned out through mathematics performed on values provided in-game, without the use of a parsing program to record your actual damage output)

[Viridiana]

This is a discussion for another thread, tbh, but out of curiosity, how many data points are you getting where you find those min/max values? You only seem to record those two rather than the full range, so I'm just curious how many times you're using these longer cooldown skills like Sneak Attack.

As many as it took to get the ratio very close to 21/19 for max/min. Not a big deal for, say, the Mutilate numbers, which I could usually get in about 30 seconds or so due to the low spread. Getting the SA numbers…let's just say there's a reason I don't also have TA numbers there. But to speak to the accuracy of the SA numbers, my strategy for determining accuracy is this. Take max/min (1593/1440) and subtract it from the 5% spread ratio (1.05/.95, or 21/19). If you get a negative number, you almost certainly have the full spread, because the spread is already bigger than ±5%. If it's positive, take the reciprocal and compare to the max value. In this case, that's 1826.5 vs 1593. I'd say that the spread is accurate to one point in about 1827 and so I wouldn't expect another data point.