A DPS Simulator 2 - SimFF

Quote:

---

Originally Posted by pandabearcat

2. I need to know the general damage "spread" % (instead of an attack always doing the same damage) so we can do more realistic sims. Statistically it won't make much difference in the average but it would obviously affect the deviations of data and can show general lower and upper bounds of damage more easily.

---

It's a uniform distribution or close to it. I don't know what Dervy's on. The standard deviation in that image should make it pretty clear--sqrt(1/12*(263-238)^2) = 7.22.

For what it's worth, I could never explain to my satisfaction why the standard deviations I have from actual data seem to indicate a greater spread than the minimum and maximum values observed. I'm not a statistician by trade.

And yeah I actually have no idea where the rounding is or how it works. Nobody seems to. Some data wants to be restricted to 4.97% spread. Some wants to be 5.1%.

EDIT: from a small experiment that took about 2 minutes, the standard deviation from actual data being larger than expected can probably be explained by the data being discrete instead of continuous. Don't know why I didn't test that idea sooner.

Quote:

---

Originally Posted by SunnyHirose

And yeah I actually have no idea where the rounding is or how it works. Nobody seems to. Some data wants to be restricted to 4.97% spread. Some wants to be 5.1%.

---

Well, one thing I can comment on is that the x1.5 for critical hits occurs against an unrounded number. That is, I have a data set that can't work if the formula is "figure out base damage, round it, multiply by 1.5 for crit, round it again."

I do too. There are some tests I've done where crits attested to a maximum or minimum value that would round to a value non-crits never reach. Although it's very rough to test this way intentionally, it's good when it happens because the value is bigger, thus the error is smaller.

If, and this is a strong if, rounding for damage works the same way for tooltips, couldn't you get a Strength value where if it's multiplied by 1.03 (party buff) and the 10th decimal would be above 0.5 to be able to see if it rounds up, or just cuts the decimal off?

Party buff is 1.03, right?

http://puu.sh/immZD/5fa63d1e2c.png

Would the potential of the decimals being completely cut off, fit your data? It was just something I observed when I was playing around with my chicken, lol.

So, is this where I ask how we're going to figure out/handle the fact that potency doesn't appear to be a straight multiplier? Or rather, that there appears to be some subtraction happening after the multiplication? >_>

Just to expand on this, take this section from the OP:

Code:

---

[12.66]        True Thrust        (560.59)
[12.66]                        LNC_TRUE_THRUST_COMBO applied to P1        (10sec)
[12.85]                Chaos Thrust (DoT)        (118.91)
[12.85]                Phlebotomize (DoT)        (101.93)

---

Everything here is under B4B buff, and Disembowel is affecting the TT but not the dots, right? 560.59/1.1 to account for Dis puts the B4B damages at 509.62 for TT and 101.93 still for Phleb. TT is 5x the potency of Phlebotomize, so you would think this checks out, but I'm betting it doesn't in-game. I'm betting that the Phlebotomize center point, even after accounting for buffs, will be 1-2 points below what you get if you divide the TT center point by 5, based on my NIN damage numbers.

Quote:

---

Originally Posted by SunnyHirose

For what it's worth, I could never explain to my satisfaction why the standard deviations I have from actual data seem to indicate a greater spread than the minimum and maximum values observed. I'm not a statistician by trade.

---

There are many options. We can't say where rounding happens (if at all).

The game itself could also work with the min-value for example, and just multiply *1.1 to get the max (one less operation). Which imo also makes more sense, one less calculation, and for balancing using the min seems pretty necessary.

Quote:

---

Originally Posted by Dervy

Would the potential of the decimals being completely cut off, fit your data? It was just something I observed when I was playing around with my chicken, lol.

---

I've only tested it a little before, but it appears to be just like VIT and PIE.

Oh yes this is the answer to counter parser domination? Considering how many parsed logs is the most reliable form of information and how many times wow's version of sim craft has been wrong I am not sure this will gain any ground. A more advanced parser would be welcomed. Se seems to use internal parsing over sims it least from the statements they talked about.

I'm not sure you understand the point of a simulator.

If you are going to think simcraft's standstill, patchwerk, single target fight sims are going to supposedly be gospel vs a dynamic, mechanic laden boss fight parse, then you are going to be disappointed.

Then again, based on how many people think that their dummy parse is useful, sims are far superior to that.

The point is to have a baseline for performance, to settle those age old questions of "my opener is better than yours" and to determine optimal rotation, not to say "oh the sim says monks do 2% more damage than dragoons, but on this fight i parsed 5% higher than my buddy, hahahhaahaha the sim is wrong". Also to settle the "should i go crit or det?" questions. Not only are they more accurate than a model or anecdotal evidence, they can be personalized to YOUR rotation and YOUR stats.

Panda , i`m interested in helping you with this, if you still need help, anyway that i can contact you ?