A DPS Simulator 2 - SimFF

[KroLeXz]

I wonder if devs read this thread and chuckle to themselves when we get closer and closer - Keep up the good work guys.

[Dervy]

It's 0, or could be 1.

http://puu.sh/jti0V/a0bf0e75cb.png

Then Broken Weapon Min/Max Values were

MAX: 52
MIN: 47
AVG: 49.5

I guess it kinda fits 1 a bit better.

I'm doing more testing anyway, so I can confirm it again with a much larger spread of Weapons, going up to 72 WD.

[pandabearcat]

I am on vacation starting today.

What would be very helpful is if...

1. People went and tested the sim damage output numbers, and if they actually correspond to same numbers ingame.
2. People went and put in whatever hardcoded rotation, and tested it to see if it is what you would expect in game. The priority rotation right now is something like 1013 dps, which is pretty decent w/o food buff, I think it was simmed with ilvl 179 gear and hive weapon.
3. People test if they can actually make a hardcoded rotation. It should not be hard with the new macro system.

See you guys in 1 week.

[KroLeXz]

Dervy I have a new parse with different gear if you need it

https://youtu.be/xpqQABzzOuw

[enil]

Here's a quick one off parse I did.
http://i.imgur.com/arNvXnP.png
And my modified version of the base script. (modified rotation and my current stats)
http://pastebin.com/QWv05mM6

One thing I notice directly comparing the numbers - what did you set leg sweep cd to? there's a significant difference between the parse and sim (for reference leg sweep is 20 second cd).

[pandabearcat]

Idk. I was probably not paying attention while writing the numbers down. All the more reason to dump all that data into a spell database file so people can change it when I make dumb mistakes.

I don't have the source in front of me so I cannot check.

Numbers seem off. Prob have to revisit formulas and/or check to make sure how stacking buffs works.

Relative damage contributions seem actually fairly okay, so rotation and/or general potency check is about right.

[Kenji1134]

I found the neural network toolkit in Matlab. Fed it my data, trained it, got very low errors... so now how do I get a meaningful equation of out it?
It seems like all the neural network does is train itself in order to process similar data. Is there any way to get the kind of equation we're looking for out of that?

[pandabearcat]

http://www.mathworks.com/matlabcentr...nction-for-gen

I didn't read this cause if I was doing it I would just feed it into the neural network every time =D but this is what you want probably.

[Kenji1134]

But that kind of defeats the purpose doesn't it?
We can already make models which match the inputs to under 0.5%, but they are only good for the input range, and maybe another 3-5% outside of the input data.
Granted since it is not overly difficult to feed new data into the model, we could continue to update the model to obtain stat weights.

But the real equation would be so nice to have. =)

[pandabearcat]

Not sure what the question is.

If you mean "can we somehow guarantee we will have the right formula, the one the devs use" the answer is no.

If you mean "can we get a formula that will fit the data perfectly" the answer is yes.

In reality, there is no difference between the two as long as we can test for the accuracy of inter/extrapolation.

Its sort of how you can get the taylor expansion of any function, there is a polynomial approximation with some error.

You want less error, you put in more terms (in our case, perhaps more hidden layers, and more data to train).