WOTLK - Fury Warrior - Odd Weapon Swap


With the above snapshot, it is telling me to swap Black Ice for Sword of Justice, which doesn’t seem right.

I put this into wowsims as a cross comparison and compared it to what I currently have equipped and it is indeed a DPS loss not a gain.

Perhaps I’m missing something here?


Seems like they must be very close, at the least. Agility isn’t much of a stat for fury. A slightly lower dps weapon in the off hand with better stats sounds believable to me.

I haven’t used wowsims, so I don’t know how accurate it is or what assumptions/settings are being used. I would have to see all the settings used to verify an apples to apples comparison. In general, our model picks all the gear one would expect - so it would be odd for it to be off in one specific case.

Thanks for the reply.

In wowsims I have set everything to be self buffed, ie food, flasks etc. There’s some other settings in particular to the fight which is duration in which I have matched Mr. Robot.

Running the sim based on Mr. Robot reveals a DPS of 3053.20 + or - 186.99DPS over 3000 iterations.

Manually adjusting my own gear in wowsims to the below:
Main-Hand: Black Ice
Off-hand: The Jawbone
And adjusting a few gems nets me 3132.42 + or - 178.90 using the same settings.

It’s not much of a difference to be honest, but I am curious why there is a difference and what is actually better.

Thanks again

There is a difference because these are two different models of the game. We aren’t using simulation, we are using a formula-based model.

For a difference this small, it is unfortunately not really possible to verify for sure which model is better. This is because it is impossible to set up an in-game test that could measure this for us and get enough controlled trials to overcome statistical error.

Since there is no way you’d ever be able to tell which of these setups is “better” just by playing the game… Our position is that you are safe using either model to pick your gear.

We use a formula-based model because it gives us a lot more flexibility than a simulation model. It can calculate solutions way faster, which then allows us to add more options to the website to let people adjust the gear advice to their preferences. We also build the models to smooth out noise and jumpiness in the results that are artificial and due to the model itself.

Simulation is not good for comparing one setup to another. Simulation is good for generating large data sets and then aggregating the results to find statistical trends. This is because there is a lot of “noise” in simulation data, and we have no way to measure the structural error inherent in the model. We have been arguing for years that the way the community uses simulation to compare one set of gear to another is an inappropriate (mathematically) use of simulation.