Edit: Beta is available now at this location - Reforge Calculator


This was discussed briefly in the Rogue 4.0.1 faq thread and very possibly in many other threads.

How do you optimize your gear using reforging?

Most classes have a "least desirable stat" and one or more "most desirable" stats. Those stats can change based on individual gear sets, hard caps, soft caps and possibly individual play style. The core concept of reforging is very simple, you can take one stat that's less desirable and convert 40% of it to a stat this is more desirable.

Things become more complex though when you try to decide - what's the very best variation of reforging I can select. For example: DPS casters will be shooting for 17% hit rating which is 446 hit rating. Suppose you could reforge two items to get 445 hit rating, is that the "best" you can do? Maybe it'd be possible to reforge 3 other items and get exactly 446. Or possibly there's multiple solutions that achieve 446 hit rating, but one of the solutions also let's you achieve at least a certain benchmark haste rating that gives you enough haste for an additional tick on your dots.

I'm working with a guy now to create a Reforge Calculator. It's in the Beta stage now (and you can PM me for access if you're interested in helping test). But we're looking for feedback on WHAT people are looking for in a reforge calculator. What do they need? What are you trying to achieve when you reforge?

The basics of the Calculator would include
1. Loading gear set from armory (which includes base stats) or possibly wowhead profile
2. A manual calculator / sandbox showing you each piece of gear, it's reforge-able stats and letting you manually click to reforge and see changes
3. A display of key stats from gear (expertise, hit rating, haste, etc.) and how each reforge changes those totals

the more Advanced features would include
4. Optimizer - you put in specific targets for ratings (say 200 expertise, or 446 hit rating) and press "go" to get a list of best options
5. Optimizer - provide EP weights for individual stats and calculator will reforge least desirable stats first to get specific targets on a rating, or maximize desirable ratings
6. Optimizer - Allow user to change gems, enchants, talents and see impact on stat ratings.
7. Optimizer - different EP weights for different levels of rating. (so like for rogue, you could set an EP for hit at below poison cap, above poison cap but below white cap, and above white hit cap)
8. Generate a link to a reforging profile so it can be accessed later.


The main point of this post is to generate a discussion on Reforging in general and the best way to reforge specfically. Most advice I've read on reforging consists of the basics: reforge your highest priority stat to a set target first, then reforge the rest to maximize your total EP on stats.

So is there a better way?
What else could a Reforge Calcuator do?
What else *must* it do?

This is definitely something that will be included. Right now, the calculator is assuming that a stat value can only be changed by a few key variables. Essentially:

stat = base + gear + gems + socket bonus + enchants + talents

Is there anything else that can increase the secondary stats? For instance, I read something about spirit affecting hit rating for some classes?



But yes, considering gems is going to a key part of getting the "best" solution. On my warlock, I needed 133 more hit rating and 10 blue sockets that needed to be re-gemmed. If I'd just followed the typical solution of "gem for hit til at least 446, then reforge" I would have been at least 7 hit rating above the hit cap. So instead I reforged 67 hit rating OFF my gear so that I needed exactly 200 hit rating and then got ten +20 hit gems in my blue sockets.

This was the scenario that made me first decide that an Optimal Reforge Calculator would be a huge help. It took me quite a while to come up with a solution to get exactly 446 hit, but it's truly a beautiful thing.

And since hit rating requirements are going to consistently change in Cata with each new Tier, I don't see this need for a calculator going away.

I can tell you that this approach is not optimal with cat dps, and possible other classes too. We're seeing EP values flip-flopping a lot with minor changes in one stat, and it gets worse at certain breakpoints. If you use static EP values, it's not going to work very well, at the least you should be recalculating EP after each reforge. Preferably, the fitness function should be purely based on calculated dps alone.

Can you give me an example of the EP values flip-flopping?

Are these defined breakpoints?

For instance, if it's just that the value of hit changes as you reach certain breakpoints OF hit, that's easily handle-able. The plan right now is to allow different EP values at different stat weights.
For instance, you could set

• hit less than 289 then ep = 5
• hit < 646 and > 646 then ep = 4
• hit > 646 then ep = 0

However, if you're saying (I'm making up this example) that say at certain levels of crit then Mastery > haste and at other levels of crit then Haste > Mastery. Then we probably would have a problem calculating an optimal reforge. Not sure, really. I think we'd need to see an example.

It would certainly be doable to say "maximize stat 1 to breakpoint #1 and if you can't reach that breakpoint then maximize stat 2 instead up to breakpoint #2". Would that suffice?


I would suspect that running a performance simulation for every spec (dps, tank & heals) would be outside of the scope of what the reforge calculator could do.

Well, to a certain extent, folks are going to have to make their decisions on what's their stat priority. I see what you're saying, but this calculator is NOT designed to figure out WHAT your stat priority is. It's design is that given a set input of stat weights, here's an optimal set of solutions.

But some additional questions on calculating the "optimal set" has crept up.

We're assuming you'd always want to reforge down your least desirable stat. Is that always the case?
We're assuming you'd want to always reforge as few items as possible. Is that always the case?


Assume your least desirable stat was crit and 2nd least desirable was mastery. Would you prefer reforging 5 items that each had crit or 2 items where one had crit and one had mastery?

What would sway decision between those two options?

I think you'd want to maximise a certain number (where X = a*Haste + b*Crit + c*Mastery) where a,b and c are the scaling factors derivated from Simcraft or anything other tool, while keeping your hit at cap...

I don't believe the number of items you reforge is something to be considered, especially since the cost isn't that high.

If say you priorize haste over crit, as long as hit is capped :

You could reforge the items that have haste + crit into haste + hit/crit, enabling you to reforge items that have crit + hit into hit + crit/haste... With gems, getting closer and closer to that magical cap number means there is a lot of possibilities to consider...

On my lock, as I have really bad trinkets, I bought the two hit trinkets from badges, then reforged the hit on both to haste, giving me passable trinkets until I get my hands on something from ICC, and getting me within 0.04% of the cap without having to socket +20 hit gems, thus getting me a little bit more haste/intellect

That is certainly not the case. If my weights are

Hit = 5
Crit = 2
Mastery = 1

and I have an item with

Hit 0
Crit 100
Mastery 50

The current pawn value is 250. If I reforge 20 Mastery->Hit, it changes to 330. If I reforge 40 Crit->Hit it changes to 370.

For fixed weights, you will always reforge to the best missing stat. You will reforge away from the stat that maximizes

Quantity * (NewEPperPoint - OldEPperPoint)

In my example the choices were 100*(5-2) or 50*(5-1).

As a first pass I'd recommend you stick with just looking at Reforging for an otherwise fixed set of assumptions (gems, gear, enchants, ...). That search space is not too large. If you have 14 slots and two "plausible" reforge choices per slot, then there are 3^14, or about five million possible combinations. Add just three plausible pieces of gear per slot and that number gets squared, to around twenty trillion.

For five million choices, you can afford to do an exhaustive search. At twenty trillion, you need to do some smarter optimization (or have a simple evaluation function, preferably with no break points).

Is there a reason not to just help out the Rawr team to add this functionality to Rawr? It already does all the other stuff and I can't imagine they won't be adding a Reforging function to the program. Reinventing the wheel and all that, eh?

Yes, static EP values entered by user.

For the first version you will have to enter 1) stat to get to a specified value 2) target value for that stat and 3) weight table.

E.g. you say "figure out a way to reforge my gear so I get to Experience exactly 200, and maximize gear value using the weight table".

This approach does not guarantee best theoretical build, but it does address one cap value. If there is much demand it might get improved to do optimization more intelligently.

There is a sector of users who care enough about their performance to go to a web site to try to do some optimization, but not enough to install and figure out rawr. I think that these users might be content with just a tool that lets them experiment with outfit (reforging in this case), even with no automatic optimization of any kind.

More importantly I think rawr is too complicated for me to get involved into, while reforger so far was a day of work.

Rawr would compare to MS Word, and reforger in question be a notepad - just enough to get work done... at least at the present stage of the project.

It is not very powerful yet, but I will be interested in any feedback I could get. Please give it a try, see what you think.

pugplug reforge calculator beta

Finding a solution for 2 stats is way more computationally intense than single stat. If I think of a way to find a solution within reasonable time frame - I surely can add that.

If any of the braniacs wanna help me with the algorithm - I will code it.