[Mage] Simcraft

[Deleted]

Hey,

Pretty much throughout the expansion people have kept saying that simcraft doesn't give correct results for mages, whether it be dps rankings relative to other (mage) specs or stat weights for a given spec. However, I have never seen anyone actually provide any evidence as to what exactly it doesn't do correctly. Does anyone actually have some facts/factoids/opinions/whatever about why simcraft shouldn't be trusted. Some examples that I could see as reasonable answers:

- The rotation for spec x was not modelled correctly, because no one could be bothered to do it. It's missing the following aspect ...
- You can't model the rotation for spec x correctly, because ...
- The rotation looks correct but ability x makes up for much larger/smaller share of my damage than in actual logs. See this log for example ...
- There are no setups in simcraft that could be used to get any useful data, because they all lack ...

Or something else along those lines. Anyone?

[Neoxx]

- The rotation for spec x was not modelled correctly, because no one could be bothered to do it. It's missing the following aspect ...
- You can't model the rotation for spec x correctly, because ...

That's pretty much it.
+ RPPM fucks everything up
+ most fight mechanics can't be simulated, since there is no real patchwork fight anymore

[Frost1129]

- The rotation for spec x was not modelled correctly, because no one could be bothered to do it. It's missing the following aspect ...
- You can't model the rotation for spec x correctly, because ...
- The rotation looks correct but ability x makes up for much larger/smaller share of my damage than in actual logs. See this log for example ...

It's your first point. People made an effort to model it in the beginning of the expansion, but it hasn't been refined and never really worked well. Key problems are Alter Time and RPPM trinkets. Alter Time makes writing an action priority list an absolute nightmare. Random trinket procs are not too difficult on their own, but having to having AT conditions and non-AT conditions for every spell is really difficult to do. I've been working on them the past few weeks, and have a logic list written down on paper for each spec, but transitioning this into a working APL has been very difficult.

Anyone who says you can't model a fight / rotation doesn't understand simulations. You can model anything. Some things are just more difficult than others. Also, the idea that SimC just works off patchwerk fights is also wrong. There are multiple fight conditions you can mix and match to generate an output that more closely represents whatever fight you are interested in.

[Neoxx]

Anyone who says you can't model a fight / rotation doesn't understand simulations. You can model anything.

No you can't, that's the problem with simcraft.
If it were possible then somebody would have done it already. It's not that everyone in the world is lazy, there are many that have tried it (including myself) but there is a reason nobody has a working APL that comes close to reality.

The simulation of a Mage "rotation" is not as simple as e.g. a Hunter rotation. There is a reason why Hunter sims are working pretty good (hint: instant cast, can cast everything while move, pretty basic rotation where you keep everything on CD with some exceptions). Meanwhile the Mage "rotation" is much more reactive and you can gamble with spells for more DPS.
Try to write an APL with the following examples and you know what I mean:
- "5 pyros in 4 seconds" opener
- Pyro gambling with 4p set bonus
- AM gambling on single/multitarget

Static rotations are gone, everything is dynamic especially since RPPM.

Also, the idea that SimC just works off patchwerk fights is also wrong. There are multiple fight conditions you can mix and match to generate an output that more closely represents whatever fight you are interested in.

Maybe that's true.
But personally I never liked these movement sims, since conditions like "move every 50 seconds for 5 seconds" are completly countered by e.g. Ice Floes. But that's just my opinion. Fights like Shamans or Nazgrim can either be a movement nightmare or you don't have to move at all, depending on your tactic and random factors ... how do you gonna sim that? (maybe that's a bad example, but you get the idea )

[Frost1129]

No you can't, that's the problem with simcraft.
If it were possible then somebody would have done it already. It's not that everyone in the world is lazy, there are many that have tried it (including myself) but there is a reason nobody has a working APL that comes close to reality.

The simulation of a Mage "rotation" is not as simple as e.g. a Hunter rotation. There is a reason why Hunter sims are working pretty good (hint: instant cast, can cast everything while move, pretty basic rotation where you keep everything on CD with some exceptions). Meanwhile the Mage "rotation" is much more reactive and you can gamble with spells for more DPS.
Try to write an APL with the following examples and you know what I mean:
- "5 pyros in 4 seconds" opener
- Pyro gambling with 4p set bonus
- AM gambling on single/multitarget

Static rotations are gone, everything is dynamic especially since RPPM.


Maybe that's true.
But personally I never liked these movement sims, since conditions like "move every 50 seconds for 5 seconds" are completly countered by e.g. Ice Floes. But that's just my opinion. Fights like Shamans or Nazgrim can either be a movement nightmare or you don't have to move at all, depending on your tactic and random factors ... how do you gonna sim that? (maybe that's a bad example, but you get the idea )

Yes, you can absolutely simulate anything. It just comes down to the core sim and how it's programmed. The effort of programming a fight by fight simulation is not worth the small payoff you'd get over the current options that are available now is my guess for why nobody does that. That and people with knowledge of how to program simc at that level are probably few and far between

Pyro gambling with the 4pc wouldn't be hard to do. I'm not sure how to do the 5 in 4, since you need to tell SimC to move your character and I don't know how to do that. But I already have conditions for a proper opener set up, so it shouldn't be too hard to say "If AT conditions are met, move forward and start AT opener"

Ice Floes is already modeled into SimC. If a fight forces you to move, it activates and allows you to continue casting.

AM gambling likewise I don't think would be difficult.

The main difficulty is alter time, RPPM trickets, and cooldowns. WoD will make the APLs a million times easier, since AT will be gone.

I believe you can write an APL for mage which takes these things into account - it's just going to be complicated.

Also - SimC is a tool. If you want to know how you'll do on shamans if you get lucky and don't move much, use Light Movement sims. If you want to see heavy movement, use helter skelter.

These two will give you an idea of how you will perform in these scenarios. From there, apply what you learned from SimC to your raid environment. Do you use a wonky movement heavy strat? Then trust the helter skelter output more than the light movement. Do you always keep the two bosses stacked? Include an additional target. Maybe the fight is hectic and you fuck up sometimes. Lower the skill it uses, so it misses actions sometimes.

You mix and match conditions, and get outputs that help guide the way you play / gear / spec. The outputs should be combined with what you know about yourself and your raid to come to the right conclusion for you.

I think I'm comming off as way more of an asshole than I mean too. Sorry >.<

[maizensh]

Hey,

Pretty much throughout the expansion people have kept saying that simcraft doesn't give correct results for mages, whether it be dps rankings relative to other (mage) specs or stat weights for a given spec. However, I have never seen anyone actually provide any evidence as to what exactly it doesn't do correctly. Does anyone actually have some facts/factoids/opinions/whatever about why simcraft shouldn't be trusted. Some examples that I could see as reasonable answers:

- The rotation for spec x was not modelled correctly, because no one could be bothered to do it. It's missing the following aspect ...
- You can't model the rotation for spec x correctly, because ...
- The rotation looks correct but ability x makes up for much larger/smaller share of my damage than in actual logs. See this log for example ...
- There are no setups in simcraft that could be used to get any useful data, because they all lack ...

Or something else along those lines. Anyone?

Frost and arcane can be modeled quite correctly with modified action lists (the defaults aren't that accurate). For instance, you can add proc snapshoting and better AT usage, quite easily I might add.
Fire on the other hand is convoluted. I think a good action list for combusts is possible (since you can put conditions in there), but it'll require quite an effort, because you'll need completely separate action lists for AT and outside of AT (which is what Frost1129 wrote, a nightmare).