So I seem to remember in I want to say Legion/BFA there was a super comprehensive simulation tool in the AMR arsenal where you could leverage your PC’s computing power and totally customize any WoW related sim that you wanted to run. I think you even had to download some piece of software to your computer in order to run it properly. Anyway, what happened to that feature? I loved that, and I just noticed that I cannot seem to find it anywhere. I have taken several long breaks from WoW over the years and may have missed something…
We discontinued the simulator for Dragonflight. The new talent trees made the number of relevant gear combos we have to test explode in magnitude. Simulation is too slow for us to create our scoring functions. We moved to a mathematical model that can get us good data much faster. Since we aren’t using the simulator as the back end for the optimizer, we don’t have the time to maintain it. We have to spend our time refining the mathematical models instead.
We miss the simulator as well - simulation is fun to tinker with. It isn’t the only way to optimize gear, though!
1 Like
As a healer, AskMrRobot was the only platform where I could compare the output of specific item/talent picks by simulating setups against each other.
For me, the optimizer never worked well. It always displayed a lot of very strange percentages that would have led to one bad pick or another. (And in fact it did lead to one that could have easily been prevented: The “Darkmoon Deck Box: Watcher.” At the start of the addon, Mr. Robot said that this was by far the best possible upgrade, which turned out to be completely wrong and set me back in HoF progression).
I’m deeply disappointed that you made this decision, because the simulation was what made Mr. Robot special and exactly why I was so happy to pay the subscription.
Some out of context percentages can be obtained on several other sites for free.
Edit: In fact i just ran the optimizer to see what funny things it suggests now and it still wants me to replace a 415 “Voidmenders Shadowgem” with a 401 "Darkmoon Deck Box: Watcher” to increase HPS by 7%. What a joke.
If you have specific cases that you want us to look at, we’re happy to do so.
A single trinket showing a 7% increase seems strange… if you post a snapshot I can look at your specific case. That item certainly does not rank that high for most cases that I have tested. Instructions on how to create a snapshot here:
Well the 7% were at beginning of S1 thus i am not able to recreate the setup, but even though it still says best craftable item with +3.68% as for now. ( Which still is not realistic at all )
( 0ab7d7315e4c444da2f72e58425f2d6e )
Since i started using askmrrobot in legion i had many cases like this where it was very obvious that the optimizer values cannot be trusted without verification through simulation.
You mentioned to test specific cases. How do you do that without simulator?
I assume you’re talking about Darkmoon Deck Box: Watcher in your provided snapshot? If you are in a fight where everyone is taking damage, you can get good value out of it. We low-ball the versatility part assuming as a non-tank you won’t use up the entire shield reliably each time you use it (right now for a non-tank we assume you’ll only get the versatility buff 1 out of every 10 times you use it). We also assume some amount of waste (i.e. overhealing) on the shield (about 30% right now).
This is one of those trinkets that is hard to give a single value… it is highly dependent on your fight and situation. If you are not able to make good use of the self shield, lower the trinket’s value on the customization tab. It has a 90 second cooldown… if for example you find that you are only getting good use out of it say every 2 minutes rather than 1.5, add a multiplier of 0.75.
Your other trinkets are not amazing for healing – it seems reasonable to me that a trinket with a good shield effect would rank higher than certain stat buffs. And if you can trigger the vers buff at least sometimes in a tough fight… it’s a pretty good trinket.
When I said test cases, I meant test the optimizer with different characters and different combinations of settings, talents, etc.
As a healer I found Ask Mr. Robot at the beginning of Legion and have paid for it ever since I found batch simming. I know you said it is to hard to maintain but is there any possibility of bringing it back for a higher sub fee? I would gladly pay 2x the sub fee for the ability to batch sim. BiB has never given me accurate results and no I barely use Ask. Mr. Robot, There is no sim out there that fills the void left. Knowing that I had fine tuned a build for a purpose with every variable accounted for… It made me a better healer and honestly… it sure was fun!
I really miss that feature.
I was hoping I’d be able to bring back healing simulation, at least. I’m not ready to say it will never happen… but it will be hard to make it happen. It takes a lot of work to maintain the simulator and not very many people were using it. Of those very few people, only some even smaller number would be willing to pay to use it (it was always free in the past).
With the simulator being too slow to power the back end of the optimizer… it is really not a financially viable project. It would be a pure passion project, which means I have to push it to lowest priority 
Have you considered publishing your simulator on github?
I think if you have no commercial plans for it it would be a good solution.
Unfortunately, I don’t have enough time to make something of my own, but I could participate in development if it was on github.
The simulator allowed you to do a very important thing. Namely, fine-tune the damage for healers.
You could always take your simulation and compare the numbers with the real logs.
Now the only way to check your builds is to play them.
No, we haven’t. The code for the back end is too entwined with the optimizer part of the website.
There is a chance I could get it into a state where we could post it up with the new items/talents stubbed in and then let people use the wiki (like we always did) to update it as they pleased. That is my current hope.
Sorry to necro an ancient thread, but it seems more useful to continue the existing discussion than to start a completely new redundant one.
I quit playing WoW about 1/3 of the way through Shadowlands, but I used to spend HOURS cross-simulating my talent and gear combinations. I enjoyed the feature so much that I’ve left my annual subscription going even though I haven’t played the game in years.
I was considering returning to the game in War Within, and one of the things I was looking forward to was running simulations on the new talent trees. I’m sad to see the feature is retired, but I definitely understand the reasoning. However, it was also the ONLY simulator model out there which actually worked for tanking and healing, instead of just damage.
I had a couple of questions - would it be possible to create a limited simulator which only simulates talent tree choices against default gear sets, since that is one thing that’s left out of the mathematical model used in the optimizer currently?
Also, I remember the old simulator had a setting that seemed to be for distributed computing - letting the system run other people’s simulations on your hardware. It never seemed to be active, but I always left it on. Could we volunteer computing power to “the cause” to ease the speed problem?
Would there be anything that we users could do to contribute maintenance work or to relieve some of the drudgery of maintaining the tool? Perhaps with a one-way obligation, i.e “the tool will function when volunteers appear to update x-y-z, and otherwise will remain unmaintained”?
The simulator! I miss it.
Putting it up in a state such that all the talents and items would load and be stubbed in the wiki is a decent amount of work in itself. I’d have to talk to yellowfive, who always handled that part of the project, to get a better idea of how much that would take.
Going through and then implementing all of the talents and special gear effects is a very time-consuming task. We always imagined our simulator would be a tool for the community that members could help maintain or even take ownership over certain specs, much like what simc is. In reality, that never happened. It was seen as a competitor instead of an evolution and improvement on simulation. We thought the idea of being able to maintain it without needing to write c++ code would be attractive. Non-programmer community members could become involved in figuring out the guts of how the specs work. It ended up with me implementing all classes and specs myself, and writing the rotations. A few people helped iterate on the rotations for a few specs, which was great, but the engagement was overall very low. We even tried hiring some people to implement specs for us, but that was also met with extremely low interest!
The “speed” problem only matters if the simulator is being used to generate data for gearing strategies. The amount of data required with these new talent trees would be absurd. I don’t see any world in which we go back to that. Simulation is not the right tool for gear optimization at the scale we do. It is good for people who want to tweak rotations and experiment with talent builds on a case by case basis.
Now, if we don’t need to personally use it to drive the optimizer, that could open up the possibility of just throwing it out there as a stub like I mentioned above and let people do what they will with it. I wonder if it would ever get fleshed out or not in that case.
2 Likes
Could it be possible to create a ranking for the talents using the current code/data you have ? Kind of like what have been done in SL
I can only speak for myself, but as a CS student looking for a project as well as a huge fan and prior user of the tool itself, I would definitely put significant time into it, were it available to work on.
May not be the most sustainable support long-term, but it’s what I can offer. 
Sad that people were not able to look past “tool loyalism” and see a useful project instead of competition for the status quo.
It was a very cool project… under the hood it was written as a very generic engine. The web user interface for creating spells, rotations, stat calculations, etc. actually was the code of the simulator, minus the generic engine. The engine would dynamically compile the expressions that users entered on the user interface (for which we added syntax highlighting and code-completion!) using C# expression trees, so minus the dynamic compile time they would run as fast as regular code.
We did it that way for two reasons: firstly, so that non-programmers could work on the simulator, and secondly, so that anyone with a little guidance on the syntax of the web pages could verify that the simulation was doing what it was supposed to be doing. It certainly wasn’t simple… but it was a hell of a lot easier to read than Someone Else’s C++ Code… the dreaded worst job of any programmer is having to slog through someone else’s C++ code with little to no assistance.
We then coupled that with a first-class log output with filtering and searching and other metrics to further assist with verifying that it was working properly. On top of that, we made it such that users could easily run the simulations on their own hardware, but still get all the benefits of an easy-to-use web interface. That kept costs and wait times way down for many of our power users.
People got hung up on the fact that the core engine wasn’t open source… and completely ignored the fact that we spent tons of effort making the parts that matter for understanding and trusting it as transparent as possible. We didn’t open source the engine part because frankly… it’s hard to understand! Most of it has little or nothing to do with WoW. It is all speed optimizations, dynamic compilation and parsing algorithms, etc. The combat log output is the tool that one would use to determine if this core engine was functioning correctly… not the source code.
Same as @Swol, I’m always a little sad that the simulator didn’t catch on… it really was a good user experience. We have to focus on 11.1 for most of February here, but if we get a chunk of downtime I would be open to exploring getting the engine up and available for people to mess around with it who are interested. Maybe one day I will even get my act together and clean up the engine and separate it from our other code bases and put it on github or something… but that would be a bigger project.
3 Likes