What defines a good Moonkin 10 comments
posted 8th of Mar 2010
by Aeiedil
There is a problem that I have been thinking heavily about the last few weeks. It's a simple question, one that all players have their own take on. Put simply, what properties seperate the good Moonkin from the bad, or for that matter the good from the great? On top of that, why is this currently preying on my mind?
Why does this matter?
The first point to consider. Why does this question actually matter. Simply put, if you are reading blogs and the theorycraft threads then the chances are you care in some way about your performance. If you don't know what qualities seperate a good Moonkin from a bad then how are you to determine good advice from bad?
Too many players like to focus on tools such as Gearscore to determine a players ability. From the point of view of a PuG leader it is hard to use much more of an in depth tool without taking hours to form a group. When it comes to forming a guild you can afford to take the larger picture into consideration. Consider not just what the player has now, but what they could become.
You may find a player with a Gearscore indicating they should be at home in ICC25 Heroic. There is nothing stopping that player having 20% +hit chance and pulling out a massive 3500 DPS on Onyxia 25. There is nothing stopping a player with a Gearscore of 800 lower doing nearly double the DPS in the same fight. This example is plucked out of a raid I was in the other day with 2 other Moonkin. The point that worried me most is that the Moonkin with higher Gearscore (and lower DPS) couldn't comprehend what was so bad about being vastly over the hit cap with a low critical strike rating.
You may look at an armory profile and draw conclusions as to how well balanced a player is in a spec you know well. I know I do. Chances are if you know your class well you will be able to pick up on a few points that may help the player in question. To do this both players involved need to understand the reason behind the advice to be able to make an informed decision to follow it, or alternatively be able to argue their own point of view.
What can we get from an Armory profile?
The armory XML provides us all kinds of data. It shows us the achievements on a player. It shows us what items they have equipped. Their talents and glyphs. Their professions. Their gems. Their enchants. From this we can draw a conclusion as to how well a player should perform, and can draw a conclusion as to how a player may be able to improve their chances.
Equipment
In general equipment can be ignored. The important factor is the stat totals. The exceptions to this are procs (including on-use effects) and boosts that are variable (such as gems).
Procs are difficult to factor in reliably, however an estimation of uptime of the passive (chance on hit, etc.) procs is possible and can be made with a reasonable level of accuracy. Uptime of the player-activated procs is more difficult. What can be calculated is a maximum effect of those trinkets by assuming a use on every cooldown.
Variable boosts are an area where adjustment can be applied. Would a player benefit more from a Reckless or Potent Ametrine? Would a Runed Cardinal Ruby be of more benefit? These are the questions that need answering. I have already drawn up a post for gems and enchants, however these summaries can't take into account every variable.
Talents and Glyphs
Talents and Glyphs are generally pretty well covered by most blogs out there. The optimal glyphs seem to change from patch to patch at present, with Starfall looking to be moving back into favour. They of course tend to alter your spells functionality and so can sometimes be hard to quantify. Others are much simpler such as the Glyph of Insect Swarm which is a nice straight forward damage boost.
Talents change much more than just your spell effects though. Combinations of talents can change your entire play style. Improved Insect Swarm pushes more towards maintaining appropriate DoTs for example. If you haven't specced Force of Nature (idiot!) then you won't be including it in a rotation. If you haven't specced Eclipse (and so deserve to be deleted) then you won't be alternating between Wrath and Starfire particularly.
The upshot of these 2 points is that a players choices of talents and glyphs will affect their output and optional rotations. Combined with variances in stat balance, what works for one player may not work for another. All factors need to be taken together to calculate how well they fit. There is no point having Balance of Power if you have 24% +Hit as you are already well over the cap (for example).
What do your spells actually do
The effect of your equipment, talents, and glyphs is that your spells will not do what they originally said they did on paper. Due to the fact that they scale with your equipment they will most likely hit harder, hit faster, and crit more than they would if you were naked.
The effect of this is that a once-great trinket may become merely good (such as Reign of the Unliving, where the proc becomes less valuable when Starfire takes less than 2 seconds to cast). Your rotation may vary through a fight (such as favouring Starfire over Wrath during Heroism when you have high amounts of haste due to the GCD). Some spells may even fall out of your rotation entirely (discussion on Elitist Jerks about taking a DoT out of the rotation).
There are formula available thanks to the theorycrafters out there that allow a player to calculate what their spells will actually do based on their stats and other environment variables. This in turn allows a player to, given time, figure out their (theoretical) optimal rotation and gear setup.
Pulling it together, the Strayegg rate-my-Moonkin tool
This brings me onto my current pet project. First of all, it needs a better name. "rate-my-Moonkin" is a nasty name and I don't want it to stick. This isn't something you should expect to see appear overnight, but it is something I intend to build and have started building already. It will be a web tool that will (hopefully) do the following.
- Create a Moonkin ranking system and figure out who are (theoretically) the best Moonkins in the world
- Expose the formulae and methodology used to calculate the rankings in point 1
- Build on points 1 and 2 to enable a player to have an automated list of recommendations for how best to improve their Moonkin
- Build on point 3 to allow a tiered improvement path rather than jumping from quested blues to ICC25 HC epics without a step in between.
- Build on point 4 to include an estimated time frame so a player may see how long it may take for them to achieve their goal.
- Devise a method to tie WoL logs in to the ranking system built in points 1 and 2 to help verify or adjust the accuracy of the rankings
- In the event that points 1 through 6 are achieved, open up the system to allow it to be spanned out into other classes and specs
The first point may sound a bit grand, but in all honesty it is because I pride myself on being a well geared, both from level of gear and balance of gear, player. I am tired of seeing players above me on wow-heroes that have higher item level yet inferior (+hit trinket when above the hit cap?) gearing choices, obviously just gaming the Gearscore culture. I don't expect to be in the top 100, but it would be nice to have some means to figure out where I stand and figure out how my choices in the future affect that standing.
The eventual aim of this is to do away with the threads on forums where players simply ask for pointers as to how to improve their DPS. Such threads usually spawn a handful of the same few replies, and it would be good to be able to just provide users a link that can provide them advice for their Moonkin for the present time and for the future as and when they may upgrade.
Help needed
Now with all that out of the way. I am not a theorycrafter but I do have a strong grasp of mathematics. I could use input from the theorycrafters among you out there to help ensure that my formulae are accurate. I could also use any feedback from those of you who have opinions as to how to determine how good (or not) a particular player is based on the information within their armory profile (and eventually, any available combat logs).
If anyone would like to contribute Moonkin formulae and other such information (I will be checking the EJ threads when I get to that point) then I would be greatly appreciative. Any assistance in this endeavour would of course be accredited within the final end product.
Wish me luck!
Comments
I'm very impressed and excited about your upcoming project, I can already imagine how powerful of a tool it will be for me. While I'm not a theorycrafter, I am artistic! You could call it the Moon Machine. :P I had to say it, sorry.
But anyways, good luck on the project. :)
Great idea!
One thing that your "machine" could check from logs is what for example World of Logs refer to as "active time". Have it ignore too low % so not to flag dying progress raiders as bad players.
All the best for your effort Khad
I've not quite worked out how I'd like to bring World of Logs into it, just that it would be nice to have it somehow. I would probably have it behind the scenes to see how accurate the rankings are overall. It would be too simple to just rank by the peak DPS achieved, as everyone gets lucky streaks!
Indeed. To have it yank just raw dps out of logs would probably be missleading. Fast kills vs long kills and so on... Same with overall dmg, % of damage done etc. You probably have some sort of math solution figured out. Gear is x ammount of your "being good", spec is some and so forth. What i mean is active time could be one of the factors to judge. Less geared moonkin in less progressive guild can still have < 97% active time and could merit some "points" from it. Annyways, im pushing on to your turf.
Again. All the best, keep on posting thou ;) Khad
As I was reading this, I couldn't help but feel that this will all fall back to two things. Gear and raid composition.
Integrating WoL will cause a few glitches because faster kills tend to have higher proc uptimes and higher hero/BL uptime. This leads to skewed DPS. If you could find a way to normalize the DPS so that the things like that wouldn't matter, then it might become a little more viable.
It sounds to me like a well geared boomkin would lead your chart regardless of performance factors. This is obviously not what you are intending. Working around this can be tricky, but as you said, if you look at stat totals instead of ilevel, you might get a better picture. Modeling procs (Nibelung for instance) will create a bigger hurdle. You could always set up an SAT-esque ranking system for hard caps (hit) where anything over takes off a fraction of a point (so those just over wouldn't suffer too much but those with an extra 100+ hit would. Make it logarithmic in growth?) Add a weighted system for stats and their soft caps, giving a certain value to a stat prior to the soft cap and a lower value after the soft cap.
Also, I don't think that you can use the rss feed for accomplishments as any type of system either as it would penalize the players with lower playtimes.
Now modeling a players effectiveness might be a way to do rank them. Using their gear, all ideal raid buffs, and 0 lag, run a simulator to get their average DPS on a stand and deliver fight. Multiply this by their active time and then compare to their actual DPS. This ought to create a percentage of their max that could be compared to other druids without consideration to gear.
Just some thoughts. . .
Hi Relevart,
I appreciate your feedback on this. While I can make my own decisions as to what to show, I would rather have others bouncing thoughts and concerns also as an insanity check :)
When I say about integrating WoL I would just like to clarify that the WoL results will have no effect on the actual scores/rankings. They would be used behind the scenes to adjust the settings used to generate the rankings in the event something is totally out of line. The WoL integrations of any kind are a long way off though.
At present I have a gearscore of sorts that is made up of the important stats to Moonkin. That is combined with a simulationcraft DPS rating (the config file for which may need adjusting to get a more reaslistic number - currently it simulates fights with some distractions and some movement and gets some nice results).
I have used both these numbers to calculate a ratio (DPS/Gearscore) which theoretically indicates how well balanced that particular druid is by way of stats. A heavily over hit-capped moonkin, for example, would have a higher gearscore (due to Hit being heavily weighted) but a lower DPS than expected thus a lower ratio. This seems to touch on what you suggested about effectively punishing over-capping.
I don't intend to use Achievements in this system at all. They have their uses, just not in this system.
If you look at the top 50 (http://www.strayegg.com/top50.cfm) which is still a work in progress then you can see that currently the top players do have a higher Gearscore. Graylo is a good example though. He has a lower gearscore than me but comes out higher on the chart and this is reflected in his particularly good DPS-Gearscore ratio :)
I've got a lot more data pulling in overnight so will have a better idea as to how it looks tomorrow morning I expect :)
Cool idea! On a side note, I've been tinkering with the ruby library for Wow armory (the library is called 'wowr') and I'm trying to put together a script that will tell you what raid achievements someone has accomplished, filtering out raid achievements for tiers of content that are now easily outgearable. The motivation is exactly the opposite for me; while I can look at any moonkin's armory and determine whether or not they build their spec the elitist-jerks-approved-way, I have less of an idea for other classes.
I want to eventually include class/spec specific modules that comprise half the 'score,' so that one half of someones' ranking comes from past achievements, the other half comes from spec-specific formulas that determine how "correctly" someone gears up.
The problem with that methodology, and the reason I wish to ignore achievements, is that achievements can be done with 1 person not carrying their own weight in a lot of cases, especially with outgearing.
Aeiedil, you can correlate dateCompleted that I believe the armory xml supplies with when the content was 'fresh.' Still not perfect, but the goal is to filter out people who can't play. So if you combine how they've geared with what they've done you can probably filter out a good percentage of the people who aren't any good.
No methodology is perfect.
I'm still not happy with that.
No methodology is perfect, but so far I have some vague ideas of a route to go involving DPS simulations and gear levels (through various measures) in order to determine some kind of idea as to who knows how to optimise what they wear. Getting to that step will take a bit of time, and until that works efficiently then future steps are seeming futile.