Possible Recount/Skada Module: DPSI

[Vladimeir]

A friend and I were discussing meters the other day and differentiating how well someone performs for their gear and the usefullness of something along the lines of Damage Per Second Per Average iLevel.

Benefits:

Brings numbers to a reasonable level of comparison. Instead of comparing 30k to 31k, you'd be comparing super low numbers and millions of damage to thousands of damage.

It would be a more accurate representation of performance since the number has a lot less to do with where you place with everyone around you but where you place with the gear you have.

If leading a raid and you need to talk to someone about their performance, you can look at this number and see who just needs gear to perform or who is just doing something wrong without having to individually inspect the low numbers. 28k is acceptable for 370 but not 390. (370 ilvl has a score of 75.7, 390 has a score of 71.8 and would have to do another 1500 DPS to be comparable. See below for a problem with this example as I feel the 390 should have to add a lot more than 1500 DPS in order to be comparable to someone 20 ilvls lower performing their limit)


Problems:

Innately, certain classes and specs will perform worse than others regardless of how hard blizzard tries to balance.

There has to be some scale. For example, using the one above, 28k for a 370 ilvl is almost max performance. 29.5k, to achieve the same score, for an ilvl 390 person is about 80% performance. While the number(DPSI) should go up and not cap out (Or perhaps it should cap or be "around" a certain number that is ideal. Say based on a middle of the ground class/spec), it needs to go up slower such that performance between 370 at 95% should be equal to 390 at 90%, not 80%.

I could probably work out the math but I am at work and would need some more concrete numbers to work with.

[Hekto]

as a raid leader id be very interested in having something like this at my disposal. Of course we would have to take in other factors that it cant account for, as you mentioned class balance, and the fight in which you are comparing players to, but could prove invaluable.

how much id love to say to our regulars who are much more geared then some others,,, "John Doe is performing at a much higher percentage then you, so pick up your slack"

[Vladimeir]

as a raid leader id be very interested in having something like this at my disposal. Of course we would have to take in other factors that it cant account for, as you mentioned class balance, and the fight in which you are comparing players to, but could prove invaluable.

how much id love to say to our regulars who are much more geared then some others,,, "John Doe is performing at a much higher percentage then you, so pick up your slack"

With all meters, common sense needs to be used. The only problem is, when there are meters, there is some idiot spamming them in your 5 man heroic haha...

It, in my opinion, just provides on the fly useful information that you don't have to research to figure out. I could look at the meters, say hey, something isnt right here, then spend the next 10 minutes inspecting people, crunching numbers in my head, going over combat logs and activity reports, and asking other players of that class whats up before finally coming up with "hey... push buttons harder." Its a lot easier to gauge, if tuned properly, from a number than running around like crazy while trying to lead a raid.

[WskyDK]

The biggest issue I see with this is that some fights that punish various specs (hagara for melee) more than others. Yes it would be useful for side by side comparison of same-spec players, it breaks down after that.

As stated above your best bet is going to be common sense.

[ComputerNerd]

Context, both in ranged/melee roles, niche mechanics such as single/vs multi-target where some excel over others or simply some classes have substantial advantage or disadvantage over others on a given fight.
When will stupid people stop trying to bring everying down to numbers rather than just using their brains.
No meters will ever be a useful measure on numbers alone, never.

Those alone are reasons why meters are flawed, let alone before you add yet another flawed measure to it.

[Deleted]

Good raid leaders already know when someone is underperforming, any somewhat serious guild makes sure to either have a raid leader who has extensive knowledge of all classes / run simulationcrafts for their gear, compare to other plays of their class with equal gear with world of logs etc or if you have a "class GM" in charge for everyone else of that class who can help the raid leader out.

This meter will do nothing good and ignorant people will misinterpret it, just as gearscore was corrupted in the sense that it only calculated the ilvl, not the stat weight on the item for each spec etc and hence a person with ilvl 375 would probably do more DPS than someone in ilvl 380.

All the tools for knowing if someone is underperforming is already out there in terms of world of logs, elitistjerks etc.

[namtlade]

it's a nice idea, to have a measure for dps that takes a player's gear into account. But even ignoring encounter design, so just using it on ultrax/patchwerk fights, ilvl doesn't give enough information. Upgrading a weapon from 359->378 will give more potential dps than upgrading bracers by the same amount but will be represented in ilvl exactly the same.
It's also possible to have someone in full 397 but with no set bonuses. They'd have the same ilvl as some with the full 4set but will be missing the huge bonuses from the t13 set.

It's a nice idea, but imo there are far too many variables that this system wouldn't be able to take into account.

[Vladimeir]

I realize all that was mentioned. I just wanted some extra thoughts on it.

From a personal standpoint, it has its niches. But even so, it wouldn't replace the knowledge I already have for classes, where they should be, etc... as was mentioned in previous posts.

But, the potential for misuse > any help it could provide.

I agree with too many variables for accurate representation.

I guess it came down to being tired of people linking meters in LFR and I just /facepalm at them and I personally get shafted in pug raids (ever since gearscore and average ilvl existed) from people being ignorant and misusing the tools provided.

[Emane]

It's also possible to have someone in full 397 but with no set bonuses. They'd have the same ilvl as some with the full 4set but will be missing the huge bonuses from the t13 set.

Still if both performs on the same level, the one without the set bonuses are clearly better. Or add 397 without setbonus as default, say 100%, for someone at 397 without setbonus doing 100%, it's fair, for someone with set bonus doing 100% it's fail.

I would rather much have something that links activity to dps. As you then can clearly see that someone is wrong. If someone is top dps, but lowest on activity and someone is high on activity but lower than that person. The person with the higher activity can be the better player.

[namtlade]

Tired of people linking meters after fights in LFR? Or tired because they're linking them and then bitching about it?
Sometimes I link meters in raid chat, but I only link the dps meters on request (and limit it to top 5 to reduce spam).

I do very much like the idea of a weighted dps meter that removes the bias of gear, but as you've said as well, it's just not possible to make the numbers mean anything more than the raw dps meter would.

@Emane: If two players were at the same ilvl, one with and one without set bonuses and they did the same dps then of course the player without bonuses is doing better. But the system would need to reflect that, and not all set bonuses have the same affect to dps.
Measuring activity to gauge a player's skill is a nice idea too, but that's only assuming that they're hitting the right buttons. I could have high activity on my rogue by just spamming sinister strike and nothing else, or play my druid as balance and do it wrong but with a high activity rate.

[Flaim]

didnt read the whole thread so i dunno if this was already mentioned, but an other crucial fact about this idea is how well a class scales with gear.
e.g. a hunter has pretty much a constant dmg throughout the whole expansion, because of their kinda bad scaling, while a warrior does pretty bad at low gear, but with high end gear they're over the top slightly.
so the only useful way for this feature would be to compare people of a single class.

[ComputerNerd]

This is not a new suggestion either, and has been covered before with similar results.
And therefore as a conclusion, there is a reason why it has not been implemented before, because the author obviously does not believe in the value of it.

[a21fa7c67f26f6d49a20c2c51]

You're looking for an addon to do the following:

Measure "skill difference" between players while accounting for variation in class, spec, gear, encounter mechanics, buffs/debuffs. The way we'll measure skill is to compare the difference between "predicted DPS for your gear" and your actual DPS in an encounter.

Actual DPS is trival to measure. Predicting DPS is much harder for a few reasons:

• Certain buffs/debuffs can drastically inflate a players DPS. Consider DI on a moonkin, priest, or Fire mage. Someone being fed PI or tricks of the trade should do noticably better than somebody whos' "unbuffed"
• Certain items are much more powerful than others despite similar item levels. A caster with dragonwrath should out-perform somebody with a 410 weapon. 4 piece T13 bonsues are very good - 397 offset peices are much less powerful. LFR trinkets can be better than those from heroic ragnaros or the valor points vendor.
• Gimmicky class mechanics (like fire mages on yorsahj, rogues on gunship, etc) will favour some classes over another.

When comparing top parses we can say things like "everyone has all the appropriate raid buffs" because those groups are typically stacked to ensure that top parses have everything they need to compete. In an LFR or pug run I'm going to get FM on my affliction warlock because I queued with my mage friend: your fire mage can go to hell.

It's not hard to think of other reasons it's hard to predict DPS but the above should give us pause before thinking we can do it accurately. We'll likely want to say something like "you should do <X> DPS with ilevel <Y>, +/- <Z> percent"

The problem we have is that <Z> is quiet large when you can't control for buffs/debuffs/good luck/encounter gimics/etc. Consider just one buff: A fire mage with DI should do roughly 4500 DPS more than one without. That's the difference between a Top-10 parse and not making the top-200 on 25m normal mode. Add in things like PI, ToT, FM, and plain old good luck and I'd argue the error bar should be +/- 15%.

If we have to say "In ilevel 400 you should do 45,000 DPS +/- 15%" then we end up in a situation where somebody doing 39k and somebody doing 52k are both 'equally good': they're within the margin of error.

The more we try to tighten up the margin of error the more we punish/reward people for things outside their control. Even if we cut that back to 5% (and if this fight isn't good for your class, or you don't have some important buff: too bad) we still have a pretty huge range: a given player can vary by 5k DPS and still be within the margin of error.

Then you consider what contribution ilevel plays. Let's say that a fresh 85 in ilevel 330 quest rewards should do 10k and in ilevel 400 you should do 45k. That means each ilevel is worth roughly 500 DPS. With our "realistic" 15% margin of error that means that means you could reasonably expect somebody in full LFR gear to out perform somebody in full 410 gear: they'd both be within the margins of error.

Even with our tightend up 5% margin of error on predicted DPS: somebody in ilevel 395 could out perform somebody in 405 and that wouldnt' be unexpected. Think about that: 10 item levels-almost a full tier of gear-don't shift the 'expected' DPS levels in such a way that you couldn't reasonably explain the difference away by pointing at a missing buff.

Then we have to remember that there's only 30 item levels between just barely allowed into LFR and having full best in slot. We need a pretty huge difference item level before we can be sure it's something more than just luck causing the difference we see.

Because the signal we're looking for (difference between "expected DPS and actual DPS") is so small compared to the noise (buffs, rng, encounter gimmics, extra strong items for an ilevel) it's really hard to measure.


Sure, you can make it work in extreme cases: like somebody in ilevel 378 doing 45k DPS on ultraxion, or somebody in 400 doing 20k DPS on the same fight. In those cases it doesn't matter what gear they have: 45k is respectable on ultraxion no matter what the ilevel. 20k is laughable no matter what the gear.

When you start getting into middle ground performance (like ilevel 400 doing 38k DPS or someone in 378 doing 30k) any mark of skill gets drown out by the impact of buff/debuff/procs/luck/etc and again ilevel doesn't really matter.