(edited by Mallis.4295)
Math in a void, DPS and the Meta
1 2
Don’t divert away from the game. It’s a simple discussion here. Do the calculations scale to allow people to extrapolate the DPS calc results into achievable DPS ranges to compare between classes. No it doesn’t, because the calculations don’t take into account ANY game mechanics.
(edited by Obtena.7952)
My personal engy calculator can handle well over a million possible combinations of scenarios, including time on target, external buffs, external vulnerability etc. I can’t speak for anyone else.
That said, I’m not going to whip up a quick simulation for every party I ever get into. I usually calculate DPS for solo play and DPS for fully optimized group play and realize that my actual damage falls somewhere in the area in between depending on the factors of that particular scenario. Generally speaking, the times where I really care about my DPS though, it is fully optimized group play, so having a Patchwerk style DPS calculation to know where the maximum achievable values fall is what I find most useful.
You could certainly make a reasonable argument for a need for the popular build advertisers to provide unbuffed, poorly buffed, and totally buffed numbers because some builds scale way way better in optimized groups than others, which may provide a better benchmark for people depending on the type of groups they usually play with, but I suspect that the community at large would still latch on to the biggest of the 3 numbers and use that for the rule of thumb reference.
TLDR: Any decent DPS calculator will be able to handle multiple scenarios, but since there are zillions of possible scenarios and you can’t make a video showing off all of them, it is usually most practical to present results for patchwerk style benchmark numbers.
TLDR: Any decent DPS calculator will be able to handle multiple scenarios, but since there are zillions of possible scenarios and you can’t make a video showing off all of them, it is usually most practical to present results for patchwerk style benchmark numbers.
Maybe that’s true but that’s a diversion argument; no one is asking people to show ALL the possible scenarios. If would be nice if people would present even just ONE more RELEVANT scenario though.
For example, there is a dodging impact on melee more significant than on ranged play. Of course, we haven’t seen anyone analyze this because … not meta. Theorycrafters only going to show people the builds they want people to use.
(edited by Obtena.7952)
oh my kittening god! who cares! join a group, do damage, and dont die! you guys are over thinking this by such a huge margin its incredible. its a game. relax.
Team Deathmatch for PvP – Raise the AoE cap for WvW – More unique events for PvE
oh my kittening god! who cares! join a group, do damage, and dont die! you guys are over thinking this by such a huge margin its incredible. its a game. relax.
Yes, join a group and then get kicked because people have decided that your class sucks.
For example, there is a dodging impact on melee more significant than on ranged play. Of course, we haven’t seen anyone analyze this because … not meta. Theorycrafters only going to show people the builds they want people to use.
The reason people don’t answer questions like that is because the answer to these sorts of things is almost always “it depends”. Does dodging impact melee more than ranged? It depends. Engineers are most effective in close range (bomb radius) of a target, but when they aren’t using blowtorch, fire bomb or concussion bomb, they don’t lose any damage by fighting at range. If you have to dodge out of melee range while your short range skills are available, you’re going to sacrifice some DPS while getting back into melee range, but if you can dodge through the boss, you won’t lose any. If you can predict exactly when dodges will occur and how long it will take to get back to the target afterwards, I can certainly model the exact impact of that on damage. But if you just want to assume dodging once every 10s, the DPS impact of that is going to depend on where you are in a rotation when you do the first dodge. Modeling the results of that are still trivial if you pick a specific time to begin dodging, but the resulting number still wouldn’t be accurate for a fight where you first dodged 2 seconds earlier, so you’d have to run another scenario.
The end result is, your best method to come up with an approximation of what dodging every 10 seconds would do is to take the maximum total DPS attainable and reduce the time on target by some% to accommodate.
I theorycraft as much as anybody, and I don’t run meta zerk builds at all because my primary play area has historically been WvW where things like toughness and vitality matter. However, when I do go into PvE content I want to be optimal with what I have available, so I crunch numbers for whatever gear I have access to or feel like I’m willing to spend money to get. Mostly I crunch numbers for GW2 because I enjoy looking at systems. The ideas behind the numbers in the game and how they impact game balance are interesting to me. I don’t do math to push an agenda or force other people to play something that they don’t want.
Personally I don’t think you should take downtime and specific encounter mechanics into account when calculating baseline DPS. That’s going to be unique to every encounter and you can deal with how that affects your composition when you come to that encounter. Plus some of these mechanics can be completely avoided by more skilled players or with different strategies. Encounter specific DPS is much better calculated with logs. I think the main difference that should be looked at for DPS calculations is toughness since that’s going to create unavoidable differences in the favor of condition DPS the higher it goes. It would be nice if you aren’t going to show your full math to at least show how toughness scaling affects it. Or at least physical vs condition damage.
(edited by Mallis.4295)
Encounter specific DPS is much better calculated with logs.
Logs give you very specific data about how a particular player piloting a particular build contributed to an encounter, but do not give you any contextual information about how good or bad those numbers are – you can look at a DPS log and see 8k DPS, but is that good? Is that bad? Without a scale a particular number doesn’t mean anything.
In theory you could get DPS numbers for a variety of players using the same (and comparison) builds in the same encounter, and build a scale statistically from that data – how well do people perform on average with a build, and how does that stack up to other builds? That would be empirically valid. Without really widespread tools feeding into a database, however (can the API do that?), such data collection isn’t practical, and purely theoretical scale creation (such as the DnT spreadsheet numbers) are the only practical scale in town.
Best case scales have known flaws (namely that they have trouble accounting for the difficulty of piloting a given creation – or how particular individuals will perform piloting a build) but they at least create justifiable anchor points from which to compare to. Without a scale you might as well be having a shouting contest.
In theory you could get DPS numbers for a variety of players using the same (and comparison) builds in the same encounter, and build a scale statistically from that data – how well do people perform on average with a build, and how does that stack up to other builds? That would be empirically valid. Without really widespread tools feeding into a database, however (can the API do that?), such data collection isn’t practical, and purely theoretical scale creation (such as the DnT spreadsheet numbers) are the only practical scale in town.
Yeaaaah that’s kinda what I meant. Something like warcraftlogs. Probably never going to see that in GW2 but doing spreadsheet DPS for specific encounters is never going to be anywhere near accurate outside of Patchwerk fights. And even then you have the problem that spreadsheet DPS isn’t very accurate in the first place.
The reason people don’t answer questions like that is because the answer to these sorts of things is almost always “it depends”. …. I don’t do math to push an agenda or force other people to play something that they don’t want.
Yes, I agree with all that. The calculations, which are not accurate because they are dependent on game mechanics are still used to advertise the meta for all PVE content. Maybe that’s not you, but it is other prominent people.
The main question is what value a DPS model has that:
1) gives only the ceiling damage value in non-relevant game situations
2) doesn’t scale downwards correctly to enable profession or build comparisons
3) is HIGHLY dependent on game mechanics not included in the model
Answer: Almost none. A model is used to predict and needs to be validated. The models that are promoting these meta builds don’t do any of that. It’s garbage science.
(edited by Obtena.7952)
We just need DPS meters to settle this stupid dispute once and for all. And we also need world of logs. It boggles my mind that we don’t. I mean it’s an MMO, what MMO that is any good has no way of tracking important stats like damage done/healing done/etc.
Both sides in this everlasting argument make good points. Neither can be proven right without a DPS meter and logs. At least IMO.
… Something like warcraftlogs. Probably never going to see that in GW2 but doing spreadsheet DPS for specific encounters is never going to be anywhere near accurate outside of Patchwerk fights. And even then you have the problem that spreadsheet DPS isn’t very accurate in the first place.
I’m not sure about any other DPS sheet, but the last time I checked my engineer DPS sheet against the DnT one, we ended up within 17 DPS (out of 18k or so) of each other when using the same assumptions. I’d be willing to bet that they’re pretty accurate for Patchwerk style fights.
As for validation of these numbers, I’d look at the world first boss kill comps as something of an indication. The Vale Guardian was killed by a Druid and a ton of Engineers and Revs. I’m sure that other comps can kill the boss, but that was the setup that did it first, and I’m sure it was more than a handful of groups trying to kill the boss. Competition for things like world first kills drives people to try to find the best things available, not to rigidly adhere to some pre-determined notion of what is meta.
Note: I think that if we run into a raid boss that is truly a DPS check that a LOT of people will struggle without a DPS meter available to help determine where they can improve and what is/isn’t working. I think having access to a meter in game would be great for the game in general, and almost mandatory for raids if they make them hard enough.
I’m not sure about any other DPS sheet, but the last time I checked my engineer DPS sheet against the DnT one, we ended up within 17 DPS (out of 18k or so) of each other when using the same assumptions. I’d be willing to bet that they’re pretty accurate for Patchwerk style fights.
First time I’ve seen someone say a number for engis that didn’t seem obviously wrong. Mainly seen people talking about them doing 30k DPS. I don’t really play engi and only did a bit of testing but 30k was obviously wrong. 18k sounds realistic with alacrity and quickness.
That was 18k with 25 might/vuln all banners and EA. Engy DPS goes completely bonkers when you start getting alacrity due to the fact that you have so many strong skills with cooldowns. If you have 100% alacrity uptime, you virtually never need to use grenade auto attack. You get to use shrapnel grenade and fire bomb almost like auto attacks, which are 2 of the hardest hitting skills in the game.
That was 18k with 25 might/vuln all banners and EA.
Are you talking about burst or sustained because that does not sound realistic for sustained at all.
That is sustained. Using Viper Armor/Weapons + Sinister accessories I’m getting the following numbers over a 5+ minute fight.
Direct DPS - 4,145.37 - 20.6%
Condition DPS - 16,011.43 - 79.4%
Total DPS - 20,157
I’ll clean up my sheet a bit and post it tomorrow so you can look over where the numbers are coming from.
Are you talking about burst or sustained because that does not sound realistic for sustained at all.
The only thing in that list that cannot be maintained throughout the long fight is the presence of both banners.
There is no math in a void, there are only very generalized spreadsheets.
They are that generalized since Gw2 never used to be engaging enough to care about boss-specific encounters, so the meta was formed solely around the 2 goals to a) find the single highest dps for every class and b) find the most synergistic group-comps to allows as much dps overall as possible.
For this scenario 100% dmg uptime got assumed (no interruptions from heals & dodging), 2.6k target armor and 100% uptime on self-buffs and boss-debuffs.
And since we had the condi-stack limit for 2 full years after launch, the zerker meta formed, and was – usually – extremely effective. Once the condi-stack limit got removed people started to experiment with condi builds as well, and one such build is the engi sinister nade build.
The reason it reaches so high numbers is because – again – the specific boss mechanics get ignored when theory-crafting a build, so they always end up with over – the – top – numbers. But since each and every single spreadsheet uses the same input for their builds, it’s easy to draw comparing conclusions between classes.
Sadly this has lead to people miss-understanding the math being done and using these numbers to discriminate and ridicule players of classes who do not achieve these high numbers on paper, or come with a lack of group-synergy to bolster the party’s dps.
Still this math is not done in a void. It is done under equal conditions, so properly reflects the proportionate potential of each class correctly, even tho the raw numbers are not achievable.
For thise who do not understand the thought process behind a spreadsheet calculation & doubt that it is relevant to ingame situations, here a 3-step tutorial that shows how a build is theory-crafted, step by step, by drawing conclusions directly from the game’s mechanics.
So, I want you to consider the following points, and tell me if you agree with them, and if you not, tell me why you disagree with them:
-disclaimer, none of these questions care about critical damage at this point. This comes later, to the end.
1.) The game’s engine uses [on the wiki] documented formulas to calculate power & condition damage ingame for each skill / ticking conditions. True or not?
2.) If the toughness values of any given creature are inserted into the power-dmg-formula, it will give out the accurate dmg-range (since weapon-strength is partly RNG) of any given skill, BEFORE applying buffs. True or not?
3.) If that dmg-range is altered by -33%, it accurately reflects the influence of protection on a mob. True or not?
4.) If that dmg-range is altered by +25% it accurately reflects the influence of vulnerability on a mob. True or not?
5.) If the dmg-range of a skill is divided by its full animation duration (pre-cast & aftercast included), added with the full condition-damage (duration x stack.amount), you can determine its Burst dps. True or not?
6.) If its determined burst dps is higher than the dps of your AA, you will prefer to use the burst over the AA. True or not?
7.) If you add the cd of a skill to its pre & aftercast duration, you can determine its full cycle-duration. True or not?
8.) If you know the burst-dps, the full cycle duration & the full animation duration of each skill a class offers, you can engineer a chained segment of burst-skills, while filling the gaps with your hardest hitting, available AA. True or not?
If you do not disagree up until this point, we agree that rotations can be engineered entirely theoretically within a spreadsheet, accurately reflecting ingame numbers, if the correct stats in your build get inserted here.
Further below you find a theorethic example how spreadsheets can give you more absolute numbers for each boss encounter, and i recommend using the below described method to engineer builds for raids, once we know the boss-mechanics.
9.) It is possible to estimate a boss’ toughness value by hitting him 100 times (or more), calculating the average hit-dmg, and then resolving the – in the wiki documented – dmg formula to give out toughness, while inserting the previously calculated average hit-dmg, the power value of the build used, and the weapon-strength of the used item. Further present protection can be compensated by increasing the calculated average hit dmg by 33% before inserting it into the now-resolved toughness-formula. True or not?
10.) If the calculated toughness value gets applicate to the dmg-calculation of any given skill, it will accurately return the true hit-dmg-range of that skill on the same boss ingame (in case of irremovable perma protection, after reducing it by 33%). True or not?
If you agree with me up until here, we agree on raw dps being determinable for each boss specifically, just by a spreadsheet.
11.) All bosses have skills with specific cd’s. True or not?
12.) A dodge takes 0.75 seconds. True or not?
13.) If you identify the skill of a boss that apply greater dmg than you can outheal them, then their overall cd (actual cd + precast + aftercast), determines the time-interval in which you loose 0.75s dps. True or not?
14.) You can calculate the percentile dps loss if you know how often you have to doge within your own rotation. True or not?
15.) Invulnerability-phases of bosses follow a determinable pattern, be it by a given skill-rotation, a HP-trigger, or simply by the cd of the invulnerability-spending skill. True or not?
16.) Group-stuns, like given from the colossus fractal’s final boss can be treated as “invulnerability-phase”. True or not?
17.) The existence of said pattern allows for developing boss-specific formulas that give out the rough percentile vulnerability-time to dmg, which can be used as further, dmg-reducing percentile factor (F.E. determined dmg-uptime is ~50%, so power-dps gets reduced by ~50%). True or not?
18.) The kind of invulnerability is important to note, because some just become completely invulnerable/evade, and others still take hits with zero dmg, allowing for condi-application. True or not?
19.) Because of this behavior, condition dmg & power dmg uptime always must be calculated separately per skill. True or not?
20.) The combined dps loss from dodging & boss invulnerability phases can be utilized to determine your boss-specific dps-uptime. True or not?
21.) Calculating the final dps up until this point with crit-chance, crit-dmg, dmg-modifiers, and crit-related procs allows for accurate, boss-specific results. True or not?
If you agree with me up until this point, we both agree on builds being able to be compared in theoretical, boss-specific scenarios within boss-specific spreadsheets, coming up with dps conclusions that are very close, but still slightly above reality.
The rest-discrepancy to reality is caused by individual player-skill & therefore require mechanical perfection from a player wanting to achieve these numbers. Still the above described methods allow for coming up with builds that not only reflect very realistic numbers, but can be used to analyze a wide range of boss-encounters, without ever recording your dmg ingame (beyond probing the values for this build-process).
Additionally the presence of cc-spammy builds (for the breakbar-stuns) and healers supporters (especially with frequently available group aegis) greatly improves your groups dmg-uptime since less dodges are necessary. Also alacrity will greatly influence your full skill-cycles, while quickness alters your burst-dps, making the presence of mesmers /chronomancers very interesting as well.
So in this kind of spreadsheet there is actually room for higher dps than it calculates, given that you form a optimized group, with reasonable support & cc-capabilities.
Yes, dps-meters would greatly improve our capability to engineer these builds, but currently there is no viable dps-meter supported by A-net (or even rendered a break of the TOS, endangering you of loosing your acc if used with your client).
Again, there is no “math-in-a-void”.
This is just a empty phrase, made up by a individual, solely for the reason to draw attention to his youtube channel & spawn controversy around him.
Engineer is love, Engineer is life.
(edited by Arantheal.7396)
If you agree with me up until this point, we both agree on builds being able to be compared in theoretical, boss-specific scenarios within boss-specific spreadsheets, coming up with dps conclusions that are very close, but still slightly above reality.
The rest-discrepancy to reality is caused by individual player-skill & therefore require mechanical perfection from a player wanting to achieve these numbers. Still the above described methods allow for coming up with builds that not only reflect very realistic numbers, but can be used to analyze a wide range of boss-encounters, without ever recording your dmg ingame (beyond probing the values for this build-process)
Just saying people do this in WoW trying to be as accurate as possible but sims pretty commonly have great deviation from logs and really aren’t that practical. Looking at simcraft right now it estimates 72k DPS for a subtlety patchwerk fight with 720 item level. On the current raid tier’s patchwerk fight the top DPS parse from an ilv 720 sub rogue is 124k. DPS meters and logs will always be much much much better than sims.
If you agree with me up until this point, we both agree on builds being able to be compared in theoretical, boss-specific scenarios within boss-specific spreadsheets, coming up with dps conclusions that are very close, but still slightly above reality.
The rest-discrepancy to reality is caused by individual player-skill & therefore require mechanical perfection from a player wanting to achieve these numbers. Still the above described methods allow for coming up with builds that not only reflect very realistic numbers, but can be used to analyze a wide range of boss-encounters, without ever recording your dmg ingame (beyond probing the values for this build-process)
Just saying people do this in WoW trying to be as accurate as possible but sims pretty commonly have great deviation from logs and really aren’t that practical. Looking at simcraft right now it estimates 72k DPS for a subtlety patchwerk fight with 720 item level. On the current raid tier’s patchwerk fight the top DPS parse from an ilv 720 sub rogue is 124k. DPS meters and logs will always be much much much better than sims.
At the end of that post:
Yes, dps-meters would greatly improve our capability to engineer these builds, but currently there is no viable dps-meter supported by A-net (or even rendered a break of the TOS, endangering you of loosing your acc if used with your client).
We’re on the same side of that argument, but until A-net updates their TOS, spreadsheets are all we can (are allowed to) work with.
Engineer is love, Engineer is life.
(edited by Arantheal.7396)
We’re on the same side of that argument, but until A-net updates their TOS, spreadsheets are all we can work with.
Well I mean you can use Jaxnx.
But yeah they should really just whitelist some parsers or implement one in game.
We’re on the same side of that argument, but until A-net updates their TOS, spreadsheets are all we can work with.
Well I mean you can use Jaxnx.
But yeah they should really just whitelist some parsers or implement one in game.
Since it works via Print Screen, (welp, as overlay) it is horribly inaccurate if it comes to condi-dmg or very frequent hitting skills (like flamethrower or grenades on multiple targets.
Yes, for single skills (or some small combat-segments) it works reasonably well, but for an entire boss-fight it just skips too much dmg.
To draw accurate conclusions for improving your builds, it’s simply the wrong tool…
I’d love a ingame solution for recording dps properly…
Engineer is love, Engineer is life.
For me it’s simple, if you make a build and claim X class can hit Y dps reliably, then I want video proof of it being done. Often this can not be provided or when it is, it’s someone blasting away on some golems in spvp or a video where the dps is far below a reasonable margin of error making you question why it was posted to begin with.
Occasionally the stars do align though and someone’s claims are matched by a video, but this is such a rare occurrence, I just can’t fathom why builds are taken at face value, no matter who they are from.
Please! This is supposed to be a happy occasion.
Let’s not bicker and argue over who killed who.
Lilcoffeebean~Yak’s Bend~Perfect Dark [PD]
Let’s not bicker and argue over who killed who.
Lilcoffeebean~Yak’s Bend~Perfect Dark [PD]
A Bugatti Veyron 16.4 Super Sport World Record Edition has a top speed of 431 km/h (268 mph). Nobody ever says, “sure but show me a video of it doing 260+ in rush hour traffic and then I’ll believe you’re on to something.”
A Tata Nano has a blistering 0-60 time of 29.4 seconds and a top speed of 65 mph.
In rush hour traffic, crawling along at 15 mph, both of them will be the same speed, but I don’t think anyone is going to argue that the Bugatti isn’t the faster car.
Spreadsheet DPS is a measure of DPS potential, much like what a car can do on a closed race track. If build A does 20k max and build B does 10k max, equally skilled players will most likely do more damage with build A. Some builds are more forgiving than others, but these theoretical numbers are a good anchor for comparing builds.
If they put a DPS meter in the game Nemesis would probably complain it didn’t work. Spreadsheet math isn’t an expectation it’s a target, there will always be limitations of player skill and instances where optimal rotations don’t pan out and that’s fine, just don’t take anything that Nemesis says as a consolidated fact because it really isn’t.
I’m sure you’ll eventually realise that Nemesis is just a controversy show-boat and ends up on the Teapot dudes discussions purely for click-bait and views.
very special guild tag [tX]
I posted my Engineer DPS sheet in the engineer forum.
If they put a DPS meter in the game Nemesis would probably complain it didn’t work. Spreadsheet math isn’t an expectation it’s a target, there will always be limitations of player skill and instances where optimal rotations don’t pan out and that’s fine, just don’t take anything that Nemesis says as a consolidated fact because it really isn’t.
I’m sure you’ll eventually realise that Nemesis is just a controversy show-boat and ends up on the Teapot dudes discussions purely for click-bait and views.
Actually, the dude has stated openly that he plays up the drama to keep peoples’ attention because he knows they’ll get bored with pure numbers/analysis (which is largely true).
That doesn’t mean his analysis is disingenuous and made only for views. I’ve watched something like 4 videos with him in it and already, it’s clear to me that he either cares a lot about a truthful portrayal of GW2 builds or he is so good at conning people that I will gladly be conned by him because he deserves it.
Or words to that effect.
That doesn’t mean his analysis is disingenuous and made only for views. I’ve watched something like 4 videos with him in it and already, it’s clear to me that he either cares a lot about a truthful portrayal of GW2 builds or he is so good at conning people that I will gladly be conned by him because he deserves it.
He measures his very own dps, sells it as THE dps that you will reach with that build, calls everything outside of his DPS measurement fake math, and calls out people as trolls if they present valid points to debunk his claims.
One example:
He thinks that his necro doing higher dps than a phalanx strength warrior earns his necro build a valid spot in a meta-comp in exchange for the warrior.
When presented with the argument that phalanx strength warriors are not there for their dps, but for enabling everybody to crit with 100%, group might stacking and raw stat-group-buffs, he either ignores you or calls you a troll.
You should still be able to find that thread if you go back 2 or 3 months in the warri subforum. It got closed by a mod eventually, so it should be easy to spot.
This guy wants clicks, and live the youtube-money dream. If you don’t believe me, watch his channel intro and see for yourself.
Engineer is love, Engineer is life.
1 2