Skip to main content

World of Theorycraft

  • strict warning: Declaration of views_handler_filter_node_status::operator_form() should be compatible with views_handler_filter::operator_form(&$form, &$form_state) in /var/www/drupal-sites/all/modules/views/modules/node/views_handler_filter_node_status.inc on line 0.
  • strict warning: Declaration of views_plugin_style_default::options() should be compatible with views_object::options() in /var/www/drupal-sites/all/modules/views/plugins/views_plugin_style_default.inc on line 0.

Why DKP systems are evil

Nenicirene's picture

All I see in this 'DKP Points' threads are people whining about not earning the same as other people that put in more work.

How can you whine about that?

A points system rewards you for what you put in.

How can anything be fairer then that?
You put nothing in, you get nothing out.
You put in a lot, you get out a lot.

What fairer system could their possibly be, why do you think that almost EVERY raid guild bar the very rare ones in EQ used this system? because it is the fairest means to reward people, you reward people for the work they put in.

No, DKP systems are a scam to retain loot within the most hardcore circle of the group that implements them. They way they work is:

You put a little in, you get nothing out.
You put in a lot, you get out a lot.

Rolling for items within the group rewards people based on the work they put in, because every raid you go on corresponds to a chance to win what you want. Go on twice as many raids, get twice as many chances.

On the other hand, a DKP system serves to lock out the people who go on less raids. They never get enough points to compete with the people that go on more raids, and thus the loot stays within the inner circle. However, fairness doesn't matter, as it's the inner circle that sets the rules, and they will naturally come up with a system that grossly favors them and then argue that it's "fair". Thus, the DKP systems won't go away.

I disagree with this...Niveus Lepus uses DKP, and items get doled out pretty fairly IMO.

I jumped in late and can't do Onyxia, so I lose out on two sources of DKP, yet I still manage to pull a fairly regular amount of items. I don't get them FIRST, but I get them pretty often. And I like to spend my points on silly things. ("What? Awesome sword when I already have an Arcanite Reaper? 'kay!")

We have people who gets stuff on their first run because so much of a certain class set has dropped that EVERYONE IN THE RAID of that class is in negative points.

Sure, there's the eventual inflation as everyone picks up what they need and nobody needs X or Y, but all that leads to is that newcomers get that stuff with no hassle...it drops them into negative DKP like mad, but they work out of it and don't get first crack on the next new stuff, whereas the people who've been around forever do get first shot at it. Which I consider fair.

Maybe our DKP system is different than ones you've seen, I have no idea, but it works out nicely for us. I've seen more than one first-timer get a nice Epic piece out of Molten Core.

Nenicirene's picture

It sounds like NL system is actually deflationary—that is, the points coming in are less than those going out, leading to most people being in the negative, and thus favoring new folks. I didn't address such cases in my original rant, mostly because I doubted they actually existed in the wild, as they work against those in charge of maintaining the system, and the my cynical self finds it hard to believe people would support power structures that harm them.

I did fail to take into account the loot-wave phenomenon, in which there's a finite number of items suitable for any given character, and people fill up on these, then move on to higher tiers of loot, and are thus not competing with newcomers for the entry-level loot, allowing the newcomers to get those comparatively quickly. My analysis was taking a more open-ended view, which assumes that all players retain interest in all appropriate items, as opposed to having a finite capacity to use loot.

All that said, I'm still opposed to any systems that are not zero-sum, as they're elegantly self-balacing without administrative tinkering.

Well, considering we rate at very low points-per-hour and give out shares of the items that drop...a single item can wipe you out for a good two weeks of raid gains.

My very first raid I got a necklace, and I was paying back for it three weeks from being in the hole, and it was cheap as some of our items go.

Reward Normalization System: A Zero-Sum Second Sealed-Bid Loot Distribution Mechanic

Nenicirene's picture

Nearly every dedicated raiding guild uses a DKP system to award loot, but most such systems are seriously flawed from the point of view of fairness. (However, since they're always biased towards the implementors of the system, they live on.)

Most DKP systems award points to players for all matter of participatory activities, which are then spent on the acquisition of items. This is fundamentally flawed. Since the amount of participatory activities such as showing up and killing bosses does not correlate one-to-one with the items obtained, any such system will either be inflationary or deflationary over time, in that the points generated by the participants over time will either be more or less than the points required to purchase the items obtained over that same time span. Thus, points should only be given to players based on the actual items that they help to obtain. Any other option will cause problems over time.

As to the value of items, assigning arbitrary values to the utility of items, taking into account potentially varying rarities, is very difficult. The best way to assign such values is by letting the participants decide through a bidding system. Since the system needs to be zero-sum, the cost paid by the winner needs to be distributed to the losers. As to the auction itself, it should be a second sealed-bid auction, as this is provably Pareto-optimal.

Thus, I propose the following system:

All participants begin with a priority of 1.0.

Whenever an items drops that participants desire, everyone who wants it secretly makes a bid between 0 and their current priority.

The highest bidder receives the item. His priority is reduced by the second-highest bid. Everyone else's priority is increased by that second-highest bid times their own bid divided by the sum of all the non-winning bids.

Note that in cases where only one participant desires an item, that participant's priority does not change.

Example: The Blade of Sashimi, an uber-rare weapon, drops for a raid group. Arcadio, Bob, Camilla, and Detritus all desire it, so they secretly make bids. This is their first raid, so they all have priorities of 1.0, and bid 1.0, 0.6, 0.5, and 0.4, respectively. Arcadio wins and pays 0.6, the second place bid. His new priority is 0.4. Bob's priority goes up by 0.6*0.6/(0.6+0.5+0.4) = 0.24. Similarly, Camilla goes up by 0.6*0.5/(0.6+0.5+0.4) = 0.20 and Detritus by 0.6*0.4/(0.6+0.5+0.4) = 0.16. Note that total priority is still 0.4+1.24+1.2+1.16 = 4.0.

Thus, intuitively, individuals are rewarded with increased future considerations every time an item that they want goes to someone else in the group. The points for a given item only flow between the individuals who actually want that item, preventing a net flow of points from groups whose desired rewards are more common relative to the number of members who want them. (That is, if priest and rogue items are equally common but there are twice as many rogues as priests, it is not the case that points will flow over time from priests to rogues, which would matter if rogues also had to compete with warriors for some items or priests had to compete with mages.)

An Addendum

Nenicirene's picture

On further examination, it occurs to me that the system I proposed is not immune to abusive behavior. Specifically, an individual can place a bid they know has no chance of winning on an item they don't really want simply to siphon off some of the winning bid into their own account. That is, if Blades of Sashimi go for between 0.4 and 0.6 normally, a person would can't make use of it could still bid 0.3 on one in order to increase his priority.

Additionally, there is a positive feedback problem in that the rich get richer and the poor get poorer. Specifically, an individual with a priority of zero can never get it to increase. More generally, if a subgroup winds up with high priority scores, they can place inflated bids on items that will get redistributed preferentially to members of that group, keeping them wealthy.

Thus, it looks like I have to toss the proportional redistribution part, and simply distribute the winner's bid evenly among all participants. However, that means there can be inflation and deflation in certain categories. In the example I gave with equal drop rates for class-specific gear and twice as many rogues as priests, there will be a net flow of points from priests to rogues, which will impact other classes that compete with them at times. The end result is that rogue items will become more expensive than priest items, which makes sense if there is greater relative demand. This looks like it calls for some simulation to test long-term behavior of the system.

Are slower weapons better?

Nenicirene's picture

In general, weapon speed has no effect on DPS. Slower weapons will have higher burst damage, balanced out by periods of lower damage. Technically speaking, DPS is your mean damage, while weapon attack speed correlates with the variance in your DPS over a given window. High burst DPS is useful in PvP to overwhelm enemy healers' ability to react. In PvE, it is substantially less useful, and more consistent damage over time is arguably preferable.

Where weapon speed matters is with extra attacks and abilities that otherwise change your attack speed. Shaman abilities that grant extra attacks are based on a chance of occurring on each attack, so weapon speed does not affect the DPS they generate. Slower weapons have more powerful Windfury procs, but they happen less often. Flurry boosts your attack speed for the next few hits, but with slower weapons you crit less often. A comparable-DPS-but-slower weapon is only better if you have abilities that outright grant extra attacks without being based on a chance per attack you make, like warriors and rogues get.

Lastly, there is the issue of attacks disrupting enemy casters, where faster weapons are straight-out better.

Slower weapons may have less Windfury procs, but not by a huge margin...Windfury cannot proc more than once every three seconds, so it's impossible for REALLY low speed weapons to get "runs" of repeated Windfury procs.

This is different than, say, Sword Specialiazation where your proc can trigger a proc, can trigger a proc...same with Ironfoe, Flurry Axe, Thrash Blade...etc.

Nenicirene's picture

It is my understanding that the three-second limit was once true, but was removed at some point and that there are no limits on the frequency of Windfury procs at present.

We will bury you.

Nenicirene's picture

The /who command leads to an accurate count of active characters in a specific area. If you want to tally your point by /whoing every area, doing it daily for a month, counting it up and finding a mathematical ratio, fine; otherwise, don't give me that junk about accuracy, because even then it can't be truly accurate.

That's exactly the sort of data available at WarcraftRealms.com.

So tell me then, how does the census program know if the character it just counted is your main or your 8th alt? How does it know whether or not the character it just counted was an alt you made on a different server just to play once a month?

It doesn't.

Until it does, it can not give an accurate count.

The data on WarcraftRealms.com includes activity-over-time graphs displaying the average number of characters seen at different points in time during the day (which is more relevant than the total number of existing characters). These figures still bear out a ratio of between 2 and 3 active Alliance characters for every active Horde character on PvE and RP Realms, and between 1 and 1.5 active Alliance characters for every active Horde character on PvP realms.

If you want to look at numbers of characters and ignore alts, then consider looking at just level 60 characters, where the figures still hold up. For every level 60 Horde character, there are 2.8 level 60 Alliance on RP servers, 2.2 on PvE servers, and 1.1 on PvP servers.

Try to conceive that for EVERY Horde there are 2 Alliance…DOUBLE. Truly put that idea into your mind; if that were true, no horde raid would ever win, the Undercity would be constantly sieged, and Ironforge would be so flooded you couldn't see.

You have just described a typical day on Argent Dawn.

Strength or Agility for Shamans?

Nenicirene's picture

7 Strength adds 1 DPS, regardless of your character's level. About 20 Agility (at level 60) boosts your melee crit chance by 1%, which amounts to boosting your average DPS (including Attack Power, both natural and from your weapon buff) by 1%. (Note that this means the value of agility relative to strength increases as your melee damage increases.)

If you have flurry (and it were working correctly for a full three hits), then a crit boosts your attack speed for the next three hits by 30% (assuming it's maxed, which if you have it, it almost certainly is). Fudging it, that's like getting 90% of an extra blow in over the next three hits, so your crits are really 190% extra damage, so 20 agility is then a 1.9% DPS boost. (Note that for sufficiently high crit percentages, your flurries will start proccing while you're still under the effect of other flurries, leading to diminishing returns.)

End result:
1 Str increases DPS by 0.143.
1 Agi increases DPS by 0.095% of your total DPS including attack power, if you have full flurry.

These two values become equal at 150 DPS. So, if you're breaking that with your attack power and weapon buff, agility is better. If not, strength.

However, this doesn't take into account the bonus that DPS added via strength already gets from your existing crit rate. To really figure things out:

Average DPS = (Base DPS + Strength/7) * (1 + 1.9*(Base Crit Chance + Agility/2000))

The end result is that you probably want both strength and agility, as their effects multiply each other. This is particularly true given the items available in-game, as it's easier to build up a reasonable level of both than a really high level of either.

I really like this analysis of Strength vs Agility for shamans, and if I every get to log into the boards with my main, with your permission I'll include it in the FAQ.

One thing to keep in mind is that using the windfury weapon buff affects this analysis. Windfury extra attacks count as attacks as far as flurry is concerned. So, if you get a crit, and then a windfury proc on your next hit, you will have used up all 3 of your flurry hits, and so really you only got that speed bonus for 1 hit instead of 3. Additionally, you can get a windfury proc on the second hit instead, which would lead to a speed bonus for 2 of the hits instead of 3.

This would change the DPS increase to 1.3% and 1.6% respectively, instead of 1.9%. The chance of windfury occuring on any single hit is 20%. Thus, the chance of the first situation occuring is 20%. The chance of the second situation occuring is 80%*20%=16% - 80% chance of the first situation NOT occuring, and then a 20% chance for a windfury proc. The chance of a full increase is thus 80% * 80%=64%.

The expected value of the DPS increase of a flurry proc when using the windfury weapon buff is thus 20%*1.3% + 16%*1.6% + 64%*1.9% = 1.732%, not a full 1.9%.

A similar analysis could be used to calculate the exact value when considering "lost" flurries that occur when a crit happens when a character is already flurried.

What about the defense agility adds? Since it's just "Strength vs Agility for shamans, shouldn't that be incorporated? Or am I mistaken as usual? :D

Nenicirene's picture

Yes, Agility, does add defense, but I was just looking at the effect of each on damage in this analysis.

Who to Soulstone?

Nenicirene's picture

In groups of 5 to 15, paladins are preferable, as they are much less likely to die outside of a wipe, which would waste the Soulstone. In the Molten Core, priests, as they are much less likely to die in a bad spot.

Soulstones are wipe protection, and druids can't recover from a wipe due to the timer on their rez. (It's stupid to soulstone the druid just so he can burn his cooldown on rezzing another rezzer first.) If you have multiple warlocks or druid combat rezzes available, then you can consider using them in combat as a means to get vital party members back in after they die, but this is a rare scenario.

I completely agree with SSing the Pallies and then Priests. Are you ready to hear something crazy? When we do Onyxia, and if we have more than 4 Warlocks, we usually toss a spare SS on the main tank, a Warrior! I've seen it work before, when he died he lost aggro so everyone Ceased Fire. He popped the stone, got healed/buffed up while he moved back in to regain aggro. It was sweet. Again...only use if you have a plethora of Locks.

Zero-Sum Systems and Member Rotation

Nenicirene's picture

Maerlyn asks:

It was actually reviewing that page when the thought hit me, Delaw. The whole concept behind zero sum is that zero sum... but if you snag 40 people out of that listing at random and put 'em in a raid you don't get a zero sum.

Yes, points are re-distributed among the raiders present when an item drops and the bid goes through, but those not there at the time DON'T get those points re-distributed to them, and you do in fact start getting the more regular/frequent raiders accumulating more points in toto.

I still may be missing something, but it's still looking to me like a system predicated on a belief points are being re-distributed evenly, when in fact--when you actually stop to examine it--they're not (in anything outside a 40-man static raid group). I just can't see how zero sum is in any way compatible with a rotation.

The intent is not to distribute points evenly to everyone ever entered into the system, but to distribute them evenly to people based on their participation in successful effort. (The successful part is causing some issues in BWL, but that's another issue.) The reason for using a zero-sum system is to prevent inflation. Since points are redistributed between people as opposed to created by raiding and destroyed by winning loot, the average point total of all individuals in a given raid will average close to the starting amount of 1000 points.

This should ensure that your points are worth the same amount regardless of how frequently you raid. Yes, someone who raids twice as often will earn twice as many points, but will have twice as many opportunities to spend those points. The key is that the net effect point-wise of long-term combinations of earning and spending is zero. Thus, the person who raids twice as often will wind up earning twice as much loot, as it should be.

Systems that are not zero-sum tend to have inflationary or deflationary pressure over time. Specifically, in inflationary systems, the points coming in exceed the points coming out, and thus if you're not raiding at every possible opportunity, the relative spending power of the points you do earn will not keep up with the people who are raiding as much as possible. Under a deflationary system, which is quite rare in practice, different sorts of problems crop up depending on what sorts of rules are in place regarding the ability to purchase items and variation in prices, but the net effect is that newcomers and infrequent raiders tend to win more than their fair share of items. (And since most systems are designed and administered by the most frequent raiders in the group, you can see why such systems are rare.)

Will the 1.10 changes make +healing gear better or worse for priests?

Nenicirene's picture

Better.

The casting time of a spell only affects how much of a bonus the spell gets from +healing (or +damage). The bonus you get is constant, as long as you're casting anything.

Due to the scaling by the casting time of the spell, +healing gear effectively adds constant healing per second (and similarly for +damage). That is, +X healing is normally +X/3.5 healing per second regardless of what spell you're casting. (We'll ignore Heal Over Time spells here.) When Greater Heal was 4 seconds, it wasn't actually performing at that rate. Also, spells whose casting time has been reduced by a talent receive bonuses from +healing gear at their original casting time, not their new, shorter casting time. This means that for a constant reduction in casting time, you get more effect the shorter the original cast time.

Looking at the actual numbers, if you have +100 healing you get:

Old Untalented GHeal: +100/4 HPS = 25 HPS
Old Talented GHeal: +100/3.5 HPS = 29 HPS
New Untalented GHeal: +100*(3/3.5)/3 HPS = 29 HPS
New Talented GHeal: +100*(3/3.5)/2.5 HPS = 34 HPS

The Granularity of Raid Rewards Discourages Casual Behavior

Nenicirene's picture

An actual intelligent post from the General Forum:

Why Casuals Whine: The Paradigm Shift at 60
As for casuals vs. raiding, here's what casuals want.

Log on whenever they feel like it. Might be for 5 hours, but they want to be able to log on when they want, not because someone told them to log on.

Either make some progress solo or get a group together and go do something for the next 3-4 hours that furthers their character.

Expecting this is not unreasonable, considering the entire 1-59 PvE game has trained them to think this is feasible. They don't want to have to deal with scheduling when to play the game when they haven't had to for the last 59 levels...

The paradigms for both raiding and PvP break this expectation.

In Raiding, casuals no longer can log on when they want. Instead they must re-arrange their lives around a game, a concept that many of them find paradoxical. "If I'm playing a game to have fun, why should I rearrange my life around the game?" "Why should I have to turn down a spontaneous invitation to go out with my friends because I have to attend a raid? Why couldn't I just attend that raid the next day? At level 45 I didn't have to go to Uldaman only at set times."

A good solution to this is changing the lockout system drastically, going so far as to eliminate lockout entirely, instead replacing it with a concept I like to call "lootout." If you loot a boss, you can't loot that boss for another (what the current timer is for the raid). If you loot Ragnaros you won't be able to loot Ragnaros for another 7 days, but you'll be able to loot something you need off of Garr in a different raid.

This would be a huge step forward as it would remove the regimented and schedule aspect of raiding that is a big turn off to many casuals. It isn't that they don't have the play time, it's that they oftentimes don't know when that play time will be, because of commitments to work, family and friends. WoW is something they greatly enjoy. They can't rearrange their schedule around WoW, however.

They could play WoW when they wanted to from level 1-59 and don't understand why, for raiding, WoW requires them to play at certain times not of their choosing. Basically WoW's own game mechanics are coming back to haunt it.

A similar problem appears in PvP.

You rank up based on your standing per week, which means if you miss a week entirely due to going on a business trip, you derank, losing all your progress. This idea is abhorrent to the game mechanics that WoW has from levels 1-59. If a casual doesn't log on for a week at level 45, he doesn't LOSE levels. Instead, he gains rested xp so that when he comes back he will be able to more effectively GAIN levels. In PvP, it's backwards from what WoW has taught casuals to expect, and thus they whine. They want to be able to log on whenever they want, play for 1-6 hours, and make progess. Right now if you try PvPing like that, you'll peak out and Rank 6 or so, and not be able to go any further until you devote more and more time to PvP, until Rank 13/14 when you're going to have to be playing 100+ hours to get those top standing spots just to rank up.

Solution: Move to a progression system more like XP, where sum contribution influences rank, not quantum contribution per week.

The current raiding and PvP game is completely different from what WoW has taught players to expect from levels 1-59 and that's why they're complaining.

A tiny side note here: the Tier 0.5 quest mechanics are excellent. Casuals can log on at their choosing and make progress towards compleing their set. This is in tune with the level 1-59 game and is a huge step in the right direction.

WoW has dramatically expanded the size of the MMOG market; most WoW players, I would venture to guess, have never played any other MMOG. That means they don't understand when you say 'oh well it was like this in EQ' and don't understand why it has to be that way. WoW was advertised as casual-friendly, and it is, amazingly so, up to level 60. Then suddenly the game rudely changes its primary game mechanics at level 60, much to the surprise and disappointment of the casual player.

Good points in general. Changing to a more pickup-friendly endgame experience, however, would require a restructuring of the current raid reward mechanics. Most of the goodies you want are single drops off bosses, which must be assigned to a single individual out of a twenty or forty person group. While random assignment is technically fair, it has a huge variance in individual fairness over the short term, so people develop point systems that act as smoothing functions for the loot received.

In order for pickup-style raiding to work, the reward mechanic needs to be smoothed by the game itself, not by player agreement. Moving toward raid rewards based on faction (which everyone gets) and token collection (which breaks up a single item into a bunch of smaller drops, effectively keeping the mean amount of loot you get constant while reducing the variance) are steps in this direction, but to truly make raiding with different groups of people something other than a total crapshoot, basically all the raid rewards need to be moved to this system, with individual uber-drops eliminated. (Alternately, the game could implement some sort of built-in point system, which players can pay each other in order to have the rights to a single good drop—a sort of second currency only useful for BoP items, enforced by game mechanics, though there's potential market abuse issues there unless it's very carefully designed.)

Lead, Follow, or Stand in Back Healing

Nenicirene's picture

There are really only three roles:

1. Kill Things
2. Convince Things to Kill You Instead of your Friends
3. Keep Things from Killing your Friends

We're firmly in camp 1. Camp 2 is actually a synonym for warriors, end of story. Camp 3 is anyone with even the most pathetic tertiary healing ability. Everyone has some slight flavor variations on how they fulfill their role, but they're actually pretty minor. Hate it or love it, it really is that simple.

but im in all 3 roles!!!111!

Just like a breeder reactor…

Nenicirene's picture

Hell I would support a resetting hardcore server which of course is more than the low-pop whiners could stomach.

I think it might be a good idea to have "seed" servers. Say, three servers each of which resets every three months, staggered so there's a fresh one monthly. The key, though, is that when the reset happens, existing characters get to transfer to permanent servers. This is basically how the ladder seasons in Diablo II worked. It gives the reroll locusts somewhere to go and stop pestering the rest of us, and, in the PvP case, caters to people who want to level without waves of high-level gankers with nothing better to do. Additionally, it doesn't require continuously adding more and more hardware, and hopefully injects more people into current ghost-town servers.

Syndicate content