Burning Crusade crunch had finally wound down… and those of us that composed the quiet little Blizzard guild decided we would take on the challenge of defeating Illidan together. Jeremy, our raid leader during Molten Core and BWL, decided that he wanted to relax a bit more and asked for volunteers to help run raiding teams for Karazhan and beyond.
Joe and I, both raid designers who missed our former raiding guilds (back in those days, if you were part of the raid dev team, it was considered inappropriate to remain part of a competitive raiding guild – so we’d both quit when we started at Blizzard) volunteered to run 10-man groups.
There’d been a fair share of bumps along the way this expansion – there’d been overcrowding in Hellfire, server performance issues in Shattrath due to the lack of Line of Sight blockers, but overall, the expansion had gone pretty well. Players were happy and the experience of seeing the content we’d created enjoyed by millions was a huge thrill.
After weeks of running our individual groups, Jeremy declared it was time to forge our groups back together to take on Gruul, Serpentshrine Caverns, Tempest Keep and beyond. It had been a long time since I’d had a chance to organize the warlocks (which had been the role I held in my previous AQ raiding guild). So with a relish I started planning out curse macros, ability and specialization plans based on the team we would field and figuring out exactly which spec we each should play. My inner control freak was having a blast – we’d be ready to kick ass in no time.
Selecting the best geared Warlocks from our guild, it was decided Mather, Kandroth (my orc lock, as opposed to Kanrethad, my Alliance Warlock) and Garlock would be attending. Mather loved Affliction, so she was assigned the uber shadow curse, while Garlock was assigned Destruction and curse of elements, leaving me with the despicable, yet my favored spec – Demonology and the final curse slot to assign based on the individual fight.
It came as a bit of a shock when after a few nights of raiding, when Nathan, the Shadow priest assigned to the Warlock DPS group, whispered me and said… “I think you are playing demonology all wrong.”
Defensively, I replied: “What do you mean? I am keeping up my dots and using Doom only where appropriate.”
Nathan: “Yeah… but why are you pushing anything besides shadow bolt?”
The One-Button Warlock
Needless to say I was at first arrogantly derisive towards the suggestion, but after being challenged to a DPS test, sure enough, Nathan was right. by specing deeply into destruction and always sacrificing my Succubus, I was able to achieve consistent, high and visually satisfying damage by just spamming Shadowbolt.
I went in the next day to Kris, the numbers Guru, to figure out what was going on.
Kris: “Bah! I TOLD someone this would happen eventually.”
Me: “Can you explain it?”
Kris: “Of course! What do you know about scaling?”
Me: “Uh, just that you want things to be about the same power.”
Kris: “There’s a lot more to it than that.”
Stepping over to the whiteboard, Kris, a former middle school teacher, flipped into education mode.
Kris: “Scaling is the rate at which power grows. Let’s take a look at how damage increased in Classic wow.”
He pulled up a chart of Shadowbolt’s min and max damage for levels 1 through 60. He created a series of dots on the rough chart, one of each max damage and spell rank.
Kris: “As you can see, the damage of Shadowbolt increased at a rate of 10-20 damage per spell rank from rank 1 at level 1 to rank 11 at level 60. This is very flat, linear growth.”
Kris drew a flat diagonal line going from 40 to 220.
Kris: “It’s very controlled. No problem. However, raiding gear changes things.”
Kris: “Dungeon and Raid content at level 60 demanded that the offensive power of casters increase. As you may have noticed back then, Rogues, Warriors and Hunters were incredibly strong, because their gear allowed for their damage to continue to increase with the Strength and Agility stats.”
Kris: “Casters had +Fire, +Shadow, +Arcane and similar gear leveling up, but then +Magic damage was added to reduce the amount of gear that needed to be made for dungeons.”
Me: “I follow you so far, but okay, the power of the gear increases your spell power by some amount. So what?”
Kris: “Well, it wasn’t enough for the gear to just increase in power. We did some research and realized that just increasing spell power by 10 doesn’t feel very good each tier of raid gear. Going from 10 to 20 feels great… and going from 100 to 110 feels okay… but going from 320 to 330 feels insignificant.”
Kris: “After some experiments and trying out variations with different groups of QA testers and friends, we came to the conclusion that people don’t notice or appreciate increased in power if they are lower than 12.5% – let’s call it 13% to make it easier.”
Me: “This makes sense…”
Kris: “So… during the original level-up phase, gear was based upon the level at which you acquire it. Level 10 white gear is generally what you would have from a vendor at that level. Green, Blue and Purple (Magic, Rare or Epic) gear tends to have a flat bump up from that level. So an rare at level 30 might be equivalent to gear you’d find from a level 50 vendor.”
Me: “With you so far… this is really starting to drag on, how does it connect back to Warlocks?”
Kris: “Hang with me just a bit longer. In dungeons at level 60, we could award blue gear. Then in Molten Core, we release epic. Naturally those are item level 70 and 80 respectively. But what happens when BWL comes out?”
Me: “You start giving out Legendary gear?”
Kris: “Right, but then we would have to create legendary gear for everyone and suddenly, Sulfuron, Thunderfury and Atiesh become far less special and memorable.”
Me: “I see. So instead, you just cranked up the item levels instead.”
Kris: “Right! Due to the way Bob (a programmer from Starcraft who’d worked on WoW back in the day) setup the initial data tables for WoW, 1 item level is roughly equal to 1% more powerful.”
Me: “So this is why gear goes up by 13 item levels between raid tiers.”
Kris: “Bingo. Likewise, spell power increases by 13% per tier of gear.”
Me: “Soo…..”
Kris: *sigh* “Alright. Let’s sit down and look at the breakdown of Warlock talents in the Sacrifice, Shadowbolt spec. “
Talents
Warlock Talents
Minimum Required Level: 70
Required Talent Points: 61
Demonology Talents – 21 point(s)
Improved Healthstone – rank 2/2
Improved Imp – rank 3/3
Demonic Embrace – rank 5/5
Improved Health Funnel – rank 1/2
Fel Domination – rank 1/1
Fel Stamina – rank 3/3
Demonic Aegis – rank 3/3
Master Summoner – rank 2/2
Demonic Sacrifice – rank 1/1
Destruction Talents – 40 point(s)
Improved Shadow Bolt – rank 5/5
Cataclysm – rank 2/5
Bane – rank 5/5
Devastation – rank 5/5
Shadowburn – rank 1/1
Destructive Reach – rank 2/2
Improved Immolate – rank 5/5
Ruin – rank 1/1
Emberstorm – rank 5/5
Backlash – rank 3/3
Soul Leech – rank 1/3
Shadow and Flame – rank 5/5
Analysis
Me: “So the big one is Demonic Sacrifice.”
Kris: “Is it really though?”
Me: “Hrm…. well, it’s one point that provides 15% of damage. That HAS to be the broken one.”
Kris: “Sure, it looks like that, but you need to factor in more than just the talent. What did it cost you to get there…”
Me: “Well, let’s see… there’s improved armor, which is good, improved health at the cost of spirit, then about… 10 points in talents that are used to improve your pets.”
Kris: “10 points which now no longer do anything. Add in the extra 5 points of value you should be getting from ‘gold medal’ talents and suddenly you’ll see that you’re getting 15% of Shadow damage at 20 points worth of talent cost.”
Me: “Strange… it seems like it’s worth it though.”
Kris: “Kevin was going for versatility when he put together the demonology talents. Letting you have the better pet talents and also be able to kill-off a pet for raw damage, was an idea to let Warlocks still be able to be Demonology and go raiding.”
Me: “That’s also why the demon pets all got AoE damage reduction, right?”
Kris: “Yeah. No one’s going to take the time to heal a pet.”
Me: “Okay, so where’s the problem.”
Kris: “If the problem isn’t in column A…”
Me: “… yes?”
Kris: “… never mind. Take a closer look at Destruction’s talents.”
Me: “Alright:“
- ISB – increases direct shadow damage by 20% from 4 sources, triggers on brits.
- Ruin – increases critical strike damage bonus to 100%.
- Backlash – increases critical strike chance by 3%.
- Shadow and Flame – increases the bonus spell damage for Shadow Bolt and Incinerate by 20%.
Me: “A lot of direct damage increases and crit boosters.”
Kris: “Do you remember what I was saying about scaling factors? Spells grow from two sources – base rank-up (leveling) and gear (spell power).”
Me: “Ah, it has to be Shadow and Flame, right? That increases the growth rate of Shadow Bolt and Incinerate”
Kris: “Shadow and flame is a definitely a big factor. But it alone isn’t the big issue. Otherwise, you’d at least be putting up Immolates, Incinerating then throwing up Conflagrate at the end of the spell.”
Me: “So what’s left?”
Kris: “You forgot something – Bane.”
Me: “Bane…? That just makes Shadow Bolt not take a long time to cast, which means you get more per time, but it doesn’t change the amount of max mana you have. It just means you have to start life-tapping sooner.”
Kris: “Well, yes, but look at it this way. What’s the spell power coefficient (multiplier) on Shadow Bolt?”
Me: “Well…. huh, it’s 0.877.. roughly 6/7. ”
Kris: “Yeah. Back in the day, Warlock Shadow Bolts and Mage Fireballs were 3.5 second cast time spells. This was picked as the “Baseline” for spell damage.
Me: “So spells with a damage coefficient less than one have a cast-time less than 3.5 seconds.”
Kris: “Right. At some point the cast time of Shadow Bolt was reduced down to 3.0 seconds… but the damage coefficient wasn’t slightly too high. This information isn’t shown in the tooltips, so it went unnoticed for a while now – spell power didn’t get up to huge levels, so the difference was hidden by the overall randomness of the spell’s damage.”
Me: “Until it was increased by 20%.”
Kris: “You got it. Suddenly, a spell with a 2.5 second cast time has a 1.02 spell damage coefficient. Let’s compare some alternative spells and rotations.”
- Immolate: X damage + 0.2 coeff + 0.13 coif per dot tick (5)
- Incinerate: X damage + 0.57 coeff
- Corruption: X damage + 1/6 coeff * 6 ticks
- Curse of Agony: X damage + 1.0f coeff
- Curse of Doom: X2 damage + 1.0 coeff
Me: “Huh…. nothing has a coefficient higher than 1, not even Curse of Doom.”
Kris: “Yeah… Curse of Doom should probably have a much higher coefficient… I’ll make a note to fix that. But for now, do you see what’s going on?”
Me: “Yeah. Because there’s no other button with a better coefficient, there’s nothing better to cast. I mean, technically, keeping up corruption would be a good idea, but since you can’t pick up all of the talents to make it instant and get the uber Shadowbolt, it kind of washes out.”
Kris: “Yeah, it’s a marginal increase instead of a significant one. Furthermore, Corruption doesn’t benefit from Improved Shadow Bolt… so suddenly a 1.02 becomes a 1.14 coefficient spell.”
Me: “My head hurts. I just realized that as crit chance increases, not only do you get more critical strike damage – but you also boost the power of the other warlock’s shadow bolts even more frequently.”
Kris: “Yeah. It’s kind of broken. But we aren’t making any big changes until we can patch, so good luck and go have fun with your shadow bolt spamming.”
Me: “Heh.”
Kris: “We’ll take a look at how to go about fixing this sometime in the future.”
Segue – interesting and strange talent choices
If you take a close look at Demonology in Burning Crusade – you’ll notice a number of talented related to defensiveness. Including a 3% anti-crit talent that also reduced pet damage taken by 15%. Demonic resilience was created with the idea that like in AQ, Warlocks might be called upon again to be temporary or caster tanks.
In the end, this ended up being a mostly PvP talent, with marginal benefit.
Serpent shrine and Beyond
I went back to the Warlock guild channel thrilled to let them know we could just focus on one-button spamming for the remainder of the patch. The reaction was not as I expected.
Garloc: “Whoo hoo! Easy mode!”
Mather: “Oh…. really?”
Me: “What’s wrong?”
Mather: “Well, I was really enjoying affliction. I liked the DoT cycling and always felt like I was getting better at not wasting time and mana. Life tap makes me feel kind of unsafe, so Dark Pact was a perfect fit for me. “
Me: “Huh… yeah, you know, I really miss having Soul Link and being able to live longer from an in-a-pinch Voidwalker to save me from a wild trash mob.”
Garloc: “Yeah, that happens when Born isn’t tanking.”
Me: “Mather, you know, we’ll still want the 2% extra shadow damage for us locks and the three shadow priests. So if you stay affliction, I think it will be alright.”
Mather: “Oh good, because I was going to do that anyways.”
Me: “lol, nice.”
Garloc: “But yeah, it kind of sucks – we’re down all of the way in destruction, but we aren’t even getting to use the super cool AoE stun ability. Like… its kind of lame.”
Me: “I hadn’t really considered that.”
Takeaways
- Talents are tiny, but potent.
- Especially multiplicative talents. 2% more doesn’t mean much. 2% stacked multiplicatively 30 times is 80% stronger.
- When building a spell rotation, you need to look at base damage and coefficient contribution.
- At a baseline, if there’s a better button to push (better coefficient per cast time), you should push that instead.
- All together, if you attach too many bonuses to the same thing, they’ll get out of control.
- Balance and rewards are important
- Warlocks who aren’t into that style of play will give up on mastering the play style they enjoy
- Bonus effects brought by a spec or talent tree, if big enough, can give those players an excuse to stay in their play style even when its lower in damage.
- A fixation on damage throughput degrades the play experience and denies the player of interesting and thought provoking tools.
The main question I have from this is: you guys clearly knew about this fairly early on in the expansion. Why, then, was it never fixed until the end of the expansion? Were you guys just prioritizing not rocking the boat re: existing content balance so you let the status quo that raid teams built themselves around stand?
I’m seeing a lot of the same puzzlement as when Greg Street (aka GhostCrawler) expressed surprise that players would optimize for maximum efficiency. My first reaction was: are you kidding me?
Of course we do. This is the Skinner Box that was built for us. We’re all aiming to do our best in raids, and analysis of combat logs and damage meters and sims will always lead us down the same paths.
It’s kind of a counterpoint to my earlier comment about how I loved the more complex and flexible talent trees. The reality is that there were a few spec combinations that paid off, so you went with the one you liked the best out of that handful and hoped that skill would make up for any deficit.
This is not game theory as much as social engineering, when it comes down to it. You practically have to have Hari Seldon on staff.
Uhm… are you… alright?
Whoa, pardner, didn’t mean to freak you out there. Don’t you fret.