Thursday, October 1, 2015

[WoW] Revisiting 10 vs 30 Player Raid Difficulty

A while back I wrote a post on why 10-player raids were mathematically more difficult than 30-player raids as they were designed at the time from a DPS perspective. This was written as we were about to move into Blackrock Foundry from Highmaul.

Well, it's nearing the end of Warlords and apparently either the WoW developers read my blog post and agreed with it--which may have been likely, given it got Reddit bombed and makes up like 40% of my total lifetime traffic even today--or came to the same conclusion as I did on their own.

Raid Size Related Hotfixes

Let's take a look at all of the fight nerfs made after each raid was released to account for small group sizes. Note, I am ignoring changes that are basically scaling bug fixes rather than changes in the encounter fabric, but I did consider adding changes where they prevent mechanics from overlapping (ie: bosses putting two effects on the same person). Clearly that would affect a smaller raid disproportionately compared to a large raid, but it still could significantly affect a large raid if they had a string of bad luck, so I opted not to include those.

(6.0.3 Hotfixes)
  • Gruul now uses Petrifying Slam against 5 targets in a 10-player raid, scaling up to 10 targets in a 30-player raid (used to be 8 targets for all raid sizes).
  • Reduced the damage that Gruul’s Inferno Slice deals to 10-player raid groups on Normal and Heroic difficulties, but increased the rate at which damage scales with raid size. Larger raid groups should find the damage relatively unchanged; smaller groups will find that the ability now deals less damage.
(6.2 Hotfixes, More Hotfixes)
  • Adjusted scaling to reduce the amount of health by roughly 10% on Xhul'horac, Omnus, Unstable Voidfiend, and Vanguard Akkelion for a 10-player raid on Normal and Heroic difficulty. Health should remain the same as before for raids with 30 players.
  • Adjusted scaling to reduce the amount of health by roughly 10% on some Hellfire Reinforcements and some Felfire-Imbued Siege Vehicles for a 10-player raid on Heroic difficulty. Health should remain the same as before for raids with 30 players.
    • Hellfire Reinforcements: Contracted Engineer, Gorebound Felcasters, Iron Dragoon, Hulking Berserker
    • Felfire-Imbued Siege Vehicles: Felfire Artillery, Felfire Crusher, Felfire Demolisher, Felfire Flamebelcher
  • Health of adds for [Kilrogg Deadeye] have been reduced by roughly 20% for raid groups with 10 players on Normal and Heroic difficulty. This should help with the difficulty experienced by smaller raid groups losing their DPS when sending players down into the death phase. The health of adds scales back up and should be roughly unchanged from before as raid group size increases.
(6.2.2 Hotfixes)
  • We have made a number of changes aimed at reducing the difficulty of the Archimonde encounter for smaller raid groups on Normal and Heroic difficulty, with a focus on the final phase of the encounter in particular. These changes are offset by increased scaling with respect to raid size, so that the experience for larger raid groups will be mostly unchanged.
    • Infernal Doombringer's health has been reduced by up to 15%.
    • Shadowed Netherwalker's health has been reduced by up to 15%.
    • Living Shadow's health has been reduced by up to 15%.
    • Reduced the damage of Wrought Chaos by up to 20%.
    • Reduced the damage of Shackled Torment by up to 20%.
    • Living Shadows now spawn from each Nether Tear at a reduced rate for smaller raid sizes.
It's interesting to note how many more changes they made to Hellfire Citadel specifically versus Blackrock Foundry, though it's possible they didn't really have the data they needed in BRF to adequately decide that small raid sizes were struggling versus larger. Or the mechanics in BRF just didn't tend to lend themselves to scaling issues.

Most of the fights in BRF didn't involve non-tanks handling adds for the most part, aside from Blackhand or Furnace, and only Oregorger really had an interrupt rotation and it was pretty simple; adding more players wouldn't change the difficulty there.

Iron Maidens Case Study

Iron Maidens suffers the same issue that Kargath suffers except worse: you lose half your raid every boat phase. Because the boat phase always has the same number of players on it, the boat itself shouldn't scale with raid size, meaning that you're losing a disproportionately large part of your raid for the same amount of time if you're a 10-player raid. In fact, if you look at's data, you'll see that Uk'urogg does have 5.45M health on all raid sizes on Normal difficulty, but Admiral Gar'an Normal's health scales up linearly ~5.5M health per player, so you should always send the same number of players to the boat, regardless of your overall raid size.

If you assume 2 boat phases, you're looking at about 45s for one boat phase, and 1m15s for the the other at 50th percentile DPS, give or take a few seconds from either, and in that case about an 8 - 9 minute fight. Meaning 25% of the fight, you're missing literally half your raid, and most of them DPS, which means you're not really putting damage on the bosses aside from the tanks, or 1 DPS and a tank. For ease of calculations, let's pretend you always send 4+1 tank, so about 4.75 DPS out of a total of 5+1.5 DPS (2 Tanks).

For 10 players, you're losing  ~73% of your total DPS for 25% of the time, so a net loss of 18.2%.
For 11, ~63.6% of your total DPS for 25% time, so a net loss of 15.9%.
For 12, ~55.9% for 25% time, so net loss of 14.0%
For 13, ~50.0% for 25% time, so a net loss of 12.5%
For 14, ~45.2% for 25% time, so a net loss of 11.3%
and so on.

Clearly an inverse exponential curve in terms of lost DPS. Case closed, right? Not quite. In a twist, it turns out that while the boss health does increase linearly, Blizzard has it increase somewhat  faster than the amount of DPS you're adding, ostensibly to handle the fact that 10-mans have it harder. Also interesting to see just how hard adding a healer hits due to the linear scaling a little faster. Healer bumps are kind of arbitrary, but don't affect the overall curve.

Iron Maidens Normal. It's backwards!
Hans'gar and Franzok Normal. This is what we usually expect boss health per DPS player curves to look like.
It's not really perfect, because the low end still requires exponentially more DPS to make up the loss, and the compensation is linear, but it's clear that Blizzard thought something needed to be done. The question is, though, was it "good enough"? Probably, though as usual unless we have access to oodles of failure rate data for raid size vs. attempts, I couldn't tell you for sure.


I'm pretty sure this is the part where I get to say I totally called it, if I weren't so modest. ;) 

That being said, Blizzard is clearly now designing their encounters with raid size in mind, or at least nudging them after the fact, so hooray for us smaller raids! Perhaps they should consider adding, "How does this encounter play with 10 players?" to their testing/design review checklist so we don't need to wait a month and a half for hotfixes because they're waiting for data rather than performing in-house analysis.

And let me say for the record that Xhul'horac as a 10-player raid is absolutely Xhul'horrific. Half your raid dealing with surge debuffs every 30 seconds, not enough raiders to have folks reliably set for clearing fire via explosions, not enough people to deal with adds while that's all occurring. It's clear to me at least that Xhul'horac, despite the fact they compensated him health-wise, still sucks the big one mechanically for tiny raid sizes. Even having 1 extra person the next week made the fight immensely easier.

There's still the open question on raid cooldowns. 3 healers for 10 players means fewer tools than a 20-player raid (both in the literal and ideally the figurative sense, but that's my bias against large raid sizes showing). Being able to chain, say, Devotion Aura for twice as long is a pretty big bonus given it perfectly scales with raid size. Perhaps that'll be my next project one day, analyzing healing and damage output to see how that changes over raid size and compare that to tools available.

But in the meantime, I'm glad to see that efforts are being made around smaller raid sizes. WoW is pretty much the only game out with variable raid sizes with razor balance, so it's new territory with new design problems.
#WorldOfWarcraft, #Theorycrafting, #GameDesign


    1. "For ease of calculations, let's pretend you always send 4+1 tank, so about 4.75 DPS out of a total of 5+1.5 DPS (2 Tanks)."

      Wait, what? Why wouldn't it be 3 DPS + 1 Healer + 1 Tank?

      "But in the meantime, I'm glad to see that efforts are being made around smaller raid sizes. WoW is pretty much the only game out with variable raid sizes with razor balance, so it's new territory with new design problems."

      Except WoW isn't -- that's precisely why Mythic is exactly 20 people because they want to get razor balance. They couldn't even manage it "right" with only two raid sizes to consider (10 vs 25).

      I mean, something like Mythic Mannoroth requires two DKs for mass grips on the imps. How would that possibly work for a group of 10 people? You can't even have one of every class in a group of 10 people (and if you had more than 10 people on your raiding roster then you'd have them in the raid since it scales) but you can have 2 of every class in a 24-25+ raid roster for Mythic. Does that mean an Imp wave would come half as often (and you're just screwed if you don't have a DK at all)? How would that scale -- you can't just decrease the timer if you add more people as it's a binary thing. Either you have mass grip ready or you don't. An imp wave every 59 seconds is the same as an imp wave every 30 seconds as far as a DK with a 1 minute mass grip is concerned -- can only get every other one in both cases.

      I mean, sure, WoW actually is *trying* to loosely tune the content in comparison to many games where the 20 person raid could be done with 10 people because they have no idea of tuning but...hardly razor balance with flex tech.

      1. Ah, whoops, yes, thanks for the correction on the numbers. That does change things slightly, but really brings it a little closer to parity with their fixes. Good call.

        Balance numerically is different from balance mechanically, noting that mythic brings new mechanics that basically require specific class abilities. Normal/Heroic fights are specifically designed to not require specific class abilities.

        I suppose razor balance is probably too "generous" for Normal/Heroic, but my point about being the only game in town trying on this aspect is still true.

      2. I'll take your word for it about the only game in town even trying, but yes, I think you're being quite generous -- especially given how widely specs can vary on fights. Look at, say, Archimonde for example. A group of mostly single target/large AoE specs is going to have a very different requirement compared to a group with mostly cleave/multi-dot specs (since there's so much cleave/multi-dot on that fight).

        And since Normal (and even Heroic to some degree) aren't supposed to require certain specs or anything, you can't really make the argument that people "have" to switch to certain specs on that fight, should be able to grab whatever and go. Which means a group could easily be required to output 10% more or 10% less in effect (skill wise) compared to another group.

      3. Eh, we have enough trouble on fights where our comp is unfavourable that I think we needed to overpower it with gear--and ideally more skill, but I think we're actually fairly above-average for Normal raids given our statistics so far in terms of pulls/first kill. Kilrogg and Xhul'horac are a couple in mind.

        That being said, I agree with you in principle. That's just not necessarily reality, and sometimes I wonder what principle Blizzard is aiming for, because sometimes it feels like they have 8 different designers all with different principles making their bosses (and quite likely, they do :P)

      4. If you're in the raid, I'm pretty confident your group is above average skill-wise for Normal.

        In general, I think Blizzard just wants to make different and varied boss fights...and figures precise tuning for flex Normal/Heroic isn't an issue (and that was one reason why it got easier with more people -- law of averages and intentionally "less" scaling per person).

      5. Aw thanks. We did just literally one-shot Mannoroth Normal despite at least 33% of us never seeing him before period, and pretty sure most of the others only saw him in LFR once or twice, so that was exciting.

        And agreed about the varied boss fights. Really, that's where the fun is anyhow.