stat progression in MMOs, 01 Jan 2025.txt

Random essay on stat progression in MMOs

Why have stats on gear at all, instead of just having "armor, and better armor"? Stats make players think and give them an opportunity to choose what they want their character to be good at, while also encouraging players to play the game more to get armor with unusual stats.

How should stats support character progression? Consider a game with two categories of abilities: magic abilities, and physical abilities. At any point in time, a character is only performing one ability at a time, and each type of stat on gear supports either magic abilities or physical abilities. If there are no downsides to always using one category of ability, players will tend to exclusively use one category of abilities and wear gear that supports only that type of ability.

We have the same situation if we increase specialization on gear, such as separating magic abilities into frost damage, fire damage, shadow damage, and so on, and making gear only support one type of magic damage. A game can be successful doing this, and it may seem like an obvious thing to do: letting players make their characters different from other characters. But it isn't the best way to do things.

If a game has five types of magic damage and three main types of doing physical damage, then players who fully specialize into one of these can be divided into eight groups. They're completely different from characters in other groups, but will tend to be very similar to characters in the same group.

If, on the other hand, there is just one type of magic damage and one type of physical damage, but gear that supports both types of damage is almost as effective as fully specialized gear, then instead of just two groups of characters, players can choose any point along the entire continuum from fully specialized in magic damage to fully specialized in physical damage. They end up with more choice, and more customization, than the system with eight different types of damage where players cluster at the extremes of specialization into a damage type.

Two Thirds of People who Clicked on This Post Already Stopped Reading

I'm basically coming from a World of Warcraft Classic perspective: "why is leveling with a shaman so different from endgame, where most shamans are forced to healbot?" A big part of the answer is gear. World of Warcraft has a rudimentary system of encouraging diverse stats on gear, in which an item's budget is sort of calculated with a Root Mean Square system, except with a different exponent like 0.7 or something. This can be seen with random suffix items like "of Strength", vs "of the Gorilla", where the Gorilla item has something like 40% more total stats than the Strength item because the budget is split into two types of stats. But this system isn't enough to ensure that gearing for multiple roles is viable. The penalty for hybrid gear is still too high.

This can be fixed. In World of Warcraft, the Season of Discovery version of the game combines stats like chance to hit, and chance to critical hit, into a single bonus that affects both physical attacks and spells, whereas in the original game they were split. So let's suppose a game lets you either get 100% scaling in a single category of abilities, or 90% scaling in physical damage, spell damage, and healing. In WoW Classic, shamans are an example of a class that normally uses all three of these categories while leveling but is forced to gear for one of them at the level cap.

If we do this, we run into another problem: healing efficiency and mana pools. It's common for games to have a stat that increases the number of spells, such as heals, that can be cast, before running out of mana. (Scaling can cause this in other ways as well: for example, in WoW Classic, paladins recover mana on critical heals, which scales faster than linearly with gear.)

I'll just mention the only other MMO I've played other than WoW: in Aion 1.x, heals did not scale with gear. This resulted in a nerf to PvP damage because heals were not keeping up, but it deserves mention as an attemped solution: mana pools may increase, but if healing per second stays the same (and mana pools didn't matter much in the first place because of low potion cooldowns), then the larger mana pools don't affect PvP much.

WoW is different. Healing-per-second is generally at least twice the damage-per-second of spells, while also usually having slightly better effectiveness per mana spent. So both in PvP and PvE, it makes sense to use healing spells. The power of healing is balanced by mana pools. In WoW Classic, 1 stamina gives 10 HP while 1 intellect gives 15 mana: if stamina and intellect are roughly equal, then mana pools will be about 1.5x health pools, and if healing efficiency is about 2 healing per mana, then a healer can heal 3 full health pools without mana regen.

A typical stat for an MMO is "increases the effectiveness of heals by X amount." This does not increase the mana cost of spells. So let's give characters gear that doubles their effectiveness for a single category of abilities or increases it by 90% for all damage and healing. (Using a more complicated version of WoW's root-mean-square-like calculation.)

Without mana or health increases, a pure healer can now heal 6 full health pools, while a hybrid healer can heal 5.7 health pools. If we also double health pools, then we're back at healing 3 health pools, and the game is fine.

But if we also double mana pools, then the pure healer can heal 6 of the larger health pools, although it takes twice as long.

It may be possible to balance PvE for this, but the pacing might not be fun. It's more difficult to balance PvP. A lot of the time in Classic WoW, PvP basically means fighting a healer's mana bar. Consider a rogue fighting a shadow priest. The shadow priest simply has to cast Shadow World: Pain, an instant damage-over-time ability that lasts 24 seconds, and then stay alive. If necessary the shadow priest can drop shadowform and use heals: since heals have about twice the healing-per-second that a rogue should have sustained damage-per-second, this should be effective as long as the priest chooses a good timing (otherwise the rogue can burst damage while shadowform is off, with its 15% physical damage reduction). The larger the priest's mana pool, the greater her chance to win, if the rogue's dps and the priest's dps and healing all scale the same.

All Of That Explanation Just To Say This

Mana-increasing stats should have a scaling cost to avoid mana pool inflation from multiple raid tiers. If half of the stat budget of an epic item is devoted to increasing mana, it should increase mana by the same amount whether it's from the first raid tier or the last.

There are other stats that should probably work the same way: basically, anything that was converted into a "combat rating" in World of Warcraft's The Burning Crusade expansion in 2007, like critical hit, hit, and parry.

WoW's stats weren't designed to work this way. The basic concept was just to have the five main character stats: the only increase to spell dps was a tiny boost to spell crit chance from intellect. Spells were supposed to increase in damage from higher ranks (with a slight boost from leveling as well), not from gear. And so it's easy to compare the quality of items even without an addon that shows item level: if one ring gives 5 stamina and 5 intellect, it's a higher quality than a ring with 5 stamina and 3 strength.

If instead we make it so 10 intellect on a lvl 60 ring is 50% of the budget no matter the item's quality, we have the unintuitive result that a ring with 20 stamina and 10 intellect could be better (higher item level) than a ring with 30 stamina and 5 intellect.

Basically, the reason this is confusing is because of UI: the five base stats including intellect in Classic WoW are displayed on items with white text and grouped together, instead of being listed below in green text and with a more verbose description. So we are led to think that these five base stats are comparable in terms of how they work with the item budget.

Why Not Increase the Mana Cost of Spells?

Mana pool size is an example of multiplicative, non-linear scaling. Total output = throughput (output per second) times duration. If throughput stats also increase mana cost of spells, then duration remains the same and the number of health pools a character can heal in better gear remains the same.

This is bad because of spells that are not affected by throughput stats. In WoW, this is spells like Blink or Dispel Magic. Why should the mana cost of Dispel Magic increase because of equipping an item that increases healing done by spells by 50?

We can consider an alternative: make spells that don't damage or heal scale with total mana (increased by gear), not base mana. But we get weird edge cases: talented Rank 1 Frostbolt in Classic WoW costs 25 mana, takes 1 sec to cast, applies a 50% snare for 8 sec with a 36 yard range, and can proc other effects like Frostbite. It is comparable to the TBC mage ability Slow, also with a 50% snare, but with this system we would have Slow's mana cost increase only with total mana, while Rank 1 Frostbolt's (trivial) mana cost would increase only with damage throughput stats, even though it's used for the snare.

We also have the complication of mana (and health) regeneration: if mana regeneration is independent of total mana, and the mana cost of non-damaging abilities scaled with total mana, then higher total mana would actually be bad to have in many situations.

Should Mana Regen Work the Same Way as Mana Pool Stats?

I'm honestly not sure. PvP is probably the easiest thing to break, and mana regen (spirit) is viewed as bottom-tier importance in Classic WoW PvP except in healer-vs-healer duels, where a stat point of mana regen arguably provides better returns with talents than a stat point of damage: example, 1 agility for a rogue might increase dps by 1 because of talents and abilities, but 1 stat point in 'spell damage' (about 1.1 damage) for a Holy priest specced into healing might only increase dps by 0.4, while 1 spirit could potentially give enough mana from non-casting regen for more than 1 healing per second.

But there is the argument that if mana regen becomes too easy to get, then mana pool stats become a needless expense in the item budget for high-quality items.

Should Mana-Related Stats Scale With Character Level?

Option 1: a lvl 30 ring gives 5 intellect (+ other stats), which increase mana by 75 (base mana 750): a lvl 60 ring gives 10 intellect (+ other stats), which increases mana by 150 (base mana 1500). Option 2: both rings increase mana by 10% (or by an amount depending on class that works out to the same total increase for every class).

With Option 2, a lvl 60 ring with really good other stats (obtained from raiding) would still be good at lvl 65. With Option 1, a lvl 65 green quest ring would give more intellect and might be better, which is bad, but a lvl 60 ring with poorly-allocated 'other stats' is more likely to be better than a lvl 50 ring with optimal 'other stats'. Option 1 is like WoW TBC's combat ratings, but mana is not inherently percent-based like "critical hit chance" so no extra abstraction layer is added and it doesn't feel bad or confusing. I would pick Option 1.

Stat Diversity and Class Preferences

Classic WoW sort of had the intention that different classes would want different stats. Priests want spirit; mages want intellect; warlocks want stamina. In practice, everyone wanted stamina for PvP, but, that was the vision. 'Retail' or modern WoW has gear stats that change based on the class that's wearing it or something; basically, stats don't matter. Classic WoW was based on the tabletop role-playing game (like Dungeons and Dragons) concept of you roll a character, literally rolling dice to determine their stats, and then pick a class for that character based on which attributes of the character were the highest, with each class having minimum stat requirements and preferring certain stats: clerics got bonus spell casts from Wisdom, while wizards got bonus spell casts from Intelligence. (Heroes in Warcraft 3, from which WoW's art style is derived, also have a preference for one of three stats, with their per-hit damage scaling based on that stat.)

This can have a nice effect on the game: an item drops and it has stats that a certain class would want.

I think it could make sense to combine mana pool stats and mana regen stats, if both should have a scaling cost in an item's budget. In WoW terms, this is like moving mana regen from spirit to intellect, while spirit retains out-of-combat health regen. The issue then is how it affects which classes want spirit or intellect, and whether further changes are needed. To put it another way: how to get players to care about more stats than just damage and HP?

THIS POST IS TOO LONG

Let's look at how well it worked in Classic WoW. "Warlocks want stamina": at low levels, intellect is still good for warlocks: 1 int = 15 mana, 1 stam = 10 health, which can be converted to 10 mana with Lifetap. If a warlock knows he will go out of mana in a fight, int is better than stam.

At high levels, the warlock will have +spell damage gear. This increases the mana efficiency of Drain Life, which means mana is less of an issue. Friendly healers will also have more efficient heals from +healing gear, making them able to easily heal the warlock after he life taps: so, stamina is definitely better than intellect.

But as mentioned, at high levels every class wants stamina in PvP. Priests don't value spirit much in PvP, even in a healing role (which has more excuses not to cast, benefiting more from spirit). So the main differences between primary stats end up overriding soft classes preferences. It's different for strength and agility: warriors get attack power only from strength, while rogues get it only from agility, so their stat preferences remain distinct: they are both choosing damage, but getting it from a different stat.

Spell schools offer an example of how to differentiate items, if done properly. A relatively bad item for WoW: 50% of budget is stamina, 50% is +frost damage. A mage can spam frostbolt, but they often have a reason to use other spells like Fire Blast, Arcane Explosion, or (with clearcasting talent) Arcane Missiles. A better item might have 50% of budget as stamina, 40% as intellect, and 10% as +frost damage: with a root-mean-square type system, 10% budget ends up as more than 20% of the amount of stamina. This much frost damage provides a small, nearly-free bonus to mages using frost spells: it doesn't help a mage when using other spell schools, and doesn't help priests or warlocks at all, but they can still use the item when they couldn't if the item was class-restricted to mages only.

Example calculation, using RMS-type exponent 2, so we use 1/2 = 0.5 here when starting from item budget share: (0.5 budget)^0.5 = 0.71 stamina, (0.4 budget)^0.5 = 0.63 intellect, (0.1 budget)^0.5 = 0.32 frost damage. We go from 0.71 frost damage to 0.32 in exchange for 0.63 intellect. (Frost damage would then get a multiplier of about 1.25 since it's cheaper than a primary stat.) Alternatively, we could go from 0.71 frost damage to 0.63 in exchange for 0.32 intellect: the lowest stat is nearly free.

In general, we can imagine stats that would be the best possible stat for no class, but wanted more by certain classes than others. These stats could be used in small amounts (therefore cheaply) to make an item attractive to certain classes. The problem is when an item designer doesn't understand this and gives an item only these non-optimal stats (like a 'mage wand' with only +frost damage). Another consideration is the opportunity cost of non-ideal stats: an item that drops in Deadmines with +8 intellect and +4 agility (calculated total budget: 8^2 + 4^2 = 80) might actually be better for a priest than one with +9 intellect (calculated total budget: 9^2 = 81), but it is definitely worse than one with +8 intellect and +4 spirit.

So the conceptual challenge is making a stat that defines a class, like "intellect", but the class not wanting to stack that stat above any other stat: because a lot of the time players just want to stack damage, and making every class-defining stat just being "the damage stat" is kind of boring. The solution might be in well-designed restrictions on stat combinations when creating items: perhaps the +8 int, +4 agility item can be attractive because int + spirit items are not allowed, or some other less extreme rule.

(Random observation: agility would be better for casters in WoW if caster melee dps was competitive with wand dps, since agility gives dodge but not while casting or wanding.)

How to Make Items That Scale Hybrid Classes

The key is to combine the right types of stats in the item budget. For example, combine all 'physical damage' stats that can potentially stack with each other using a low root-mean-square(RMS)-type exponent, close to 1. This could be strength, critical hit chance, hit chance, and attack power: if two stats, like strength and agility, mostly substitute for each other (not sure if feral druids in Classic WoW get AP from agility), they might be combined first with a higher exponent. Separately, combine magic damage stats with a low exponent. Then combine the result with physical damage with a high exponent.

Healing stats can be a little complicated, if magic damage also affects healing to a lesser degree (or, like in WoW TBC, there is a +healing and also 1/3 +damage stat) but I'm sure you get the picture. It's combined with physical and magic damage in the same step, giving a combined 'throughput' part of the budget. This then gets combined with other parts of the budget, like 'defensive stats', stamina, and the scaling mana pool stat etc., with appropriate RMS-type exponents.

Examples without real math: in a simple system, an item might have 10 strength, 10 critical strike rating, 10 hit rating, 10 attack power, 10 armor penetration (or some other dps stat), and 10 stamina. Five dps stats and one health stat, each taking up 1/6 of the item budget: we calculate a simple budget of 10^2*6 = 600. If we remove stamina, we estimate the other stats go to 11 to get about the same budget (605). If we triple stamina, we get 900: overbudget; if we double stamina, we get 20 stamina (400 budget) + 6.3 (40 budget) for each of the other stats, total of 31.6 dps stats, down from 50 dps stats.

In the improved system, we combine all dps stats together. Let's just say we use exponent 1, aka we just add them. If we could actually afford this item, it would be 50 dps stats = 2500 budget + 100 stamina budget = 2600. For the same budget, we could triple the stamina and get 30 stamina (900 budget) + 41.2 dps stats (1700 budget). This is using the same exponent of 2 to combine the groups as before: if we use a higher exponent, we further discourage putting all the stats into a single category like dps, because other stats like health or mana have a bigger discount in small amounts.

We can see just from this simple example how if there are a lot of types of dps stats, stamina gets crowded out of the budget and we get damage that scales all out of proportion from health: bursty PvP. Altering the budget calculation allows us to get a lot more stamina in the budget for only a small cost in dps stats, meaning the item with extra stamina would be attractive to players.

The key is that an item with little or no stamina, or little or no mana or defensive stats, could still be available. A player who wants to optimize for some aspect of the game could still use that item: but it would provide a much lower advantage than in the simpler item budget system. Rather than 'max dps at the expensive of survivability' being optimal for 90% of raiding, it might be optimal for only 5%, combined with other changes like to threat.

This is how you make gear that gives players the option of 100% scaling for a single category of abilities, or 90% scaling for multiple categories like physical damage, magic damage, and healing. If you don't have a system like this then either players must split stats and get something like 50% scaling in two categories of abilities or 33% in three categories, or you just don't allow gear-based specialization and all abilities scale 100% with all gear, which can be boring.


___

Do you like gear specialization and if so, how much specialization?

So I wrote a really long essay (20k characters, 3700 words) but the only thing that was really new in it was that "the cost of mana-increasing stats in an item's budget should scale with higher item qualities from raiding, so that total mana does not increase". I have condensed that very long essay into the following poll.

Suppose you have a character who can specialize into different things or can try to be proficient at many things. Based on your class, your character can choose between one and three of "physical damage", "magic damage", "healing", "tanking", "magic damage of school A", "magic damage of school B", "magic damage of school C". How well would you want your character and all other characters to scale in each of these categories with specific pieces of gear (i.e. using the same items when swapping talents)?

TO RESTATE THE QUESTION: Should you be able to sacrifice the effectiveness of some of your abilities to boost certain other abilities by wearing specialized pieces of gear, and if so how much of a boost should this give? But see original question: how well should your character scale?

1) 100% scaling all the time in up to three categories
2) 100% scaling in dynamic category based on current talent spec, very bad in others
3) 100% in one static category, very bad in others, OR 35% in three with hybrid stats
4) 100% in one static category, very bad in others, OR 60% in three with hybrid stats
5) 100% in one static category, very bad in others, OR 90% in three with hybrid stats