Talent scaling consistency

[lukep]

There are a few talents that scale quite differently than most others. Most use CombatTalentSpellDamage (or similar), but some use a formula like:

Code: Select all

local dam = 30 + self:getCun() * 0.8 * self:getTalentLevel(t)

This is bad because it circumvents the scaling that is applied to almost all other sources of damage, and the inconsistency between them can lead players to the conclusion that a talent isn't worth investing in, as the starting values are too low, despite the fact that the high end could be very worthwhile.

The main places I see this are in Traps, Glyphs, Slow Motion, and Aim, though there are other places as well.

Do others agree that this should be changed? I may do a quick run through of many of the talents' code, and it would not add too much effort to find and change them.

[SageAcrin]

I'm not sure Aim/Rapid Shot balance should be mucked with-they are actually surprisingly balanced against each other on close examination, and any change would likely make a less balanced choice there.

Traps and Glyphs could use overall overhauls as categories IMO, as they both have some wonky scaling issues and both have weird niches that I don't feel are really that well done. Traps conflicts with Stealth, one of the more unique draws for Rogues, while Glyphs is area control on a durable survival-oriented mage, which is somewhat overkill. (Neither's really bad, but they have clashes with playstyle due to this.)

To be honest, Freeze strikes me as the worst case in the game of it. Not only does it scale insanely on the PC end for damage, it proceeds to scale even harder upwards past PC-obtainable levels. The thing that drives Elandar from general threat to ridiculous one right now is that weird scaling Freeze has, which allows him to just bust out 1k damage whenever he feels like. (I get the logic behind it, since the cooldown goes *up* with levels. But that doesn't make it any less dangerous when enemies toss it out.)

[Zonk]

Speaking of inconsistencies and misleading players...most talents have diminishing returns, but the two weapon fighting ones(the passives that determine your offhand damage)work the opposite.
Meaning that going from 4 to 5 can be a quite bigger bonus than 1->2, something that might not be obvious at first. It's also likely to make a huge difference for NPCS who have these talents at a very high level.

[Hachem_Muche]

The main places I see this are in Traps, Glyphs, Slow Motion, and Aim, though there are other places as well.

Do others agree that this should be changed? I may do a quick run through of many of the talents' code, and it would not add too much effort to find and change them.

I whole heartedly agree and have already done some of it. I have just about finished rescaling nearly every talent in the game as part of my Infinite500 addon. (Over 450 talents updated in the svn version). While my priority has been to keep gameplay balanced at high levels -to character level 250 or higher- to allow excursion to at least dungeon level 500 in the I.D., part of this effort has also been about making sure that damage scales consistently across talents. I have made a number of changes to the talent formulas to achieve this, while maintaining game balance for characters near the end of the Maj"Eyal campaign. (Just to be clear, this addon is completely compatible with normal character levels.)

I have focused mainly on high talent levels, but all of my updated formulas employ continuous functions that incidentally smooth out the kinks in some of the existing formulas. (The offhand damage formulas are already corrected, for example.) Since the code changes are already in place and are cut-and-pastable with most of the existing lua code, adjusting the formulas is very simple from this point. I'd be glad to look at any suggestions you have for things are a bit wonky at low level and make adjustments.

As an example, I changed the formula for the Freeze spell from:

self:combatTalentSpellDamage(t, 12, 180) * (5 + self:getTalentLevelRaw(t)) / 5

to:

self:combatTalentSpellDamage(t, 10, 350)

These two formulas give (by design) the same damage for a character with talent level 5 and 100 magic stat, but the first one scales super-linearly with character level while the second does not.

I am currently working on making sure traps, stealth, and attack and defense scale consistently. So I could use some advice on what needs fixing with them as well as the archery talents.

[tiger_eye]

Do others agree that this should be changed?

Not really. "Broken" talents that are unfair and deadly at high levels should be fixed, and there have been efforts to do so (thanks lukep; were high level spider summons (iirc) ever fixed?). But, changing a talent from scaling linearly to slightly sub-linearly isn't really a big deal for most of the mentioned talents.

This is bad because it circumvents the scaling that is applied to almost all other sources of damage, and the inconsistency between them can lead players to the conclusion that a talent isn't worth investing in, as the starting values are too low, despite the fact that the high end could be very worthwhile.

Uh, not really. If the starting values are low for one method then they will be low for the other method too.

Using a linear relationship (as is done for Traps) results in then being stronger at high levels then they otherwise would be using a CombatTalentDamage function.

Traps and Glyphs could use overall overhauls as categories IMO, as they both have some wonky scaling issues and both have weird niches that I don't feel are really that well done.

What's wrong with the scaling of Traps? They're great early, and remain awesome late. Perhaps you and others would enjoy a "Traps no break stealth" addon?

Speaking of inconsistencies and misleading players...most talents have diminishing returns, but the two weapon fighting ones(the passives that determine your offhand damage)work the opposite.
Meaning that going from 4 to 5 can be a quite bigger bonus than 1->2, something that might not be obvious at first. It's also likely to make a huge difference for NPCS who have these talents at a very high level.

Offhand damage is capped, so it does not make a huge difference for NPCs who have it at very high level. Same for "Slow Motion".

The main places I see this are in [...] Slow Motion

Slow Motion is unlike any other talent in the game. It does not do damage, so why should it scale as a typical damage talent?

So, I'm ambivalent regarding the suggested changes, and don't think they need done or would affect the game much if they were done. However, this does reveal a long-standing issue: how to communicate the scaling of a talent to the player? If talents were more or less done equivalently, then some sort of scaling coefficient or what-have-you could be displayed in the talent levelup description.

Anyway, thanks all for taking a detailed look at the internals. The game wouldn't be what it is now if not for the community support and involvement !

[lukep]

Uh, not really. If the starting values are low for one method then they will be low for the other method too.

Using a linear relationship (as is done for Traps) results in then being stronger at high levels then they otherwise would be using a CombatTalentDamage function.

My worry is that people see a talent (A) that starts at 10 damage (for example), and will completely ignore it in favor of a different talent (B) that deals 20 damage to start, despite the fact that the scaling is different. If talent "A" scales like a trap, it will go from 10 to 50 damage at level 5.0, while talent "B" (scaling like most of the rest) will only scale from 20 to 40.

Slow Motion is unlike any other talent in the game. It does not do damage, so why should it scale as a typical damage talent?

Lots of sustained non-damage talents use CombatTalentFooDamage as their scaling formula. For example, Icy Skin, Fungal Growth, Anscestral Life, Chants, Hymns, Energy Decomposition, Entropic Field, Feather Wind, and Stone Skin all do. Basically all wild gifts and spells (except Vim? use it, but Psi talents use a different formula (combatTalentIntervalDamage IIRC, which scales linearly with talent level) and stamina sustains are often (base + [talent * stat * multiplier]). It seems to work for some talents, but not others.

So, I'm ambivalent regarding the suggested changes, and don't think they need done or would affect the game much if they were done. However, this does reveal a long-standing issue: how to communicate the scaling of a talent to the player? If talents were more or less done equivalently, then some sort of scaling coefficient or what-have-you could be displayed in the talent levelup description.

For most talents, this isn't an issue. From level 1 to 2 they gain 33% power, 2->3 is +19%, 3->4 is +14%, and 4->5 is +11%, for a total of double power at talent level 5.0 vs 1.0. It's just how CombatTalentFooDamage works, and a large majority of (non-melee/archery) talents use it, especially for damage talents. The players may not know the precise numbers, but it doesn't take long to get a feel for the scaling. When it doesn't fit that mold, things can get unpredictable for the player.

[tiger_eye]

For most talents, this isn't an issue. From level 1 to 2 they gain 33% power, 2->3 is +19%, 3->4 is +14%, and 4->5 is +11%, for a total of double power at talent level 5.0 vs 1.0. It's just how CombatTalentFooDamage works

Right. The difference between the two approaches is how talents scale with talent level. The thing is, though, it's easy to see how things scale with talent level, because one can just respec' talent points into (and out of) them to see what happens. Both methods, however, scale linearly with the associated power (i.e., Cunning, spellpower, etc.). This is what's difficult to see, and using one method or the other doesn't make a difference. In other words, your proposals don't clarify the primary issue.

So, here's a new idea: in the levelup screen, have the option to see talents w.r.t. the current character, or a reference "Mid level" and "High level" character. The mid level character could have stats of 50 (or whatever is appropriate), and the high level character could have stats of 100 (or whatever is appropriate). I tend to code-dive to see how suitable a talent will be for me as my character develops. Most players don't--and shouldn't--do this, and instead learn how talents scale by playing the game and leveling up their characters. I suppose this exploration is part of the fun for some players, but seeing what a talent may be like for a mid-level and high level character could be beneficial too.

[Hachem_Muche]

So, here's a new idea: in the levelup screen, have the option to see talents w.r.t. the current character, or a reference "Mid level" and "High level" character. ...

I don't think this is needed. With all the rares and random bosses now in the game, you can get a pretty good feel for how effective talents are at higher level by inspecting these mobs or by experience with being on the receiving end of them. I much prefer this kind of discovery rather than too much hand-holding.

The far more serious issue is talents that just don't scale reasonably, like one-point wonders, etc....