Fixing high level Talents

[lukep]

With NPCs cheating on talent levels, it is very possible for them to get extremely high leveled skills, that breaks them in various ways. The ones that I think need fixing, listed by the reasons behind it are:

Should be short range, gets too long:
Frozen Ground, Nova, most Breath attacks, Flameshock, Blastwave, Quake, shattering shout

Roughly linear scaling (of at least part of it):
Dual Weapon Defence, Shield Wall, Sweep, Multishot, Momentum, any speed effect, most mindslayer skills, all summons, Martyrdom, Twilight Surge, Time Prison, Slash

Power approaches/crosses 0 (mostly cooldowns)
Darkest Light (neg. regen), Rampage tree (cooldown etc), Supercharge Golem (cooldown), Mortal Terror (lvl 10 = Daze at 0% damage), resistance effects (Juggernaut?), Rush (cooldown), Pity (range), Suppression

There are at least a few that I missed, and I didn't check if monsters could use all of these talents. One thing that would fix the majority of these would be to change self:getTalentLevelRaw to only return a value between 0 and 5. The rest would need to be changed individually.

EDIT: added Rush, removed swallow from third set.
EDIT 2: Added Pity and Suppression to third set

[bricks]

Rush still hasn't been addressed, and is a considerable problem. I think it's actually possible for an NPC to get Rush to a cooldown of 0.

[edge2054]

Swallow is no_npc.

Darkest Light I'll fix so it caps at 0. The rest of these I don't know how Darkgod feels on. (I personally think seeing big dragons breath across the whole screen is cool but the radius should probably cap at 10 anyway since that's the LOS limit. )

[bricks]

Perhaps breath attack damage should fall off with distance, but get a boost for short-ranged use. It would make them more useful for one-on-one, give players a means of mitigating damage via positioning, and it logically follows that a spreading cone would lose effectiveness the more "spread out" it became. Other cone skills could get the same "shotgun effect" treatment; even ball effects could be stronger in the middle and taper off with increased radius. DoomRL has this effect and it makes for very interesting tactics.

[Silvermoon]

Perhaps breath attack damage should fall off with distance, but get a boost for short-ranged use. It would make them more useful for one-on-one, give players a means of mitigating damage via positioning, and it logically follows that a spreading cone would lose effectiveness the more "spread out" it became. Other cone skills could get the same "shotgun effect" treatment; even ball effects could be stronger in the middle and taper off with increased radius. DoomRL has this effect and it makes for very interesting tactics.

I like that idea, and it probably wouldn't even be too hard to implement, assuming you do so universally. Define a minimum and maximum effectiveness for any AoE talent, and an origin point. The further from the origin point, the less effective. For breath attacks, this is the end of the cone touching the player. For radial AoEs, it would be the center point. I'm not aware of any other AoE patterns in use. For abilities where you don't want them to be affected by the spread, just define the minimum and maximum identically.

[Gliktch]

For breath attacks, this is the end of the cone touching the player.

Some cone-area talents (was Flameshock one? I haven't played any of the relevant classes yet, I just remember it being mentioned elsewhere) actually do 'hit' the player though, so they can cause their positive or positive+negative effects on the player too - ah, it was "healing flames" or something actually. In those cases you probably don't want to player to be hardest hit.

Fully agree that AOE should scale with distance though - I believe Explosion Expert implements something like this, but I really can't parse its talent description Any hints on that one people?

[eronarn]

This kind of scaling is particularly bad on insane. For example, the dredge frenzy ability comes off of cooldown before expires - a dredge captain is invulnerable if it keeps triggering it (you have to stun it so it can't use it to be able to kill it). Talent levels beyond 5 should in general be subtle increases rather than scaling the same way 1-5 do.

The acid blood sustain is one that scales ridiculously, by the way.

[edge2054]

Frenzy doesn't grant any sort of immunity to death so it's not quite invulnerable. Very hard to kill on insane sure but not invulnerable. That said I'll double the cooldown, it's pretty low now. This will stop them from chaining the talent on normal difficulties. I'm not sure how Darkgod feels about 'balancing' insane though so I'll have to talk to him before I start putting duration caps on monster talents.

[lukep]

There are quite a few more talents that scale linearly, in addition to acid blood, Precise Strikes, Aim (no NPC), among others. I didn't have a consistent criteria throughout, so some got left behind.

Some cone-area talents (was Flameshock one? I haven't played any of the relevant classes yet, I just remember it being mentioned elsewhere) actually do 'hit' the player though, so they can cause their positive or positive+negative effects on the player too - ah, it was "healing flames" or something actually. In those cases you probably don't want to player to be hardest hit.

IIRC, they hit the tile the caster is on, but do not hit the caster.

[marvalis]

I can't help but wonder: why do npc's need to cheat on talent levels in the first place?

[Grey]

Yeah, why can't they fight fair? If they didn't get crazy talent levels we wouldn't have this problem.

[darkgod]

because I'd need to generate them with a full set of correctly matched equipment to their spec.

[Grey]

More loot, yay! :D

[Aquillion]

Hrm. Here's how I'd do it:

Define certain specs for each class, with minimum talent and stat levels.

For each spec, define a 'benchmark' -- typically some function of the damage dealt by certain key talents, while factoring in their HP, resistances, etc. So, say, average the damage of their three key skills and their HP with some sort of weight applied to it.

Generate a few sets of gear, and for equipment slot, choose whatever maximizes the benchmark. For weapons you'd have test both duel-wield and double-wield with the generated items, but that's not too hard to account for.

A bit complicated, but doable, and it'll generate characters that very closely resemble PCs (since that's how most PCs choose their gear!)

But that's not all. An added benefit of this is that you could then use a variant of this same benchmarking system to provide general advice to PCs who want to choose between different gear options.

[marvalis]

because I'd need to generate them with a full set of correctly matched equipment to their spec.

Just let them cheat on damage then?

I don't know if that would work out as intended, just an idea.