User:Raine Valen/Mass Balance/Artificial Intelligence

Index

Warning: This page does not follow the proper format for pages of its kind. Do not base future formatting on the current formatting of this page.

Proposed:

Raine's Proposal:

• AI-controlled characters will be more likely to use anti-melee or anti-caster skills on foes that more frequently melee or cast spells. Additionally, AI-controlled characters will make initial assumptions based on their foes' primary and secondary professions, rather than their equipped weapon.

Getting Empathy and Soothing Images put on me when I monk because I'm wielding a spear is just silly. By the same token, not getting Backfired when I monk because I'm wielding a spear is just silly.

• AI-controlled characters will be less likely to melee or use offensive skills when they are not causing a net decrease in Health on their current target. In the event that there is no net decrease in health on any available targets, AI-controlled characters may stop attacking altogether.

If they're not accomplishing anything on their target, they should stop attacking. This is common sense: if you're just getting hurt by attacking the 600, stop attacking the 600. If you're not hurting the Defy Pain tank, stop attacking the Defy Pain tank. This would do a lot to stop many forms of farming, as well.

• AI-controlled characters will be more likely to melee or use offensive skills on targets on which they are causing a greater net decrease in Health.

The converse to the above, also common sense: if you're being very effective, do more of what you're doing. This is basic.

• AI-controlled offensive characters will be more likely to change their target to that of an ally if that ally is causing a greater decrease in Health on their target. Additionally, AI-controlled characters will be more likely to select an ally's target as that target's Health gets lower.

Again, this is one of the things that should just be obvious. If your teammate is doing well, they say, "Hey, guys! This target is really easy to kill!" and they get support. AI doesn't do this, just like bad players don't do this. They're both fairly terrible.

• AI-controlled characters will have their awareness limited to their own actions and actions of nearby creatures and their target's actions and actions of nearby creatures.

That's right. AI won't be able to dshot any target on the map on recharge. No one knows what's going on at all times everywhere. That's just degenerate. They will be limited to awareness of things going on near them, and near their target, as any player is.

• AI-controlled characters will have their awareness of their target's status limited to effects placed on that target while they were being targeted. However, they will be aware of whether or not their target is under the effects of a Hex, Enchantment, Weapon Spell, or Item Spell, regardless of when that status was placed on the target. Furthermore, they will know which conditions are on a target.

How would you know that someone has Life Bond on them unless you saw it cast on them? You wouldn't, unless you had omniscient AI like Gwen. You would know that they were enchanted. You might even know that it was a Monk enchantment, if you looked closely. Some enchantments even have specific displays (Protective Spirit, Shield of Absorption, and Life Bond aren't easy to miss when you hit something under the effects of one of those enchantments. It's not hard to think "Holy Veil" when your hexes take twice as long to cast. In fact, some enchantments are plainly visible: Holy Veil, for example, is a little blue ball with a faint blue trail that orbits the person on whom it's being maintained.), but a lot of them simply don't. AI shouldn't be aware of things that they'd have no way to be aware of.

• AI-controlled characters will initiate interrupts 0.3 to 0.6 seconds after they become aware of a target using a skill. This number will be assigned to each individual AI-controlled character upon instance creation. Also, AI-controlled characters will not attempt to interrupt skills which they have repeatedly missed interrupts on.

Godly interrupt reflexes were one of the things that made dealing with AI grieving. Not even the best players can Cry HCT Patients (not without great prediction skill, at any rate). Getting Power Leached on Spirit Bond right off the bat is just stupid, and nothing should be able to do that.

• AI-controlled characters will remember any skill used more than once on or by a creature that they are targeting. AI-controlled creatures will remember the user instantly if the target was the user, and after 2-5 (randomly) uses if the effect was placed on the target by an outside source.

Yes, people remember skills. If some monk is putting SoD on all your targets, you learn pretty quickly to target swap when you see SoD go up. Unless you're controlled by GW AI, in which case you continue to train them indefinitely. When something significant happens, AI should have some sort of flag go up.

• AI-controlled characters will be more likely to act based on remembered skills that are more significant, and this significance will be based on the amount of damage, the amount of healing, or the degree of shutdown that a specific skill causes.

This means that they can aim shutdown at important things. That's right! WoH is a bigger problem than Orison, so AI should work harder to PD WoH than Orison. Again, common sense that AI simply lacks.