The Forums Are Now Closed!

The content will remain as a historical reference, thank you.

Enhanced AI Mod v0.26.35 Release 2/16/2011

By on November 27, 2010 11:12:28 PM from Demigod Forums Demigod Forums

Peppe

Join Date 03/2006
+49

Enhanced_AI_v0.26.35:

http://peppe-dg-mods.googlecode.com/files/Enhanced_AI_v0.26.35.zip

http://peppe-dg-mods.googlecode.com/files/Enhanced_AI_v0.26.35.exe

 

 

AI will work hard to get map control.

AI will actively work to keep portal counts even or in its favor. 

AI will save for critical upgrades like currency, so it can purchase it at or near warrank 3.

AI generals will consider what healer idols the rest of the team owns when making purchases.

AI generals will send one of their healing idols to an AI teammate on summoning (when 2 are summoned) and on teleport.

Locked Post 398 Replies +3
Search this post
Subscription Options


Reason for Karma (Optional)
Successfully updated karma reason!
November 30, 2010 1:34:58 PM from Demigod Forums Demigod Forums

 

Thanks for the playtests and the feedback.   I have only played QoT once or twice, so don't know much about her.  I'll take a look at her actions and post the logic.

 

First you are correct she has one build that i pulled from the forums somewhere.  If you would like to see other builds I can add them.  If multiple favor items are listed it randomly selects between the two.

        shield_spike = {
            'HQueenBrambleShield01',    # 1
            'HQueenGroundSpikes01',        # 2
            'SAVE',                      # 3
            'HQueenBrambleShield02','HQueenGroundSpikes02',    # 4
            'HQueenSpikeWave01',           # 5
            'SAVE',                      # 6
            'HQueenBrambleShield03','HQueenGroundSpikes03',    # 7
            'SAVE',                        # 8
            'SAVE2',                      # 9
            'HQueenBrambleShield04','HQueenGroundSpikes04','HQueenSpikeWave02',    # 10
            'HQueenShambler01',          # 11
            'HQueenShambler02',          # 12
            'HQueenConsumeShambler01',    # 13
            'SAVE',                       # 14
            'HQueenGoddessofThorns','HQueenSpikeWave03',  # 15
            'HQueenConsumeShambler02',    # 16
            'HQueenShambler03',          # 17
            'HQueenConsumeShambler03',    # 18
            'HQueenCompost01',            # 19
            'HQueenCompost02',          # 20
            FavorItems = {'AchievementMinionInvis','AchievementHealth'},
            ItemWeights = {
    
                    # Orb of Defiance
                    # Use: Become invulnerable for 5 seconds. Cannot move, attack or use abilities.
                    # +500 Health
                    # +500 Armor
                    Item_Consumable_150 = {
                        Priority = 30,
                    },
            },
        },

 

To go into packed form: She has to have > 30% health.  To go into unpacked I don't see any conditions.  Not sure what this means behaviour wise.  She may just randomly switch packed vs unpacked if she is greater than 30% health.

We can put some logic on that.  

So if there is no enemy hero threat and no tower threat she should be packed? 

Saving ground spikes and spike wave for attacks when heroes/towers are around?

Unpacked has a survival bonus attached to it (tells the AI to use unpacked to survive).  Is she faster in that mode or something?

 

She has two shield actions:

Shield herself if she is less than 75% health and an enemy demigod is around.

Shield friendly when enemy demigod is around and the ally is under 75% health.

 

Looking at the skill it may make sense to drop the health check as the shield doesn't wear off.  Maybe put a mana check and spam it on allies if we have >50% mana?  Saving the rest of our mana for ourselves?

Reason for Karma (Optional)
Successfully updated karma reason!
November 30, 2010 1:49:10 PM from Demigod Forums Demigod Forums

Ah yes, they do tend to ignore potions during a retreat.  Need to add a periodic loot potion check to the retreat action, definitely.  Cycling every 5 seconds at radius 15 should be enough to prevent the AI from going way out of its way to grab a potion, although weighting the decision by the potion's distance to the health stone compared to the AI's and also by potion strength would be good.

How exactly do you implement something like this?  Create a sensor for it, and have the health sensor trigger it?  Clearly the health sensor already does some potion management, it just tends to ignore them when it hits its retreat threshold, which is really when it needs them most.  Could it just be a matter of tweaking weights?

 

Re: QoT, that's backwards.  She has shield, mulch shamblers, and increased armor and hp/mana regen when packed.  There's no speed difference.  She should go into packed for survival, and try to cast shield/mulch.

Reason for Karma (Optional)
Successfully updated karma reason!
November 30, 2010 2:37:34 PM from Demigod Forums Demigod Forums

 

Ok i'll play with queens packing/unpacking.  Think the issue is packed was set to like 5 goal sets and unpacked was set to 'All' goal sets.  Sounds like that should be flipped or different sets defined for each. 

Hero attacking is in the Assassinate goal set, so if that is the only thing to use unpacked form for then we could keep her packed a lot of the time by just giving unpacked that one goal set. 

 

Edit: On shield usage i found every unit maintains an 'Absorption' value, so I am going to try adding the condition of using shield anyone without a shield or with a shield below 700.

 

The reason it doesn't pickup potions on retreat is its already in the 'Move to Health Strone' action.  It has a dynamic timeout equal to the walking time from the start point to the statue.  To get it to stop and pickup a health potion that action has to be killed when a potion gets in its radius.   Now if we kill the action it may go get the potion or just select retreat as its action again and be stuck cause a potion is still in its radius    Trying to create the setup to test against it, but working on queen at the moment

If you want to test in your local copy miriyaka you can try adding this to the move to health stone instant status function (pulled from pickup health potion action in MiscActions):

        local PotionSearchRange = 30  (30 is the range from misc actions)
        local potion, healVal = AIUtils.GetHighestValuePotion( unit:GetPosition(), PotionSearchRange, 'Health' )
        if potion then
            WARN('Abort Retreat Found Health Potion')
            return false
        end

Reason for Karma (Optional)
Successfully updated karma reason!
November 30, 2010 3:36:07 PM from Demigod Forums Demigod Forums

 

26f:

http://www.filefront.com/17586623/Skirmish_AI_v26f.exe

http://www.filefront.com/17586625/Skirmish_AI_v26f.zip

 

Queen updates.  Should be in the closed form most of the time now switching to attack...  Shield use more often (maybe too often).  Condition to use shield on self or allies is if there is a hero or tower threat in the area and the target does not have a shield above 500 points already.

AI in full retreat may stop to pickup potions.   <--- Needs testing.  I suicided a bunch in an area to get some pots out and then fought the AI until it retreated and it got a potion, but we may have battled near enough the potions for it to not be in a full retreat yet.

 

Enjoy.

Reason for Karma (Optional)
Successfully updated karma reason!
November 30, 2010 3:43:33 PM from Demigod Forums Demigod Forums

Nice.

Reason for Karma (Optional)
Successfully updated karma reason!
November 30, 2010 4:11:22 PM from Demigod Forums Demigod Forums

 

Added potion check to the moving loop as well.  Seems to abort the action 100% now and go get the potion.

 

26g:

http://www.filefront.com/17586763/Skirmish_AI_v26g.exe

http://www.filefront.com/17586764/Skirmish_AI_v26g.zip

Reason for Karma (Optional)
Successfully updated karma reason!
November 30, 2010 5:40:33 PM from Demigod Forums Demigod Forums

First off – thanks again for all of the effort on this mod.  I think it is a very nice value add to demigod for single player and should be quite useful for MP (well, players starting the game with AI at this point).  I think you’ve accomplished quite a bit towards making the AI useful.  In MP terms, its skill isn’t great, but because it’s intelligence is increased, people can spend more time enjoying playing and less time focusing on getting upgrades and wondering what batshite crazy thing the Ai will do next (play a 3v3 without the mod and choose rook as your character.  Walk to the middle of the map on cata and take the flag.  Drop a tower and watch at least one of your teammates tp in for no explicable reason about a minute into the game).

Ok – so, hopefully we are about to the point where we can do some major testing on more than a few settings.  There has been quite a bit of rhyme and reason for me choosing the settings I have been (normal with high towers).  I choose normal so that the ai would have to learn how to manage its money intelligently for upgrades and high towers so the AI would HOPEFULLY learn not to run past towers on a regular basis to go after flags.  The assumption is that money management will be great and much better on a hard/nightmare AI if the logic is sound on normal (certainly seems like it is) and that hopefully the AI will be smart enough to handle towers if they are on a normal setting.  So, I’ll start testing whatever the most current release is tonight on the various maps and keep taking notes.  If I don’t come across anything that’s a big deal (read: a problem and not just an enhancement request), I’ll keep on testing and hopefully we can get to a release candidate.

I’m thinking a name change for this mod is in order to make it a little more clear what this mod does, especially if we get it included in the game.  I’m thinking something simple:  Enhanced AI (we really don’t need to make this change until we have a final release version).  I’m up for other suggestions, of course, but I think AI Skirmish probably won’t make a lot of sense to folks that check their modding folders.  

I think you asked about this sometime ago, Peppe, and I’m not sure if I ever answered you, but here goes:  I’m looking to include your updated AI as a mod and not actually hard coded into the game.  I’m guessing there are more than a few changes with what you’ve coded and if we took a shot at hard coding it into the game, I’m guessing the QA work would be through the roof and we probably wouldn’t see implementation.  So, the goal is for inclusion as a mod.  I’m not sure if they could make it so the mod would be enabled by default or not, but at least it would be there for everyone.  I’d also like to see about getting your lobby change hard coded into the mod (where you can change the AI’s default levels and fill factor, but I’m not sure what exactly is involved on that coding side of thing – is it something pretty simple?).  

And last – there is still one thing I would really like to see incorporated here that we haven’t really discussed yet.  Can we get this mod  working well if a player drops from an ALL MP game and the AI has to take over?  Would the AI still save for upgrades?  Would its build go nuts or be garbage?  Etc.  That would be an excellent feature to add, but I’m not sure how that would pan out.  I can do some testing, of course, but I don’t know if we are already assuming this would run poorly, do ok, etc.  

Reason for Karma (Optional)
Successfully updated karma reason!
November 30, 2010 6:07:56 PM from Demigod Forums Demigod Forums

I think you asked about this sometime ago, Peppe, and I’m not sure if I ever answered you, but here goes: I’m looking to include your updated AI as a mod and not actually hard coded into the game. I’m guessing there are more than a few changes with what you’ve coded and if we took a shot at hard coding it into the game, I’m guessing the QA work would be through the roof and we probably wouldn’t see implementation. So, the goal is for inclusion as a mod. I’m not sure if they could make it so the mod would be enabled by default or not, but at least it would be there for everyone. I’d also like to see about getting your lobby change hard coded into the mod (where you can change the AI’s default levels and fill factor, but I’m not sure what exactly is involved on that coding side of thing – is it something pretty simple?).

The Lobby/UI mods I did with auto-fill and difficulty level are done in 2 files if i recall, 3 tops.  There is nothing in the game to load these as modded files.  They would have to be included as a patch.   Looks like they had plans to allow total conversion mods that could affect the menus/lobby but never implemented it.  You might be able to get them to QA and test the lobby changes with fewer issues than the whole AI mod.  As the options drop downs are super simple and the only code change is in the auto-fill button click function:

it reads the options for game size and difficulty

then uses a copy of the tournament mode AI selection to select unique demigods for that size game/difficulty.

It opens slots as needed and if the button is clicked again it clears all non human slots and fills them again.

It would be nice if they dropped their mod protection on tournament mode.  Seems to be hard coded in the exe to block gameplay mods in tournament mode.   I tried everything I can think of, but no matter what when the game sends the singleplayer launch command to a function in the exe even if gameplay mods are in the list to be loaded and show up in the game launch log as active they never get hooked in tournament mode.

 

And last – there is still one thing I would really like to see incorporated here that we haven’t really discussed yet. Can we get this mod working well if a player drops from an ALL MP game and the AI has to take over? Would the AI still save for upgrades? Would its build go nuts or be garbage? Etc. That would be an excellent feature to add, but I’m not sure how that would pan out. I can do some testing, of course, but I don’t know if we are already assuming this would run poorly, do ok, etc.

Mod should work fine if a player drops.  AI will focus on flags/map control and should also take over upgrades.  If the player had weird items it may sell/replace items to get its core set (scalemail, scale helm, banded, unbreakable, planor bc, nimoth, vlemish).  It won't have a skill plan, but there is a 'weighted' random skill plan for every class that it will fail over to.  We can adjust the weights if needed, but basically it selects skills based on all the previously selected skills.  Skill wise it should continue somewhat well from where the player left off.

Reason for Karma (Optional)
Successfully updated karma reason!
November 30, 2010 6:24:20 PM from Demigod Forums Demigod Forums

 

Someone mentioned earlier the AI may be too suicidal still on towers. 

The primary action that takes them into towers is the kill demigod actions.  There isn't an abort mid action, so it will attack for at least 3 seconds and re-evaluate.   Right now it is set to do something different when the attacker is in range of 2 or more enemy towers.  So on cataract you should see them pursue past the corner tower, but turn around at the range of the gold towers or thereabouts.  

We can continue to use that or make a more advanced formula.  Like if its current health is greater than 3000 it can go past one tower.  if its health is greater than 5000 it can risk 2.  Any more towers and it should abort the attack.  We can also add an abort mid action to cancel if it is running into too many towers.

 

Tough to balance if we want it to pursue at high risk for a kill or abort when it would be risky.

Reason for Karma (Optional)
Successfully updated karma reason!
November 30, 2010 6:42:29 PM from Demigod Forums Demigod Forums

2 human + 1 ai v 1 human + 2 ai test comin up.

Reason for Karma (Optional)
Successfully updated karma reason!
November 30, 2010 7:40:14 PM from Demigod Forums Demigod Forums

AI skirmish (mod version) .26g/uberfix 1.03 (last version published by miriyaka)

Settings:  cata/High Towers - all other settings on default
AI settings:  normal
3v2

rook (human)/ub/senda v ub (human)/sed (human)

Observed results: 

ub stopped one time when i was on a flag and did not engage (ub fell back to the tower behind the flag and just waited)... ub was at about 85% hp

sedna should not use cloak ever (cloak logic is bad - hops in whenever she feels like it to attack and then lacks any logical strat)

senda heal/silence build... doesn't seem to heal until late... I think I should revisit the current ability builds.  I DON'T THINK the ai should use cloak of night at all... it doesn't use it intelligently - would be much better off with a different favor.

ai tried to take gold with tower still up in front... retreated, but should have focused on the tower first (eg dg pushes past a high tower to take a flag still =  problem.  If a tower is nearby, do NOT go for a flag (unless late game > 8 = go for portal). 

Impressions:  I won the overall game.  I won because my ai, though it died a lot, bought the vast majority of the things I needed.  I think I chipped in and bought angels and giants to make a final push faster, but the ai covered the basics well enough for me.  AND, and its a big and, after I died pushing a portal on 1 side, drawing 2 ports, the ub on my team took a shot at their other portal and capped it... in short, late game, and soccer had acclimation, they had to spend way too much time covering way too many locations.

 

Next test - 3v3 - they get an ai.

 

Reason for Karma (Optional)
Successfully updated karma reason!
November 30, 2010 8:21:09 PM from Demigod Forums Demigod Forums

AI skirmish (mod version) .26g/uberfix 1.03 (last version published by miriyaka)

Settings:  cata/High Towers - all other settings on default
AI settings:  normal
3 (1 human+2 ai) v3(2 human + 1 ai)

rook (human)/ub/senda v reg (human)/oak (human)/qot

Observed results:

ub stayed back at full hp behind a tower while i was chasing a dg (eg too cautious for no clear reason)

AI SHOULD NEVER EVER EVER EVER EVER GET CLOAK OF NIGHT - PLEASE CHANGE ALL OF THE BUILDS TO BLOOD until we get a better replacement.

I was routed in this game.  Dark choose a reg and kited the ai a lot.  The ai needs to learn "the point of no return."  This is VERY important, but an advanced tactic.  If there is no guaranteed hope of escape, a good player will turn around and fight it out.  I can imagine something like this will be very tricky to program, but the bottom line is that if we don't get something in place, human players can just kite the ai (eg attack a little - then back up, then chase while the ai runs and wastes pots or dies).  The goal is to not die, yes, but to also NOT waste potions/sigils over and over.  There were several times where the ai on my team would approach the human players and run.

Oh - ai qot is great.  We need her to only go blood or anklet or anything that gives mana instead of cloak, though.  She doesn't use it well.  But she is spamming shields. The changes you made were great.  Now... can you make the AI snipe ok with reg?  If not, we should drop ALL reg snipe builds to make for more exciting games. 

Reason for Karma (Optional)
Successfully updated karma reason!
November 30, 2010 9:19:59 PM from Demigod Forums Demigod Forums

 

Achievement item switch is easy.   Glad new QoT is a good contributor. 

Do  you want it to breakoff the engagement if a tower hits it?   Or to avoid giving up on a kill allow one tower if we are higher health than the target?   But at 2 or more towers break off attack.

I can also count enemy and ally heroes near it.  Can kill the attack when its not favorable?

Reason for Karma (Optional)
Successfully updated karma reason!
November 30, 2010 10:22:17 PM from Demigod Forums Demigod Forums

AI skirmish (mod version) .26g/uberfix 1.03 (last version published by miriyaka)

Settings:  cata/High Towers - all other settings on default
AI settings:  normal
3v3

qot(ai)/oak(human)/tb(human) v sed (human)/rook(human)/qot

Observed results:

erb with cloak = retard.

occ  had ball lightning early + cloak = retard

... 1 more game

Reason for Karma (Optional)
Successfully updated karma reason!
November 30, 2010 10:22:33 PM from Demigod Forums Demigod Forums

i really need to tweak the ability builds/item builds

Reason for Karma (Optional)
Successfully updated karma reason!
November 30, 2010 10:52:19 PM from Demigod Forums Demigod Forums

Removed the cloak from all generals on my local version.

 

If you want to tweak skill builds directly in the files you can or post skill plans and i'll add/update builds in the file.

Reason for Karma (Optional)
Successfully updated karma reason!
November 30, 2010 11:04:59 PM from Demigod Forums Demigod Forums

We can continue to use that or make a more advanced formula.  Like if its current health is greater than 3000 it can go past one tower.  if its health is greater than 5000 it can risk 2.  Any more towers and it should abort the attack.  We can also add an abort mid action to cancel if it is running into too many towers.

I'd make it 2k and 4k. 

Or make it 1.5k * N + target's minimum HP. That way it'll be aggressive against an almost dead target even if if the pursuer is also almost dead. 

Another idea would be to make it sigil-aware or shield/bite aware. If it has a sigil (and it isn't on cooldown), it is more aggressive. 

Reason for Karma (Optional)
Successfully updated karma reason!
November 30, 2010 11:23:38 PM from Demigod Forums Demigod Forums

Only boost pursuit weight with a sigil if the target's health is 'low enough', whatever that means exactly.  It shouldn't sigil to chase a target with 2k health into two towers, because while it might be able to get out before it dies, there's no way it could make that kill, and that's a waste of a sigil.

Some general damage/shield ability availability awareness might also be a nice addition to pursuit considerations, if we're going to bother doing it for sigils anyway.  However, we probably don't want these conditions (which run quite often) checking dozens and dozens of possibilities that require iterating tables and comparing damage amounts etc etc because it's not only harder on performance, but increases the unpredictability and complexity of the whole system.

Maybe I'm being too negative about this, and maybe we could get some nice results from a UB that knows its spit cooldown is coming due soon, and it has a sigil, and it can sigil + continue pursuit + spit for a kill and still get out alive, but especially the survival aspect of that seems like it would be a bit difficult to manage reliably from what little I know about these AI actions.

 

Really, I think this kind of stuff is probably getting into demigod/build-specific territory.  There is a way to add AI actions for individual demigods, yes?  And I imagine this could be jury-rigged to be aware of the DG's build and current skill levels, weighting certain actions (particularly pursuit into towers) based on available and near-available (sub-3s cooldown remaining) abilities.  But probably not something that should be done in general AI logic, keep that no more specific than sigils and potions.

Reason for Karma (Optional)
Successfully updated karma reason!
December 1, 2010 12:25:57 AM from Demigod Forums Demigod Forums

Feature request - I'll flesh some of this out later.

Can we please get an option to make it so that the ai CLOSEST to any user ping will run away?  This would help cover any issues with weaker ai if a human player is watching it.

Interrupts.  Can we please make ANY AI with any iterrupt johnny on the spot to interrupt any lock or tp attempt?  This can be huge (to stop ganks or keep flag control or save a teammate).  I'd love to see this happen much more often.

Side note - we played a few games on nightmare ai and I think we had the most fun here.  I think it will be SOOOOOOOOOOOOOOOOO much better if the ai CAN'T CHOOSE CLOAK OF NIGHT (unless you find a way to code it so that the cloak is used 100% of the time for escapes and the AI actually run's after said escape).

And last - I don't have access to the modding forum form work.  If you want my feedback M-F, you need to edit the OP and move it to the demigod forum (you can move it back later, but I can't view it from work as is).  Strange world, but it is what it is. 

Reason for Karma (Optional)
Successfully updated karma reason!
December 1, 2010 12:39:37 AM from Demigod Forums Demigod Forums

Check my last 4-5 post plz - edited a few of them for clarity.  If you move this thread to just the demigod forums, I will provide more feedback tomorrow during the day.  Thanks!

Reason for Karma (Optional)
Successfully updated karma reason!
December 1, 2010 1:35:29 AM from Demigod Forums Demigod Forums

Current skills that look for interrupt opportunities to use:

DA, ShadowSwap   -Interrupt using item/skill is the only trigger for this skill.  Probably needs another condition to use the skill.

UB, Grasp - Interrupt using item/skill and if the target is below 60% fire at will.

Oak, Penitence - Interrupt using item/skill, fire at will when Oak has greater than 50% or 1500 mana.

Rook, Boulder - Interrupt using item/skill and if the target is below 60% fire at will.

Sedna, Pounce - Interrupt using item/skill, fire at will when Sedna has greater than 50% or 1500 mana.

Erb, Charm - Interrupt using item/skill, or if 2 or enemy demigods are near Erebus.

 

Nothing setup for TB, QoT, Oculus, or Regulus. 

For the above to interrupt more often we would have to reduce/remove the other usage condition.

 

Cloak has 3 actions.  Flee, warp to squad target, warp to enemy hero.  I can disable the warp ins leaving just flee.   Or make the warp to hero  condition only when the target is at like 500 health or lower.

 

What are the issues with regulus snipe?  He has 3 default actions for using the skill.  Kill shot structures, kill shot squad targets, and looks like fire at will on heroes.    Is he wasting shots on structures or squads?  Or fire at will with the 15 second cooldown wasting the skill?

Reason for Karma (Optional)
Successfully updated karma reason!
December 1, 2010 1:54:03 AM from Demigod Forums Demigod Forums

Pounce could probably be used more liberally than that.  It's a really cost-efficient ability.  I'd say 1200 mana or 30% - that'll still leave more than enough for one more use of it.

Grasp, on the other hand, is really expensive.  UB is better off using spit repeatedly due to the massive efficiency advantage and fast cooldown, and specifically saving grasp for interrupts only.  The only exception to this is on a fleeing enemy at VERY low health, like <1000 or ~20%.  Even then, it's often only useful when an ally is around to help finish the target off, due to the small follow-through/order re-issue delay making it very difficult for UB to resume pursuit on a similar-speed enemy.  1.3 helped this, but there's still a little hiccup.

 

What kind of skills do they try to interrupt?  Anything with more than a certain cast time, or anything they catch with x cast time left?

Reason for Karma (Optional)
Successfully updated karma reason!
December 1, 2010 10:26:53 AM from Demigod Forums Demigod Forums

 

If i recall you can inspect what skill or item is actually being cast and its cast time. I'd have to put loop over all the columns again to see what all is available.  I thought saving the interrupts for tp's only would be cheesy.  A player doesn't know if they have 2 seconds or .5 second when the animation starts.

There is also some delay in picking skills to use, although upping the interrupt skill weights may get them checked more often, or there might be a loop with a wait x in it somewhere else deciding when to check skills that we could shorten.

Reason for Karma (Optional)
Successfully updated karma reason!
December 1, 2010 10:30:04 AM from Demigod Forums Demigod Forums

Quoting Peppe,


What are the issues with regulus snipe?  He has 3 default actions for using the skill.  Kill shot structures, kill shot squad targets, and looks like fire at will on heroes.    Is he wasting shots on structures or squads?  Or fire at will with the 15 second cooldown wasting the skill?

My 2 cents .

If possible can you make it so that he pressures 1 target (with snipe going as much as possible) until they retreat to the crystal and then begin to pressure another (switching only when the 2 other options occur)? Currently he changes target alot and tends to pick generals to snipe (which gets healed off pretty qucikly). Possibly always choosing the demigod with the least health, and preferably an assassin would be good as well. 

 

Also, Reg seems to take staff of renewal everygame, which while useful is not as useful as blade of the serpent for him. Particularly with mana intensive builds

 

An amazing job so far guys!

 

Exx

Reason for Karma (Optional)
Successfully updated karma reason!
December 1, 2010 11:06:56 AM from Stardock Forums Stardock Forums

Cloak has 3 actions. Flee, warp to squad target, warp to enemy hero. I can disable the warp ins leaving just flee. Or make the warp to hero condition only when the target is at like 500 health or lower.

Ah - ok then, perhaps leave in the flee option only then and disable the other functions.  I can see the ai getting killed pretty easy even if it hops at someone with 500 hp, but either way I guess.  As long as its not hopping to start a battle. 

Nothing setup for TB

Can something be coding for his ice mode to try to interrupt?

What are the issues with regulus snipe? He has 3 default actions for using the skill. Kill shot structures, kill shot squad targets, and looks like fire at will on heroes. Is he wasting shots on structures or squads? Or fire at will with the 15 second cooldown wasting the skill?

Or does he generally not fire... that one.  If I had a choice, I'd only have reg use snipes on enemy dgs, preferably at max distance or if a kill shot appears.  The ideal scenario on cataract, for instance is to have a reg on one side of the map and walking some to snipe an enemy dg on the other side.  As it is, reg's don't use snipe all that well or much. 

If possible can you make it so that he pressures 1 target

That is also good.  The ideal target to harass would the the dg with the lowest hp.  Constant harassment can force a player to head back to the crystal.

 

 

Reason for Karma (Optional)
Successfully updated karma reason!
Stardock Forums v1.0.0.0    #108433  walnut3   Server Load Time: 00:00:00.0000328   Page Render Time: