Some of the AI updates coming in 1.1

Just checked this in:


----------
12.07.12
----------
+ Moved AIEvaluateAreaThreat to AIfaction
+ AI looks for enemy defenders sitting on top of targeted improvements in AIFindClosestEnemyImprovement
+ AI trait weighting reduced from 1000 to 100 for researching techs purposes
+ AI checks to make sure a given unit isn't a city guardian unit before doing AI on it
+ AI calculates the battle rank of its champions and sovereigns each turn afresh to accurately reflect how strong it is prior to operating on it
+ AI no longer uses post message to leave a city in order to ensure it has real time strength notice of a target city
+ AI caps how much it will count crystal, influence and metal when factoring what it should research
+ AI vastly better at choosing technologies to research
+ AI is more protective of its cities
+ Fixed bug where AI pioneers couldn't build outposts if they were in an army with a champion or a sovereign (hence, armies of pioneers late game)
+ Fixed AI bug where sovereigns and champions, late game, would tend to lone wolf attack cities
+ AI pioneers better about staying put if there is an area threat around them (i.e. will stay inside a city)
+ AI values armor more than before when designing units
+ AI champions now care more about whether the city is defended well enough before leaving a city
+ Fixed bug that prevented AI from casting strategic attack spells on enemy units
+ AI interacts more with players
+ Modified Average combat rating to be 100 instead of 200 in elementaldefs.xml
+ Fixed crash bug in elementalsectormapper where the ZOC would get updated in a thread
+ Fixed bug in mapgenerator.cpp in which terrain decals were being loaded and placed (we don't display them so we were just using CPU and memory for no reason)
+ Modified the prop code so that deserts display more props and other terrains show fewer (mapgenerator.cpp)
+ Fixed path finding bug that caused any AI units that intersected during movement to cancel their destinations. This prevented AI units from being able to effectively create armies or get around late game (basically crippled AI movement late game)
+ Modified the combat rating calcualtor to consider defense a bit more

 

We're going to tweak easy and below to be easier because of these changes.  My 1.05frog build (https://forums.elementalgame.com/437369/) has these improvements in it already for anyone who wants to try a private build.

 

One thing I learned from extensive testing is that the AI "quality" is heavily dependent on which factions you play against. Some factions are weaker than others (in terms of the AI's effective use).  Krax and Trogs are the ones the AI loves the most.

36,385 views 41 replies
Reply #1 Top

Awesome.  Thank you.  I will be sure to pick Krax and Trogs for my next game!

Reply #2 Top

Can we get some better strategic spellcasting as well? Most importantly Tremor and Freeze when I enter their areas when at war.

Reply #3 Top

+ AI trait weighting reduced from 1000 to 100 for researching techs purposes


What does this mean exactly?

Reply #4 Top

No diplo fixes so far? :'(

I still hope some make it in  ;)

Reply #5 Top

+ AI interacts more with players
End of quote

 

What does this mean?  Better diplomacy AI, or just more of it?  Or something else, and if so, what?

Reply #6 Top

Actually I have found that Gilden is by far the most effective AI faction. They always build wonders, and often end up with big, heavily-armored late game armies of barons that will whoop your butt unless you are faster/better at magic.

 

My first game ever: Gilden ended up on a spree of continuing to research refined defense and attack, apparently. I was forced to specialize in lots of magic use in order to kill those super barons and knights. 

 

But since then, no AI has managed to muster up a decent challenge. It is rather disappointing. The game has so much potential - hopefully the AI will eventually learn how to handle the game mechanics. :)

Reply #7 Top

Quoting Glazunov1, reply 6

+ AI interacts more with players

What does this mean?  Better diplomacy AI, or just more of it?  Or something else, and if so, what?
End of Glazunov1's quote

It means it's better at making long-winded speeches.  You'll finally get to say 'Have some light balls.' ;)

Reply #8 Top

What effect do these have?

+ Moved AIEvaluateAreaThreat to AIfaction

+ AI trait weighting reduced from 1000 to 100 for researching techs purposes

+ Modified Average combat rating to be 100 instead of 200 in elementaldefs.xml

Also, has the AI insta-build on reload bug been fixed?

Reply #9 Top

I've also had a good time against Gilden. They can field very good units.

Wither is also a spell the AI should utilize strategically. Even without shards it's good value for the cost against armies. With enough shards it's a game winner.

Reply #10 Top

Quoting Borg999, reply 9
What effect do these have?

+ Moved AIEvaluateAreaThreat to AIfaction

+ AI trait weighting reduced from 1000 to 100 for researching techs purposes

+ Modified Average combat rating to be 100 instead of 200 in elementaldefs.xml

Also, has the AI insta-build on reload bug been fixed?
End of Borg999's quote

The first point may relate to this one:

Quoting FrogBoy,
+ AI pioneers better about staying put if there is an area threat around them (i.e. will stay inside a city)
End of FrogBoy's quote

The trait weighting comes into play when the AI has a trait that affects its research - War Monger, Sorcerer, or Builder.  So it will be far less likely to obsess over one tech tree.

Average combat rating refers to what we see in the game as a Medium combat rating, so I suppose in this case an army will not be considered Weak when it's not really all that weak.

Reply #11 Top

Quoting MarvinKosh, reply 11



The trait weighting comes into play when the AI has a trait that affects its research - War Monger, Sorcerer, or Builder.  So it will be far less likely to obsess over one tech tree.

Average combat rating refers to what we see in the game as a Medium combat rating, so I suppose in this case an army will not be considered Weak when it's not really all that weak.
End of MarvinKosh's quote

 

Thanks!

Reply #12 Top

+ Fixed bug that prevented AI from casting strategic attack spells on enemy units
End of quote

Just a reminder.

Reply #13 Top

Hi Frogboy, since everyone is ignoring my threads I'd post this here for some assistance. Thanks!

 

Where is it? I paid for it 

 

Order placed on Friday, December 07, 2012    (Order #: 2363257)

Billing Address

XXXXX
US

Order Summary

Fallen Enchantress Map Pack DLC

(Digital Download)
 
 
Soo... where is it? It's not on the stardock download page. I've signed out and in multiple times. It's not on your Stardock Central thing.
 
It's nowhere to be found so...
 
yeah.. where is it?
Reply #14 Top

One thing I learned from extensive testing is that the AI "quality" is heavily dependent on which factions you play against. Some factions are weaker than others (in terms of the AI's effective use). Krax and Trogs are the ones the AI loves the most.
End of quote

Well Duh, some of these factions are also way simpler and have easily "abusable" tactics, compared to the more difficult Altar, and the completely useless Pariden.
Half of that is faction balance taking effect...

Sincerely
~ Kongdej

Reply #15 Top

+ Fixed bug in mapgenerator.cpp in which terrain decals were being loaded and placed (we don't display them so we were just using CPU and memory for no reason)
End of quote

Good to hear this.  I sincerely hope that the Dev team will continue to find ways to tighten up the code in order to use the minimum amount of memory necessary to get the job done.  As things stand, the game still crashes way to often.

Reply #16 Top

Quoting MarvinKosh, reply 8

Quoting Glazunov1, reply 6
+ AI interacts more with players

What does this mean?  Better diplomacy AI, or just more of it?  Or something else, and if so, what?

It means it's better at making long-winded speeches.  You'll finally get to say 'Have some light balls.'
End of MarvinKosh's quote

 

Flavor's never bad.  The diplomatic flavor in GC2 was very good, but it also had more diplomatic options, and most important of all, it all worked together well.

 

I suspect you're right that we're only getting the flavor part, though.  They still don't think we care that much about diplomacy.  Maybe 1.2?  1.3?  Perhaps the first major add-on?

Reply #17 Top

Something that has always irked me is that there isn't a 'How about you switch sides and help me fight X instead?' option in diplomacy.  You have to first sign a peace treaty and then bribe them to go to war.  It would open up AI trait possibilities like Loyal (tends not to betray an ally) or Backstabber (tends to jump at the chance to betray an ally).

Reply #18 Top

Will this fix the instadeclare on higher difficulties as soon as you meet them? Because exploring is stupid when playing higher difficulties, the longer you are unnoticed the better, and that seems somewhat upside down.

Reply #19 Top

Quoting Kongdej, reply 15

Well Duh, some of these factions are also way simpler and have easily "abusable" tactics, compared to the more difficult Altar, and the completely useless Pariden.
Half of that is faction balance taking effect...
End of Kongdej's quote

Periden is not useless, though they are weak in stand-up fighting -- the initiative bonus can be incredible, and the earth shard bonus counts even when they are level 1.

I expect they will be overpowered in 1.1.  :)

Reply #20 Top

Quoting dihir, reply 20

Quoting Kongdej, reply 15
Well Duh, some of these factions are also way simpler and have easily "abusable" tactics, compared to the more difficult Altar, and the completely useless Pariden.
Half of that is faction balance taking effect...

Periden is not useless, though they are weak in stand-up fighting -- the initiative bonus can be incredible, and the earth shard bonus counts even when they are level 1.

I expect they will be overpowered in 1.1.  
End of dihir's quote

Pariden is imo the best faction for higher difficulties. (unless you custom)

More essence = more Mana 

Mana = Power  (Land =/ power in this game regardless of what you might think)

Go abuse Destinys Gift and see for yourself. If you expect to fight AI with units at Ridicilous+ you will be disappointed.

But if we are talking about AI, we might as well uninstall the game if they start abusing Destiny's Gift as well ^^

Reply #21 Top

Quoting sjaminei, reply 21

But if we are talking about AI, we might as well uninstall the game if they start abusing Destiny's Gift as well ^^
End of sjaminei's quote

I've been seeing the AI use destiny's gift (and, on insane, I was dealing with sov's with higher initiative than me and of course much higher health than me), but in 1.02 the AI has so many problems that even destiny's gift does not help it enough to make a difference.

And, of course, when you take away its toys, then you have them... so here's an unbuffed, level 1 Amarian unit from that situation:

 That's 40 initiative with a ranged attack, the defenses and health did not really matter much because they rarely got hit (but later, I buffed them with evade and nature's cloak, just for fun, and I also buffed them with regeneration because when they leveled up they needed to heal to get the newly awarded hp).

Reply #22 Top

Quoting sjaminei, reply 21
Mana = Power (Land =/ power in this game regardless of what you might think)
End of sjaminei's quote

I think that depends on the world's Magic Strength setting. I have it set to "Random". The game I'm playing at the moment must be Sparse - there's probably no more than ten shards on the whole map. I've got roughly twice the territory of any other factions, and I've only got three shards. I have to be conservative with my mana because there isn't enough to go around. In this world, physical, not magical, might is king.

Reply #23 Top

Quoting wildez, reply 23

Quoting sjaminei, reply 21Mana = Power (Land =/ power in this game regardless of what you might think)

I think that depends on the world's Magic Strength setting. I have it set to "Random". The game I'm playing at the moment must be Sparse - there's probably no more than ten shards on the whole map. I've got roughly twice the territory of any other factions, and I've only got three shards. I have to be conservative with my mana because there isn't enough to go around. In this world, physical, not magical, might is king.
End of wildez's quote

Map size also matters.  I was playing the huge map mod, in the above screenshot.

Reply #24 Top

Quoting dihir, reply 20



Quoting Kongdej,


Well Duh, some of these factions are also way simpler and have easily "abusable" tactics, compared to the more difficult Altar, and the completely useless Pariden.
Half of that is faction balance taking effect...
End of Kongdej's quote


Periden is not useless, though they are weak in stand-up fighting -- the initiative bonus can be incredible, and the earth shard bonus counts even when they are level 1.

I expect they will be overpowered in 1.1.  
End of dihir's quote

I still expect Pariden with be a pushover compared to other factions until the AI learns how to properly (ab)use the Decalon trait (namely spamming Arcane Monolith to start claiming and improving strategic resourses long before they build a city to take advantage of it).

We could also go on for days about whether or not the scrying pool from Pardien's Enchanters trait truely benifits from having a sov that spent 6 points unlocking three schools of magic when Decalon eventually unlocks all but Life/Death for the sov and all other champions to use, all with no direct bonuses to mana beyond Procipinee's crown negating enchament upkeep costs, but that's neither here or there.

Also, holy crud, I'm going to be bombarded with freeze and flame pillar when I enter hostile territory now? I don't know if that's awesome or scary...

Reply #25 Top

Pathfinding needs work too.

For example:

 

-2 cities, you can see the first at the bottom of the above pic, the second one is due north at the top of the pic, connected by a road goes that goes on a diagonal  northeast then diagonal northwest.  Champion (yellow cape guy) with movement=2 was located in the southern city.  Champion can be manually moved from southern to northern city in 1 turn (manually exit on road then follow road into second city).  Reloaded, selected the champion and then just clicked on the highlighted square south of the crystal (just outside the northern city) and pathfinding had the unit exit the first city off-road then head due north (off-road, ignoring the existing connecting road).  Ended up requiring 2 turns to reach second city when it should have taken 1 turn.  If I click on a square in the city (the crystal square) then pathfinding will get the champ into the northern city in 1 turn, so that works.  Pathfinding should work similarly in both cases.

-exiting a city, too often a unit exits off-road when a road square could be used.

-if a 2-move unit and 4-move unit are stacked together at the start of their turn and the 4-unit is detached, its movement is 2/4 instead of 4/4.

 

Similarly...

I just moved the 3-horse scout unit adjacent to the yellow cape champion to recruit the champ, using up all the scout's movement.  Champ still had 2 movement left.  I wanted the champ to attack the mob 2 squares to his north.  To do this I'd have to move thru the scout's square, but doing so causes them to group up and the champ loses his remaining 1 movement.  Now, I can move into another unit's square without grouping if I click past said square, so I tried to click directly on the target mob to see if pathfinding would pass thru the scout to attack the mob (as moving thru the scout's square takes 1 movement, leaving 1 movement to attack the mob).  Nope, pathfinding moved the champ due north into the 0.5 movement-required square, using up all the champ's movement, and no attack was possible.  Is there a way to have the champ attack the mob in the current turn, that I'm missing?  The attack should be possible.

The game is much improved, and getting better with each update.

 

Nicely done!