Default+Dunefolk era balance

Discussion of all aspects of multiplayer development: unit balancing, map development, server development, and so forth.

Moderators: Forum Moderators, Developers

Post Reply
User avatar
Celtic_Minstrel
Developer
Posts: 1225
Joined: August 3rd, 2012, 11:26 pm
Contact:

Re: Default+Dunefolk era balance

Post by Celtic_Minstrel » April 2nd, 2018, 5:18 am

I imagine something gets written to the console? I've never actually done it, but my guess would be that you just need to redirect stdout to a file.
Author of The Black Cross of Aleron campaign and Default++ era.
Maintainer of Steelhive.

User avatar
MathBrush
Posts: 106
Joined: February 12th, 2018, 10:21 pm

Re: Default+Dunefolk era balance

Post by MathBrush » April 2nd, 2018, 5:34 am

Well, I just decided to watch the playthroughs anyway.

I've only done Knalgan tonight. I used wesnoth --showgui --nodelay -m --controller 1:ai --controller 2:ai --era=era_dunefolk --side 2:"Knalgan Alliance" --side 1:"Dunefolk"

and alternated the sides.

It used the default map, and Knalgans won 70% of the time. The burners were surprisingly good, and falcons racked up a couple of kills. Piercers got many kills, even advancing at one point. The biggest problem seemed finishing people off; the knalgans would get worn down, but they wouldn't die. I'll try more people soon. The AI never recruited a single healer.

User avatar
Celtic_Minstrel
Developer
Posts: 1225
Joined: August 3rd, 2012, 11:26 pm
Contact:

Re: Default+Dunefolk era balance

Post by Celtic_Minstrel » April 2nd, 2018, 5:42 am

The AI doesn't use healers too well, either, if I recall correctly. At least, the default AI doesn't; there is a healer support candidate action that can be enabled. I'm not sure if you can use it from the command-line, though.

(It also doesn't use leadership abilities well; while it understands their effect and can take advantage of it if a leader is coincidentally well-positioned, as far as I know there's no candidate action which instructs it on how to best place them.)
Author of The Black Cross of Aleron campaign and Default++ era.
Maintainer of Steelhive.

User avatar
MathBrush
Posts: 106
Joined: February 12th, 2018, 10:21 pm

Re: Default+Dunefolk era balance

Post by MathBrush » April 2nd, 2018, 6:04 am

From my personal play through a designing the Dunefolk campaign and watching these replays (there’s more I’ll post tomorrow), I think giving the herbalist a ranged poison attack would make a world of difference. Everyone else has their place.

User avatar
skeptical_troll
Posts: 408
Joined: August 31st, 2015, 11:06 pm

Re: Default+Dunefolk era balance

Post by skeptical_troll » April 2nd, 2018, 9:16 am

Although the Dunefolk have balance issues, I don't think they are bad at the level that an AI test could be useful. Work was done in the past so we are not starting from scratch. The AI does a decent job in looking for vulnerable enemies and high-defense positions, but it's not capable of more advanced tasks like efficiently retreating at night (in this case), or using mobile units to threaten/protect villages and of course using abilities such as healing and leadership. I imagine that high-mobility factions such as drakes and dunefolk are harder for it to handle compared to neutral, positioning-based factions like Knalgans and Rebels. I think we shall really try to get some human data for this. I'd be happy to play some games, but my MP level is just around average. I guess part of the problem is that the 1.14 server, where the tweaks are implemented, is not too crowded yet?

User avatar
Elder2
Posts: 354
Joined: July 11th, 2015, 2:13 pm

Re: Default+Dunefolk era balance

Post by Elder2 » April 2nd, 2018, 9:43 am

Don't expect ai to use poison well, either.

User avatar
MathBrush
Posts: 106
Joined: February 12th, 2018, 10:21 pm

Re: Default+Dunefolk era balance

Post by MathBrush » April 2nd, 2018, 1:14 pm

I think that Dunefolk are so bad that AI can help. The Dunefolk just lost 9/10 battles against the undead. Burners were of course very useful. No healers were recruited. The maces of the rider and piercer were useful.

The dune soldiers have not done much against knalgans or undead. Whether it's useful or not, I'll continue doing the tests, as it's fun to watch.

User avatar
MathBrush
Posts: 106
Joined: February 12th, 2018, 10:21 pm

Re: Default+Dunefolk era balance

Post by MathBrush » April 2nd, 2018, 2:05 pm

I tested Drakes next, based on comments saying their high mobility was similar to Dunefolk.

It was 60/40 for Dunefolk/Drakes. Dune Soldiers did great against saurians, falcons and rovers both leveled up. Saurian augurs leveled up frequently as well.

Edit:

Rebels vs. Dunefolk:

40/60 Dunefolk/Rebels. Woses killed and were killed in great numbers.

Edit2:

20/80 Dunefolk/Northeners. Poisoned leaders and death by extreme force.

User avatar
MathBrush
Posts: 106
Joined: February 12th, 2018, 10:21 pm

Re: Default+Dunefolk era balance

Post by MathBrush » April 2nd, 2018, 6:00 pm

I haven't done Loyalists yet (when I do, I'll edit my last post, to avoid overposting).

But it's clear that Dunefolk AI has a major problem with the 2 chaotic factions' AI. However, it completely neglected the herbalist.

I would suggest that if advanced players want to start somewhere, they should start here, with multiplayer matches between Dunefolk and Undead or Orcs.

I know that knowing which factions you are facing can affect play. So perhaps the ideal would be 2vs matches. Having Dunefolk + Northeners vs Dunefolk + Undead would force all players to use general tactics instead of faction-specific tactics.

I don't know if there are 4 players interested in trying this out, though.

User avatar
Coffee
Developer
Posts: 180
Joined: October 12th, 2010, 8:24 pm

Re: Default+Dunefolk era balance

Post by Coffee » April 2nd, 2018, 9:06 pm

My 2 cents also as someone familiar with the game.

Short version:
AI tests are good, but need solid grounding IMO. 4 player tests would IMO be a waste of time, as you are unlikely to get 4 players of equal skill to play over and over again to do testing.

Long version :P
Problem with AI and factions with high mobility units from my testing with the "Experimental AI" and previous AIs is that they will send these units quickly at the enemy spaced out. Village grabbing is crucial also to unit positioning for AI in later turns and AI will also not retreat or heal effectively by default.

These things can be fixed however with a huge difference in success rate for the AI. I attempted to do this with my UMC campaign "The Great Quest" where I have candidate actions to effectively scout for villages, set caution and aggression for night/day/in-between, and recruit appropriately, etc. Whilst the aim was to make the AI better and more fun against humans, if played against regular AI I found it will crush it almost every time (even with gold advantage and an extra side).

I think a better test for AI usage would be a custom created small scenario that has a condition tag on enemies killed to reset the turn to turn 1 and respawn with random leaders from the same faction again. A counter could keep score somewhere visibly or in the output of the scenario. My machine has 12 cores and can run easily 10 of these scenarios at once. An input could be the map, which can be programmed to be loaded on startup so you get a variety of maps also.

With this type of custom scenario you could set aggression/caution/healing parameters to get more realistic outcomes, as well as force recruit occasionally of units that the AI will not recruit by default, but humans might well do.

I can give an example cfg file I think sometime today if I get time for what I am saying above.

For specific matchups my own gut feeling is that the Dunefolk have an issue with Northerners because of troll whelps and levelling up to trolls. The fire unit which could take these down is expensive in comparison and all units pretty much have impact resistance issues. Adding poison to the healer unit wouldn't solve this, but introduce further issues for other matchups.

The issue you are seeing where the Undead is overpowering the Dunefolk I think will disappear with better AI configuration. The effective use of the expensive horse with impact and the fire unit should affect this balance. If Dunefolk get low on gold and insist on recruiting then the low cost units aren't useful against the skeletons, similar to loyalist/undead matchup.

My guess is also that the Dunefolk vs drakes will remain imbalanced in favor of the Dunefolk. Anyway, we shall see I suppose.

User avatar
Celtic_Minstrel
Developer
Posts: 1225
Joined: August 3rd, 2012, 11:26 pm
Contact:

Re: Default+Dunefolk era balance

Post by Celtic_Minstrel » April 3rd, 2018, 12:03 am

ElderofZion wrote:
April 2nd, 2018, 9:43 am
Don't expect ai to use poison well, either.
They use poison better than healing or leadership. However, the default AI still doesn't use poison that well; as with healing, though, there exists a candidate action that attempts to improve this.

I would actually suggest running tests with the Experimental AI (which if I recall correctly includes both the healing support CA and the poison spreading CA), in addition to tests with the default AI.
Author of The Black Cross of Aleron campaign and Default++ era.
Maintainer of Steelhive.

The Black Sword
Posts: 373
Joined: October 13th, 2008, 4:35 pm

Re: Default+Dunefolk era balance

Post by The Black Sword » April 3rd, 2018, 7:49 am

AI testing is not very useful for balancing MP factions. The aim is to be balanced so that the matchups are even when people are playing 'perfectly'. It's ok if the faction is weak or strong when used by/against bad players. And the AI would probably lose to the lowest ranked players on the ladder.

There's no harm in doing AI tests, but if you want to help balance the faction I believe getting on to the MP server and playing matches would be a more efficient use of your time.

User avatar
Elder2
Posts: 354
Joined: July 11th, 2015, 2:13 pm

Re: Default+Dunefolk era balance

Post by Elder2 » April 3rd, 2018, 3:14 pm

dunefolk AI might be the weakest AI, weaker than elf ai because of the special gameplay.

Well, i could probably play some khalifate (still the name of the faction) if somebody wanted.

And yeah guys, of course, go play matches and post replays xD

User avatar
Pentarctagon
Forum Administrator
Posts: 3547
Joined: March 22nd, 2009, 10:50 pm
Location: Earth (occasionally)

Re: Default+Dunefolk era balance

Post by Pentarctagon » April 4th, 2018, 12:22 am

I should have time for a couple matches this weekend, so I'm planning to get on the MP server and hopefully play then. Obviously it won't be remotely high level play or anything, but it'll be a start.
99 little bugs in the code, 99 little bugs
take one down, patch it around
-2,147,483,648 little bugs in the code

User avatar
Pentarctagon
Forum Administrator
Posts: 3547
Joined: March 22nd, 2009, 10:50 pm
Location: Earth (occasionally)

Re: Default+Dunefolk era balance

Post by Pentarctagon » April 8th, 2018, 7:13 pm

Attached are a few replays. As I've mentioned, don't expect any sort of high level play on my part. That said, the two main things that stuck out for me were:
  • Having liminal and lawful units makes it much more difficult to decide when to attack or how to defend, since for the majority of the time between most and all of your units have a ToD malus.
  • Not having any additional terrain defense on forests or mushrooms also makes things harder. There doesn't seem to be any sort of compensation for this, they're just worse at dodging on those terrains.
Attachments
2p — The Freelands replay - Soliton.gz
(17.94 KiB) Downloaded 49 times
2p — The Freelands replay - skeptical_troll.gz
(15.57 KiB) Downloaded 51 times
2p — The Freelands replay - random guy.gz
(16.4 KiB) Downloaded 54 times
99 little bugs in the code, 99 little bugs
take one down, patch it around
-2,147,483,648 little bugs in the code

Post Reply