Are you, in fact, a pregnant lady who lives in the apartment next door to Superdeath's parents? - Commodore

Create an account  

 
Todo list for 5.44

We've accumulated a lot of issues and ideas that'll need some sort of decision for the next update.

1. I see far too many Stoning weapons in random generated items. This is a result of Stoning being the only very cheap item power on weapons, having a high chance to appear in general, and weapons being a fairly frequent item type to show up. The game rolls a D10 for item type, and there are 3 armor types, 1 accessory type and 6 weapon types. We probably want some weighting here to make random items follow the distribution of item types in a similar fashion predefined items do. We likely also want Stoning's weight to be lower. (There are only 5 things to choose from at 125 cost that's not Stoning.)

2. Not sure if they do already but maybe the AI should prioritize ships less for buff targets. While a warship with heroism endurance etc is formidable it's not a direct threat to the player as it can't attack on land.

3. It might be interesting if Guardian added more "power" to combat spells for the defender, to allow strategies not based around units (and AIs) to take better advantage of it. Something like +1/3 strength on direct damage spells and -1 spell save on resistance based spells.

4. The AI seems to still move units before shooting when Wall of Shadows is in effect, this will need to be fixed

5. We might want to consider making Regeneration a variable amount (restore 0-1 health on lower tier units and 2-3 on higher tier ones).

6. AI should research Exorcise/Holy Word, Cockatrice and Shadow Demons faster if the human is playing Death/Werewolves. We can either trigger this on number of human Death books (1? 5? 8?) or on specific spells known (Lycanthropy or Shadow Demons).

7.Werewolves need to lose some resistance. (They are immune to all early effects but exorcise so same deal as Shadow Demons, but more figures and cheaper unit so resistance needs to be lower)

8.Werewolves might need to get the 3 armor 5 hp change undone and restored to 1 armor 6 hp. I'm still quite unsure about this one. Might not be needed at all if 5 and 6 are implemented.

9.Enchant Item should be able to have Spell Charges, but limited to weaker spells only. (commons? Uncommons? Anything below cost 12? 15? 20? 25? This is important to enable heroes to get Life Drain/Heroic Shout charged weapons.)

10.AI needs to learn to target adamantium units with Lycanthropy? Might be useless as adamantium requires alchemist guild and makes the AI build up military buildings, once the AI has those, they will usually be producing higher tier units that won't be targeted anyway due to high cost. Forcing them to make spearmen results in diluted armies with some useless adamantium spearmen mixed in.

11. Save or die per figure effects the AI should probably learn to target at lowest figure count in already lost battles, but highest figure count otherwise. (currently targets highest figure count only so those shadow demons/wraiths/death knights never really die to it)

12. Demon Lord seems to have save at -2 in effect but -3 in help and I don't remember which was the actually intended amount.
Reply

1. No idea, I'm not a heroes guy.

2. Can we specify 'if 2 ships that don't have wraithform or flight already have this buff, then reduce priority of targetting other ships that don't have wraithform or flight'? Having 2 strong ships is definitely helpful but diminishing returns on the rest.

3. I could see that.

4. Yup, in combat decisions can always be tweaked. Good eye.

5. I agree with that. Specifically I like that it's a roll with a 50% chance of regerating one less than 'normal'.

6. Base it on 8 books. I prefer not basing anything on actual spells known. Unless you think a 5 book start can still get adamantium werewolves going fast enough to beat war declaration turn, in which case base it on 5 books.

7. I'd be much happier with this than making neutral ais end up with 1 armor again. 

8. See 7. Adamantium is still the problem, not the base stats. See if anyone can have those win ratios with werewolves without playing rich myrran. As mentioned a long time ago, rich myrran is still far too different from fair arcanus. So trying to balance around rich myrran is still not going to help the baselone testing you've been doing.
Reply

2. No, space is extremely tight here. Not sure if I can even do a simple "if shop, priority-50" but hoping.

6. This is entirely up to luck if there are fewer than 8 books. Although it might be impossible to have an uncommon right in the starting choices that way? You'll have the 3 Arcane commons and then 5 from your realm (unless you selected that few books but you can't do that with only 4 retorts). Problem is the AI researches some of those spells so late, even if you don't get turn 1 Lycanthropy, you're still good. I mean, Exorcise comes after half the rares. (Shadow Demons and Cockatrices are summons so they are early. But if the AI already got another summon, they're not and get pushed to the end of uncommons.) So expect these spells to show up around 1408-1412 on more reasonable difficulty levels, by then even a late werewolf can conquer half the world. (and it's not about adamant, I'm doing fine on Arcanus with Mithril.)

Edit : we might also want to have a diplomacy based limitation on this like other similar AI decisions : only do it if having no treaty with the human? It's a waste to research exorcise otherwise although the creatures are good either way.
Reply

9. I would just include all common spells.

10. I wouldn't teach the AI too much about werewolves. Players would hate to face the super strong ones, and playing rich games would end up requiring dealing with it which reduces gameplay too much. At best I would teach them to target units with adamantium > Mithril > magic > normal weapons, but keep all cost of unit targetting the same as now, and weapons only determine target after cost choices are finished.

11. Agreed, combat decisions should keep getting tweaks.

12. I would guess -3 but I don't recall.
Reply

AIs abusing their mana and starting positions cheats (thanks to settlers beelining towards adamantium) in combo to show up with ada wws in y2-3 will be fun.


I propose item: make AI resource advantage progressive, +1%/month. No need to reduce it. Just increase it for now. The reduction will happen if needed with balancing.
Reply

Quote:I propose item: make AI resource advantage progressive, +1%/month.
This is already on the rejected suggestions list, see http://www.realmsbeyond.net/forums/showt...p?tid=9029
Reply

I can't find it unless you refer to
AI advantage scaled by cities or by turns, AI has too many nagas etc summons too early

That's an answer to the request of less AI resources at the start, I'm asking for no change and more AI resources gradually as a means to increase the difficulty of high difficulty levels.
Reply

Yes, that's the same thing with different numbers.
It's not helping, the only problem with current AI is it's too vulnerable in the early game. The game has exponential growth, if the AI doesn't lose territory early, they are usually powerful enough.
Reply

1. Sounds good, stoning ability shows up maybe too often.

2. Buffing naval units should be low or no priority. The only way it might help AI to buff ships is resist magic to war ships.

3. Something like +1/3 strength on direct damage spells and -1 spell save on resistance based spells. Sure.

4. this will need to be fixed

5. Regeneration a variable amount - I may disagree (less intuitive, biased against low armor units) unless regeneration is based on % of hp, then totally cool with it.

6. AI should research Exorcise/Holy Word, Cockatrice and Shadow Demons faster if the human is playing Death/Werewolves. Sounds good

7.Werewolves need to lose some resistance. - 5 resistance would be good

8.Werewolves might need to get the 3 armor 5 hp change undone and restored to 1 armor 6 hp. How about 2 armor 5 hp?

9.Enchant Item should be able to have Spell Charges, but limited to weaker spells only. (commons? Uncommons? Anything below cost 12? 15? 20? 25? This is important to enable heroes to get Life Drain/Heroic Shout charged weapons.). Probably just commons or spells costing less than 20. Uncommons could result in powerful spells like prayer.

10.AI needs to learn to target adamantium units with Lycanthropy? Not sure how this can be done without hurting AI in some other way.

11. Save or die per figure effects the AI should probably learn to target at lowest figure count in already lost battles, but highest figure count otherwise. Fine either way.

12. Demon Lord seems to have save at -2 in effect but -3 in help and I don't remember which was the actually intended amount. I don't remember either, probably -3

Reply

(September 22nd, 2018, 05:38)Seravy Wrote: Yes, that's the same thing with different numbers.
It's not helping, the only problem with current AI is it's too vulnerable in the early game. The game has exponential growth, if the AI doesn't lose territory early, they are usually powerful enough.

wat

I have the exact opposite experience, as long as I pass the first 4-5 years I win, AIs can't do shit once the situation is.. like the one in your latest test game, judging from the map screenshot.

AIs need help in the later stages.
Reply



Forum Jump: