btw I'm still looking for suggestions to replace "rule 4" on the AI.
|
Current balance of normal units
|
|
November 26th, 2017, 19:19
(This post was last modified: November 26th, 2017, 19:23 by Nelphine.)
For me, rule 4 plays into overall production rules.
While you've given them to us, I haven't had time to look over them in detail yet. I do think nomads should do stables instead of fighters guild, even with the strategic strength difference. I think its amusing that they're forced to build the smithy, when I almost invariably sell smithies from conquered cities. That being said, even ignoring the differences for specific races, I don't think we can avoid having AI go straight to fighters guild. The AI simply doesn't know when its units are relevant. It can't build several cavalry, several bowmen, several swordsmen, and then use the appropriate one based on the human units. It only knows strategic strength. Unlike the human who can take advantage and attack with the right unit mix. Basically they are lovely units, balanced against each other, that the AI doesn't understand and so it has to skip to halberdiers which have no specials (and so therefore the AI actually understands them). Which again leads to, we can't make any of the four low tier units strong enough to encroach on halberdiers, and the AI is going to skip them anyway. The best we can do is increase them to make halberdiers not as cost efficient. (But note, this already leads to humans using halberdiers instead of anything better due to cost efficiency, so in extreme cases this could quite literally lead to hunams never building fighters guilds, just like I never build armorer's guilds.)
Tried to look up the discussion about implementing this AI "rule 4" and realized it's nothing like we thought. It wasn't added because the AI needs the fighter's guild units to stop summoned creatures or anything like that, no. They have their own summoned creatures for that and those even stop attacks from enemy units better than halberdiers ever will.
What it was added for is, 1. to prevent the AI from overextending, that is, pumping out settlers endlessly without raising any of the cities in military tier, ending up with an empire that has 10 cities already but still produces swordsmen everywhere on turn 70 which makes it easy to take for the human and 2. To give the player some place for their own settlers instead of hogging everything like there is no tomorrow, resulting in the human not being able to plant their second settler at all. Forced production of a Fighter's Guild, an expensive building, creates a ~10 turn gap between settler 2 and 3, which is not much and won't threat the AI's performance, but does mean the human player gets a much better chance to build at least 2-3 outposts before there is no more land to build on nearby. Note that I haven't found any discussion about this around the time it was added - I merely remembered this all started with the "make the AI produce less settlers, this is unplayable" complaints. Which was bad for the AI as it was serving easy to conquer cities for the human instead of being a threat, and also bad for the human because not being able to use settlers all the time is annoying. I'm not sure if this was before or after the fixing of the AI bug where they only sent settlers to lands near the human, but I suspect it was before. So that probably shouldn't be a concern anymore - but overextending without the ability to build better units is a very valid problem. Cities will generally have units they can produce as the garrison, so if the AI does do other stuff like economy buildings first, they are just doing a favor for the human by giving them a better city. So we probably don't want it to do other buildings without other units - but mixing the two might not be that bad. Swordsmen do horrible at defending cities sadly as they are wiped out by poison and bows as well. Cavalry is slightly better, it works well vs ranged attackers, but fails against melee units like nagas or hell hounds anyway. Bowmen are semi-ok actually, as long as we are fine with Guardian Wind and Skeletons becoming stronger for human players. So giving the AI a small timeframe for bowmen and cavalry while they can also get their marketplaces and such before the FG might not be a bad thing...but I don't think it's good either, more like, about the same - assuming they do get the FG going in the capital at the very least. We can also hope the AI hits the city with the summoning circle and dumps a few summons into the garrison to improve it somewhat, so it does well without halberdiers. Honestly, I don't think halberdiers are that good for the AI - they are as weak to ranged attacks as swordsmen. They only work against certain threats like bears or nagas but fail vs ghouls. So FG first is only really a good tactic if the AI has a race with a strong FG unit like lizardmen and barbarians or the game did progress into the state when the AI has a decent production capacity to support massive stacks of halberdiers. Overall, I do think we have some room to tweak rule 4, but not too much - The FG tactic should still come into play early enough on the larger cities, but the small ones might not need it on certain races. (But definitely for others. I do think Dwarf, Gnoll and Barbarian must always go for their military buildings - steam cannons, berserkers and wolf riders are that good.) I might underrate the importance of AI halberdiers though. We haven't had a chance to see an AI without the broken, overdone "rule 4" since the release of...I think, 4.0 or around that when the new AI overland production procedure was made so I can only speculate how good or bad the rule is currently. We probably need to go through races separately. Barbarians : definitely do FG. Nothing compares to berserkers here. Gnolls : Stables here which is cheap and generated production so can't go wrong with doing that early and Wolf Riders are probably better for the AI than halberdiers anyway. Beastmen : Might do better if FG is not rushed - their FG units aren't that much better actually, so doing FG somewhat later might work better here. FG should be the building to go for though - as it is needed for Centaurs and as an intermediate step towards Minotaurs and Manticores. Dark Elf : depends on how we update the race, their halberdiers actually aren't that great as is compared to swordsmen, and the AI knows how to use the ranged cavalry. Maybe force the stables early instead? Draconian : Not sure if their halberdiers are really that much better to be worth the FG rule and they only need the FG for doom drakes as next tier racial unit. Forcing the Maritime Guild for Air Ships might be interesting. However being a Myrran race, strategic strength is important here and only halberdiers have a good one for the race. Forcing FG only if the human is not Myrran might be best? idk. Dwarf : Forced University for Steam Cannon, perfect. They don't have bowmen and cavalry to build to begin with so they have to upgrade. Halfling : They no longer have bowmen and never had cavalry while their swordsmen are subpar without buffs so they must do the fighter's guild unfortunately, even if the slingers aren't that amazing for AI use. The only other alternative would be the wizard's guild for magicians but that's way too expensive. Or Shamans but they are far too weak. High Elf : Longbowmen are great but we can't force those easily because Forester's Guild needs a Forest tile which we might not have access to. The race benefits from early granary and such a lot though, so delaying the Fighter' Guild, but forcing it later seems the best solution, similarly to beastmen. High Men : If we add a new unit, we'll probably need to teach the AI to go for that, if the new unit is early enough. Otherwise, not sure, the Pikemen aren't that good for the AI, but the rest are bad too. Klackon : No choice here, must do early FG, that's the correct play with the race, even for the human. Lizardmen : Same, no bowmen or cavalry, but Javelineers instead... Nomad : I guess stables first and FG at a later time would work better than only FG but early. Orc : Orc units are not strong to begin with, the race is more into the growth/economy. Halberdiers and even Hordes are still easy to kill. So upgrading to them might not be the best tactic, but I have no better idea. Troll : Their swordsmen are actually not that bad but being a military race, they probably should try to pump out as strong units as possible as early as possible. So unsure on this one as well.
Seravy, I've just broken through the towers in a game to find a nomad AI with all the cities full of mages. It's got bears and nagas, I guess the mages are considered a stronger unit.
So, thanks to the fact that there are only ranged magic units I can conquer them all with a single hero with high enough resistance/invisibility, or nightblades, golems, or anything else that works with this kind of opposition. For me what's more important is to try and teach the AI to use mixed troops for defence. Sure, the strategic strength will be lower, and AI vs AI fight could be worst but for humans having a single type of defence make attack too easy imho. The use of the weaker units means however that the roaming stacks could be stronger and be ready for aggression earlier! So what I'm suggesting is to tweak rule 4 and the garrison rule at the same time. Ideally I'd like them to have 1/2 shamans, some mix of melee and ranged units, and eventually to upgrade the melee from swordsmen to halberds and from bowmen to mages in general... 
Ehh, I'm having serious doubts about mixed garrisons. So they have 4 mages and 4 halberdiers?
Use the invisible unit or equivalent to kill the 4 mages, retreat, now kill the halberdiers with any army. But you don't need to resort to such trickery. Bring 9 mages of your own. You lose 2-3 on the first turn but then the remaining ones destroy the enemy mages and then also the halberdiers because 9 mages are better than 4 mages and 5 halberdiers. The AI uses magicians only because they are strictly better than mixed garrisons in most cases. There are potential exceptions, for example griffin mixed with mages works well. But there is one more concern. Units that don't stay in the city will be pushed into offensive stacks. Mages are bad at doing that. So while the magician and the griffon are equal in the garrison, it's clear the griffon is better in the offensive army and shouldn't be wasted as garrison. ...Also, magicians are easy to unlock. Other units good enough to mix with them generally require an armorer's guild or equivalent which will not be available for a very long time. So even if the AI happens to mix the units, they'll still have only magicians for a long period of time.
This is the current garrison priority table. Units with identical or similar values can get mixed - each level is worth +2 points. So a level 2 unit with 9 score can be mixed with a level 0 that has 13 score, or another level 2 with 9 score. Lower score leaves the city, higher stays.
Code: Necromancer HERO 120 Feel free to suggest improvements to this table.
I see, I don't value much all mages defences, they are just so easy to kill with their abysmal health. You need to find a way to get over the first turn barrage, but that's not so difficult... I would drag them down in priority quite a bit. Example: doom bats are much scarier imho, and water elementals or fire giants are much sturdier... Even dragon turtles make me worry more than magicians. But maybe this is to avoid keeping good units in defence? If that's the reasoning then I see the point. But then why is gorgons higher? I guess I don't have enough info on the motivation for the priorities.
OTOH, your argument about the double army doesn't really work I think. Sure, you can first attack with an invisible unit (or similar) and deal with the mages and then with the halberdiers afterwards. But that's still more effort than dealing with the mages only with your invisible unit and that's it, no? And with mixed troops you can still upgrade them by type... Wait, this is the difficulty isn't it? It would be too difficult to have categories of garrison units and to make logic that considers the categories? If this is the issue then forget all I've said (or rather add it to the issue list - mixed defenders needs categories logic which is too hard to do) On a different note, heroes are also quite easy to deal with as long as they don't have uber items. They are a kind of category with its own embedded logic (which is, max 6 units of the category in total). A beastmaster (second top in the table) without levels/uberarmor will give her bonus only for the first of 25 battle rounds. Wouldn't it be better to send them out adventuring? Is it possible to get the AIs to assess items too, like they already do levels? That sounds quite difficult, tbf. And buffs? Quote:Wait, this is the difficulty isn't it? It would be too difficult to have categories of garrison units and to make logic that considers the categories? If this is the issue then forget all I've said Categories are impossible. We can have reduced priority for additional copies of the exact same unit though. The Beastmaster and Necromancer heroes have the highest equipment priority, so they will be well equipped and very hard to kill. Especially with a stack of buffed very rare creatures ripping through whatever attacking forces you use while you try. Fortunately, this is one of the things that works better than expected. Doom Bats are low priority because they do well on offense and create amazing doomstacks. On defense they are not that great, being a melee unit with high speed they are likely to get used up and die in battles fairly quickly. As cities likely only have one or two, while doomstacks will be a group of 9, they are not a good defender. Gorgons on the other hand are quite generic - they don't excel at either role so there is no reason to force them to go offensive. Albeit, being a melee unit, they might be a poor defender after all, they are still better than something like a halberdier or even priest. Do note these tables are used only when the AI summons or produces a 10th unit. Otherwise, when building stacks, they will pull the unit with the highest strategic strength, except for leaving the best two in the city for garrison duty. So even if magicians have low priority, they are likely to stay because of their low strategic strength while at most 2 of the better unit stays. (Note the "2" can be 1 or 3 depending on AI personality.)
Completed the new, more detailed, all races included, rule 4 :
Code: ; Force military building as neededA quick summary of the code : Barbarian - very early FG Gnoll - Stables first Dark Elf - Stables first Halfling - very early FG Beastmen - lateish FG (~turn 50-60) Draconian - lateish FG is Myrran human, otherwise very early FG Dwarf - Builder's Hall and University first High Men - late Wizard's Guild (subject to change once we decide on the new unit or test the race more) Nomad - Stables first, and medium FG (faster than beastmen, slower than trolls) Troll - early FG Orc - lateish FG Klackon - very early FG Lizardmen - very early FG |