A nation the AI could use *well*

[Sombre]

It's simple enough to mod a specially built AI nation which could provide more challenge in SP games; Discounted recruitables, crazy starting sites, overpowered units and commanders etc.

But a more interesting topic might be what kind of nation the AI could use *well*. By this I mean rather than simply giving the AI better stuff, I'd like to discuss what the AI can use in a way which is more like a human. If a nation were created to be played well by the AI rather than be powerful it could also be balanced and even interesting for a human to play. Perhaps even a good one for beginners, who like the AI tend to make some odd choices and play in a rather simplistic way.

I'm not putting this in the mod forum because at this point I just want to identify the traits of existing nations/units which suit the AI well. Some thoughts:

The AI isn't capable of intelligent bless usage.
The AI doesn't tend to send battle mages with its armies that much.
The AI only VERY rarely uses stealth troops.
The AI likes to make blocks of troop types, so 'standard' units aren't a good idea for it. Remember the Honk Honk marverni hornblower army?
AI pays no attention to supplies.
AI can't really use blood properly.

The AI is generally pretty good with straightforward armies of medium infantry and missile units.
The AI can be relatively good/irritating with assassins.
Spellcasting AI in battle isn't too bad if the mage is suited to useful evocations. An AI abysia shows some appropriate mage-artillery usage. An AI TC generally doesn't.


To get the AI using more battlemages, it might work well to have all military leaders be spellcasters. So if the AI brings a swarm of 1000 troops and it mostly uses 40 or 80 ldr spellcasters, you will automatically get quite a few mages present and slinging spells. This is more aimed at NI maps, because otherwise the AI would still be using indy commanders.

Issues like supply ignorance can be countered by strong nature magic in the nation, troops that don't need to eat or leaders that otherwise give supply bonus. Any of these solutions would also be perfectly acceptable to a human player, given thematic justification and appropriate balancing.


So are there any traits/nations you've noticed that suit the AI well? Remember the idea here is to find stuff that the AI uses in an interesting and sensible way, not just stuff they win more with. How do you think the AI manages with the heavy infantry of Ulm

[Twan]

If given mages able to cast them without booster, the AI may be very agressive with offensive rituals (live wolven winter, fires from afar, seeking arrow, far summons...), which may or not be a good point. A nation with some mages specialized in rituals (ie : immobiles but powerful) and huge gem income may be good for the AI (but dont know if it will recruit these mages if others are available).

AI use priest reanimation well (except it sometimes have problems to move hundreds undead chaff when all leaders don't have undead leadership).

ME Ermor is actually one of the nations it plays correctly (because of this and the d+s mages fulfilling the evocation condition, the AI mages will spam nether bolt/dart behind undead hordes once researched).

[Sombre]

The immobile mages + huge gem income wouldn't result in a balanced nation that players could make use of though. It would probably be a good way to boost the AI if that was the objective though. I know giving EA Mictlan a starting income of 50 blood slaves a turn boosted them a lot, but of course it isn't a nation players could use.

[NTJedi]

So are there any traits/nations you've noticed that suit the AI well?

When a computer opponent is configured for 'Defensive' the AI does seem to invest more wisely in province defense.

From my MP and SP games which included AI opponents I've noticed it's good playing early age Niefielheim. IF this AI opponent can survive until late game its mages will be casting battlefield buffs such as: Mass Protection, Mass Regeneration, Anti-Magic, Rush of Strength, etc., .

Unfortunately the AI lacks knowledge for equipping magic items. Even if it only knew how to equip magic boosters its battlefield and ritual spells would be greater.

The AI also has a huge weakness for sending its pretender into the arena death match. Any AI opponent should be banned from sending its pretender into the arena death match via the pretender tag.

Sombre said:

AI can't really use blood properly.

Unfortunately true, within DOM_2 it was possible to make them use blood efficiently via mapedit commands. The method was to provide blood income for a non-bloood nation, provide them immobile blood casters, then use the #startspell to specify which spells they can cast. Unfortunately #startspell is currently broken even tho it's still listed in the mapedit PDF.

The AI also isn't aware which ritual spells are a good investment and which ritual spells are a bad investment. One example is the AI opponent will still be summoning cave drakes despite it having access to mechanical men or living statues.

One problem I'm facing is that it doesn't seem possible to provide a commander/mage with only magic leadership or only undead leadership as it needs at least normal leadership of 10.

----

There are lots of AI improvements which could be done, yet it's most likely too much work for Illwinter. Thus the reason modding for the AI turn should be made available via scripting commands. Yet even this task may be too much.

[Twan]

The immobile mages + huge gem income wouldn't result in a balanced nation that players could make use of though.

Why not ? Imagine a nation with an immobile capitol mage with the level to cast an offensive ritual (say F3 for fires from afar or A3 for seeking arrow) and a 15 one color only gems site, but handicaped in both magical diversity (2 elemental paths, the hardest to boost, fire and air) and quality of mages able to move (say F1A1 + 100% FA + 10% FA ; so a mage can always use fire or air evocations, sometimes the two, but never forge boosters).

Even if troops are good and all national leaders get some supplybonus to help the AI, it would be a rather weak nation for a player (the need to empower mages compensate the huge income, and the lack of magical diversity the research advantage fire gems can offer). But as the AI doesn't forge boosters even when possible, and like to waste gems on offensive rituals and empowerements even when it doesn't have this increased income, it's all good for its playstyle.

[Karlem]

As far as my experience tell me the best challeging AI nations are Late Age Emor and Ryleh. Freespawn rule.

Apart from that as the AI forgets about any middle/late game strategy (they seem to research random stuff, not have a clear research strategy and such based on national mages) they will never be able to compete past turn 30 unless on hard research that just moves this limit artificially higher.

[Sombre]

Twan: Fair point. It might still feel like an AI nation though, when you played it. That's something to avoid - the ideal result is a nation much like the existing dom3 ones, but that has been tweaked and altered to hopefully suit the AI well. With careful planning your idea could be good though. Are you sure the AI empowers though?



I understand that the AI has plenty of problems and can't handle the lategame at all, or certain SP tricks,.. but the thread isn't really about that. It's about working with the current AI based on our knowledge of it rather than suggesting changes to it.

LA Ermor and Rlyeh do well with the AI, that's true. Especially with no indies via map or mod. I think LA Ermor might have a special AI though.

Why do we think the AI does well with EA Niefel? It can't use thugs or a bless strat, so it would seem odd it is successful with Niefel, yet it does appear to be. It consistently does well with Jotunheim too.

[Ewierl]

Sombre said:LA Ermor and Rlyeh do well with the AI, that's true. Especially with no indies via map or mod. I think LA Ermor might have a special AI though.

I know in SP games, the issue seems to be less "AI does well as Ermor" and more "AI does very badly fighting against Ermor." Therefore, the AI Ermor consumes all its neighbors easily and grows to enormous size.

Why do we think the AI does well with EA Niefel? It can't use thugs or a bless strat, so it would seem odd it is successful with Niefel, yet it does appear to be. It consistently does well with Jotunheim too.

AI does well with nations that have big tough troops. Powerful troops like the Jotuns allow a nation to be effective without having to rely on blesses/summons/crazy tactics/crazy combos that the AI does less well. Remember that a higher difficulty setting gives gold bonuses to the AI, which skews its effectiveness even further towards Massive Army Power.

[Sombre]

Yeah but jotuns aren't actually very good base troops. I mean they aren't bad either, but I'd think ermorian or pythian infantry etc would be much more effective.

[sector24]

The reason it can't play Pythium is due to the gladiators. It recruits 800g worth of units that can only be used in a single fight. I think if you disabled them Pythium would play a lot better.

I think the AI is good at both fire and death spells. They can skelly spam, and shadow blast, plus use generic fire evocations. They're also decent at air magic. Possibly a nation of N1D1?2 and N1F1?2 mages with ANDF as the random picks, plus low gold mid-resource units that are very tough for their cost (like Alae Legionnaires perhaps).

More importantly, the AI needs to NOT have a lot of things. No gladiators, no horn blowers or standard bearers, no 120 leadership useless commanders that cost a lot, etc. Ideally they'd have a mix of all useful troops where the only difference is a high damage 2-hander, a high damage 1-hander with a shield, or a mid-range 1-hander with a shield. Lots of survival types would help, but a high enough resource cost should prevent the AI from putting 500 of them in the same army and starving to death.

A human player playing the nation would be very powerful with the Air/Fire combo, but the lack of astral would be a big offset. That's just where my line of thinking took me, maybe you have a different idea.