AI

[Seravy]

I'm wondering if it might be worth raising the turn 10 timing of moving into the city to rampage to turn 12? I've seen the AI turn around right before catching a defending unit a bit too often. Unfortunately this would slightly reduce the damage done to the city, but raise the chance of catching all defenders and conquering the city if a good doomstack is the attacker. Which also eliminates the chance to then spam spells on the doomstack and destroy half of it.

[Nelphine]

I'm ok with that kind of tweak.

[Seravy]

One thing I noticed last game but seems even more obvious in the current one :
The AI's decision to change to a better race for settlers sometimes really hurts the AI.

Don't remember what was for the last game (switching lizardmen for nomads?), this one the lizardmen switched to spreading high men and ended up with only 5 cities in 1405, half the continent unsettled. A few of the early lizardmen settlers went to my continent (and got killed/conqeured), and afterwards, only high men were made it seems.

So I was thinking something should be done about this. Either a turn limit (don't change to another race too early) or making the choice have a different table for each race, or special-case only lizardmen. Not sure which would be the best?

At the very least, lizardmen switching to high men (+4 unrest instead of -1, and slower expansion) shouldn't happen.

[Nelphine]

I'd put a table, with a turn limit, customized by race, wizard picks, wizard personality and difficulty!

Since that might be a little too complicated, any tweaks are worthwhile 

Although thinking about it, I think personality would actually be a really good modifier. 

Also personality influencing starting race (or vice versa) would also be very cool.

[Seravy]

Personality already influences starting race. (to some extent, only the worst combinations are disallowed. Like Peaceful with Barbarians.)
Later on, I think the main reason why a race should or should not be picked is unrest (as ignoring that, picking the late races is always the correct move, except for early game lizardmen maybe.)

I'll try to design a table for this then.

[Nelphine]

I disagree. Unrest only leads to lost gold and production. If you can get more gold or build better things faster, unrest doesn't matter. GENERALLY that means unrest matters. But a sole nomad city might create more gold than a sole klackon city on an island. A dwarf city on the right ores certainly gets more gold than troll or even orc cities.

Building halflings as mono life is worth far more than building any other city units, so production wise, unrest may not matter. Although halflings also get low unrest. Building dark elves against magic immunity is pretty silly (so if human has many sorcery books, trolls or dwarves may be far better.) Etc.

[Seravy]

Well, yeah but most of those aren't based on your existing race so it wouldn't fit into the table.
On the other hand things like halflings producing more research so building dwarven cities that make no magic is bad for them, works.
I think I'll also add a separate, modifier table which adds priority based on realm played by the AI. Perhaps we can add one more for the human's realm.
That should cover pretty much everything the AI is capable of understanding.

[Seravy]

Designing the table isn't as easy as I thought. This might even take more than an hour.

[Seravy]

I'm actually stuck on the lizardmen. On one side, producing late game settlers slows down expansion a lot. On the other side, if the AI does not do that, their late game is bad - lizardmen economy is medicore and their units are not that great for late game either. Problem is, the AI can't guess whether the human will attack them early, or not.
(Klackons are a similar problem except there it's the easy to counter klackon cities vs anything else that has tons of rebels and slows down the economy a lot until much later.)

[Seravy]

   

Done. First, white line is the unrest table modifier. Second, colored line is the settler priority (highest will be built).
Modifiers for realms added if at least 5 books of that realm on that player exists. If multiple realms match, all of them are added.

I'm a bit worried because the table does change the default order on quite a few races, but the realm modifiers should avoid the worst case situations (like building halflings instead of other races then running into fire storm or invisibility).

Already spotted a possible improvement, barbarians vs high elves and dwarves should be 7. Beastmen are better than both, they have good units and good magic. In fact, dwarves maybe should be lower? Combining the two races without magic sounds like a bad plan although high resistance on dwarves is very useful for the barbarian player.

I'll implement this feature tomorrow, filling this table was exhausting.