AI farmers

[Seravy]

Currently the AI sets the amount of farmers to match the amount of food needed pretty much randomly. (older cities will usually have the most farmers but that's about it)
It also ignores famine, so if it happens to pick that city for increasing food output, it might not make enough.

I believe this procedure could use a rewriting (again, as it's already new, the original could not reduce the number of farmers at all).

So I wonder, what factors should the AI consider when deciding on where to make food and more importantly how to weight them.

1. Amount of food produced by the farmer - this one is obvious, if you get 3 food for 1 farmer it's better than 2 or 1.
2. Amount of production lost - 1.5 * bonuses for most races, but 2.5 for klackons and dwarves
3. Remaining production of the town - going down from 20 to 15 production is much better than from 7 to 2!
4. Special ores - Iron and coal, production is worth more in these cities
5. Availability of military units - it's better to farm where the AI can't build strong units - but how do we define "strong unit"? Is there a difference between having only one type available, or 3 types?
...anything else I missed?

This doesn't look easy to do so I'll leave it for a future update.

[GermanJoey]

Hmmm, sounds really good.

#1 & #2 seem right.

#3 I'm not so sure about... it's important to jump-start new cities, but the AI can cash-rush so easily that the little bit of production a tiny city has is sorta insignificant. It might be better to set those cities to farmers so that the cities able to produce high-quality units can focus on that.

#4 - in addition to iron and coal, cities with adamantium, orichalum, or mithril should also avoid farmers.

#5 - a crude, but perhaps reasonable estimate of "strong unit" would be to simply use the most expensive unit, in terms of base hammers (not including discounts from coal). not perfect but, not bad. perhaps the availability of adamant/ori/mithril could add a multiplier to this number?

Also, speaking of farmers - if the AI needs to make judgments on farmers, and also finds it worthwhile to spam Change Terrain, perhaps it would make sense to up the priority for building granaries and farmers markets? Those buildings do so much to speed growth.

[Seravy]

Started working on this, considering to use the following numbers, city with the highest total gets 1 farmer added :

; +10 priority per food produced
; -3 per lost production points
; -5 if Iron, -10 if Coal, -10 if Mithril, -20 if Adamant
; -5 if High Men or Nomad
; +5 if Barbarian, Gnoll after turn 50
; +5 if Halfling
; -5 if Armorer's Guild available
; -5 if Fantastic Stables available
; -2 if sea present (ship production)

[Domon]

You should probably add a rule to promote doing this in cities where the REMAINING production doesn't become too low? Like, going from 40 to 33 is a thing, going from 10 to 3 another thing entirely.

[Seravy]

You should probably add a rule to promote doing this in cities where the REMAINING production doesn't become too low? Like, going from 40 to 33 is a thing, going from 10 to 3 another thing entirely.

Yes, I wanted that, then realized, the AI will always build a Sawmill first if there isn't one, and will always rush-buy it immediately if able, so every city will have at least ~10-15 production even without any workers, which is probably good enough, or more like I have no idea how much minimal production would be ideal if the sawmill is not enough.

On the other hand, maybe boost priority if the city is producing Trade Goods and reduce if producing Housing?
...nope this one can't be done, production for the turn is decided after farmers, not before.

[GermanJoey]

That might be backwards... I would think that cities with lots of farmers should be the ones assigned trade goods, rather than the other way around? Unless there's something that forces big cities with good resources to keep producing military?

[Seravy]

That might be backwards... I would think that cities with lots of farmers should be the ones assigned trade goods, rather than the other way around? Unless there's something that forces big cities with good resources to keep producing military?

at the moment, overland production, including trade goods is based on some simple rules to adjust weights, their objective, and random chance. It would require rewriting the whole procedure to improve it any further.

[GermanJoey]

Hmmm, might be too easy then for farmers to get stuck on cities that want to produce military, unless all farmers in all cities are reassigned fresh each turn.

[Seravy]

Hmmm, might be too easy then for farmers to get stuck on cities that want to produce military, unless all farmers in all cities are reassigned fresh each turn.

They are.

[Seravy]

This is the set of rules I ended up with :

; +10 priority per food produced
; -3 per lost production point
; -1 for each 2% of unit cost reduction from ores
; -10 if Mithril, -20 if Adamant
; -5 if High Men, High Elf, Nomad, Dwarf, Beastmen, Troll
; -10 if Dark Elf, Draconian

; +5 if Barbarian, Gnoll after turn 100
; +5 if Halfling
; -4 if Armorer's Guild or Fantastic Stables available
; -4 if War College available
; -2 if sea present

Orihalcon is ignored because it's hard to detect and only works well for magical ranged units which is a minority of unit production.
Evaluation s done individually for each farmer, so if there are 10 identical cities that can have 1 farmer producing 3 food and all others only 2, then it'll set 1 farmer in each city if 30 food is needed. However if 1.5 food is produced per farmer, it is treated as 1 or 2 based on which way it is rounded for that specific farmer, allowing the AI to pick the city where no 0.5 food gets lost from rounding first. The price for that is, if all cities round down with their next farmer, they don't have a better priority than a city that always gets 1 food per farmer. Same is true for cities that produce other fractional amounts of food, like Famine. Furthermore, even if a farmer would produce zero food, it can get chosen if no farmer with a higher priority exists, allowing the AI to use farmers that produce less than 1 food (due to famine) as a last resort. Finally, if everything else is equal, the oldest city is selected.