Age of Strategy Forums

I opened this topic as i have not found a "AI is dumb" topic here, please help me if there are any and i will merge it.

No i changed these, and i will post here my further findings as i am testing the changes:

1. Shallow water problem: we introduced the "shallow water" some time ago, and ALL "best against me" logic checked if the target can walk on water or not, and if so only the ships were checked to be built (as every unit was able to walk on shallow water)... so LITERALLY only ship production worked earlier - so ALL land units were working bad...
now i did a quick fix: now i first check the possible trainable land units and if no units to train than then the water units too, regardless of the target unit walkability
but i am afraid if there are land units set against ships than ONLY land units will be produced agaisnt ships in TC-s (eg. archers agaisnt ships not anti-ship ships)
(as probably in a TC-s there are land units are trainable too that are also in the bestagainstme list)

and the land i checked first as in code i dont know which unit is ship or not anymore
(since we have a land tile "shallow water" i can not check their walkability anymore.)

2. "Healthy army ratio": (this logic is limiting the unit types to build egarless of task needs, it will not build 150 workers as it is set only about max. 10% can be workers) in this logic was hardcoded that >20 power units are considered "siege" - so no attackers were created in AOF as in (the hardcoded) "healthy army ratio" said only a few siege units is permitted - so since AOF got upscaled (meaning many units have more then 20 power) this started not working.
3. Enemy units walking ranges: the TC occupier units were wrongly calculated - instead of their speed their ID was used that is in AOF a big value, so AI always made TC protecting units instead of anything else because it though all enemy units can walk about 200 tiles a turn.
4. Neutrals should have less importance for productions: i set on "spell task" productions too that neutral buildings not that important to kill than any other (same for "attacking task" productions)
5. "Defend TC" tasks have less importance for productions: I set that while for units the top prior is the "Defend TC" task (when an enemy is close to a TC of the AI) for factories it will not be the top importance, now goes behind the "attack enemy" tasks -> this way not only cheap tc defending units will be created but (as on normal attack tasks) good "anti" units agaisnt the enemy army.
6. productions "consume" tasks: i dont know why but prodcutions were not consuming the tasks, as a result on a case if a task to eg. "heal" a unit that was in most importance, than in ALL factories healers were created, as the first factory did not get a sign on the task that is was "taken" so next factory also used that task.


and maybe several smaller other changes.

also ALL of these changes will be in ALL ALTERNATIVES!

(bad news for me i will need to make the same changes in unity version too - as all AI is already ported to unity.)

New finding: i found it in AOF: the orc pit does not invent upgrade techs.

reason: the current logic expects that the unit level up tect is in same building as the unit training itself so it work now like this:
- factory receives a task to "attak an enemy" (read: train a unit to attack an enemy)
- tries to find best unit against enemy (that the building can build)
- and checks if that unit can be upgraded by a tech or not (have missing upgrades)
- and if so than checks if that building can build that upgrade
- and if so, first check if the factory is far from warzone and if so it invents the tech


so as a result orc pit does NOTHING with such task as it has no any attacking unit it can produce. -> resulting orc pit produces NOITHING for several turns!

(remark: for test i placed the upgrade techs to default production for orcs - and that case it starts trainaing the level up techs immediately in orc pit - but this is a "random" tech training this way - not basedon enemy "army" type to train/upgrade against that.
)

any idea what to change would be appreciated.

ANSWER: i checked in code and originally "blacksmith" techs were also handled this way (to set as "deafult" production), so in AOF we should set all blacksmith techs (and upgrade techs that are in separate buildings) to be default techs.

(now i start setting these in AOF)

diuscussion from DC, about the new "default production" handling - mostly on blacksmith like buildings (where there are no normal units can be trained so it will result in no production atm)

stratego — Today at 4:? PM
what about @Sunrise Samurai idea: to check existing units of AI and invent accordingly. would that be better? (eg. for hard AI)
only one case is not good with that on game start, there are only the starter units so the first blacksmith tech will be for the starter unit (worker, scout , archer) depending on starter units as AI has only those.
or shall this check the running prodcutions units too?
ӍαƙȺʑμωƦ³² ᓚᘏᗢ — Today at 4:32 PM
Elves:
Always must research weapon tech first.
Since weaponsmithing of elves affects attack value of all elves specifically — both melee and ranged.
stratego — Today at 4:34 PM
ok weapon techs first, can be set by the default productions setting.
question after that should it be set like you wrote above (by weights), or you also think it would be better to count the AI units and upgrade existing units? (eg. hard AI case)
...
ӍαƙȺʑμωƦ³² ᓚᘏᗢ — Today at 4:38 PM
I would love to see for ai to count units and upgrade them first.
Since for humans it is critical to upgrade while being much more complex.
For elves it is not required since weights above are enough to upgrade elves properly.
For orcs it has low priority.
For undeads and scalefolks it is not required. Weights system is enough.

For dwarves currently it is required but later due to planned overhaul for dwarves to make them a bit less op but also more unique and fascinating it will require changes.
stratego — Today at 4:42 PM
ok than i try making this:
step 1. i set the most important ones as default productions (that is always to be invented first)
step 2. if a building has no more defaults to train (that is doing nothing atm) than
a) hard AI case: i make a NEW logic: upgrade(upgrade techs) or buff(smith techs) for the existing units
a) easy AI case: does nothing here in step #2
step 3. if still has no trainable unit, than a totally random production set that the building can still train (any tech or unit it can train)
-----------
wait you say ONLY humans would need that new logic? i will defintely not implement it than
ӍαƙȺʑμωƦ³² ᓚᘏᗢ — Today at 4:43 PM
Orcs also need it.
stratego — Today at 4:43 PM
u said low prior
ӍαƙȺʑμωƦ³² ᓚᘏᗢ — Today at 4:43 PM
Dwarves will really need it later, even more than humans.
After changes we planned for them.
stratego — Today at 4:44 PM
ok, than i will check how much would that be and if not that tricky / hard, than i make it. is the above #1..#3 step sequence good?
ӍαƙȺʑμωƦ³² ᓚᘏᗢ — Today at 4:46 PM
Probably yes.

ok, aupdated version is uos 5

(in both AOS and AOF)

in this the (only the HARD) AI will figure out what tech to invent in blacksmith like buildings (unless it is defined in "default production" lists - that is stronger in priority)

First - let's clarify one thing - Currently in published version engine works like:
- unit (building) gets assigned a task
- checks buildist.json for production role related to the task taking into account race and terrain limitations
- if found corresponding role, randomly and according to weights assigns production unit from the list, but taking into account only those, that can be produced by it
- if not found, looks through DEFAULT_PRODUCTION, but taking into account only those, that can be produced by it
- if not found again, does no production

- If any unit was found in specific list or default one - there are those steps:

- and checks if that unit can be upgraded by a tech or not (have missing upgrades)
- and if so than checks if that building can build that upgrade
- and if so, first check if the factory is far from warzone and if so it invents the tech

Question:
1. Where exactly is bestAgainstMe, bestAgainstMeWater considered?
2. I have STANDARD_ARMY_SETUP_MIN_VALUES_BY_CATEGORIES and STANDARD_ARMY_SETUP_MAX_VALUES_BY_CATEGORIES - do they work?
If so when they are checked?


And the idea is to change it to how it works in the dev something like:
- steps cited (about upgrades) are not executed
- instead there would be another step before determining no production:

a) hard AI case: i make a NEW logic: upgrade(upgrade techs) or buff(smith techs) for the existing units
a) easy AI case: does nothing here in step #2
step 3. if still has no trainable unit, than a totally random production set that the building can still train (any tech or unit it can train)

So step 2 would be executed if building can research any tech, that affects units, that exist on map and step 3 otherwise?

As the steps 2 and 3 here would be are 100% on engine side I see a lot of potential problems:
- what would be the order of techs for e.g. AoF humans?
- what if some race (in any variant) depends on upgrades first (are stronger, that buff techs), but there is a lot of them and only selected few give much bigger advantage?

If bringing something like that - I'd make some way to let designers deal with exact parameters (or even if you had to tinker it yourself - without rebuilding at least).
So e.g. another productionRole like "DEFAULT_RESEARCH" in which weights would be multiplied by (1+number of units in unitList) on map?
To avoid confusion there could be another parameter in buildPriors "weightNumberMultiplier" or something, which is checked only for DEFAULT_RESEARCH (or something ascended from buildPriors to avoid another possible mixup).
As techs (including upgrade techs) can have categories - there would be a way to determine the order very precisely.

But also - could there be some way to affect priorities if techs would be in bestAgainstMe ?

1. Where exactly is bestAgainstMe, bestAgainstMeWater considered?

only on ATTACK tasks, for looking for a counter unit.

I have STANDARD_ARMY_SETUP_MIN_VALUES_BY_CATEGORIES and STANDARD_ARMY_SETUP_MAX_VALUES_BY_CATEGORIES - do they work?
If so when they are checked?

i saw the hardcoded working - i need to check if these work or not. (i have not remembered and i have not seen thee in code - maybe a not ready implementation? needs check.)

So step 2 would be executed if building can research any tech, that affects units, that exist on map

yes. ("exist on map " = "exist on map for the given AI")

and step 3 otherwise

no, step 3 will only follow if previous stpes did not give any unit/tech type to produce.

what would be the order of techs for e.g. AoF humans?

what do u mean?

what if some race (in any variant) depends on upgrades first (are stronger, that buff techs), but there is a lot of them and only selected few give much bigger advantage?

you mean if a variant dont want the AI to invent techs - or rarely?
dont forget: this new logic runs ONLY of that building would not train ANYTHIING (so it had no unitType to build based on its tasks - so a "spare time inventing a tech" is the case.

DEFAULT_RESEARCH

you have it already as if you set default production for techs it will be chjose before all we talk here.

so the total list of steps
1. get production based on tasks (most important tasks first) - if Attack task it is possible to get a unit upgrade for the same unit if that is trained in same factory.
2. IF NO any production from these (eg. a blaksmith building that trains nothing else than techs) than check the default techs (if there any tech set will be selected to produce)
3. IF NO any production from these than will try to figure out a upgrade or a buff (prop changer) tech based on the AI player actual Army
4. IF still NO any production from these it will select a random producable anything
(in all above cases only HARD AI will invent upgrade and prop changer techs, easy AI will not go that way)

i think that is the total list.

But also - could there be some way to affect priorities if techs would be in bestAgainstMe ?

i am not sure i undertand give me example please.

Stratego (dev) wrote: ↑Thu Nov 03, 2022 8:36 pm

what would be the order of techs for e.g. AoF humans?

what do u mean?

I mean on this step:

3. IF NO any production from these than will try to figure out a upgrade or a buff (prop changer) tech based on the AI player actual Army

If engine determined human swordsman to be most common unit, then what would be the order of techs researched: armors lv1, melee weapons lv 1, To Expert infantry, Shields lv1, Magic Armor lv1 and armors lv2, melee weapons lv 2, To Veteran infantry, Shields lv2, Magic armor lv2 (I may have skipped something, but only the idea is important here)

what if some race (in any variant) depends on upgrades first (are stronger, that buff techs), but there is a lot of them and only selected few give much bigger advantage?

you mean if a variant dont want the AI to invent techs - or rarely?
dont forget: this new logic runs ONLY of that building would not train ANYTHIING (so it had no unitType to build based on its tasks - so a "spare time inventing a tech" is the case.

It's not about stopping or making it less frequent (that is something reserved for weaker AI levels).
But to even utilise the logic here - we have to drop some techs from default production.
And apparently there were a lot of cases when blacksmith-like building (no units in any counters or defaults) have nothing to produce, so what if order of things would matter here greatly?

DEFAULT_RESEARCH

you have it already as if you set default production for techs it will be chjose before all we talk here.

That was the idea to make additional weights only used in the case when tech was selected after having nothing in default production (so taking into account army composition).

But also - could there be some way to affect priorities if techs would be in bestAgainstMe ?

i am not sure i undertand give me example please.

E.g. AoS new tech "Cultural Promotion" is meaningful only as a counter to enemy converters (and Onmyoji), but in case there is a lot of enemy converters it becomes very good.
So I'd set it to somehow be used as a counter to those (after already having some counter units).

If engine determined human swordsman to be most common unit, then what would be the order of techs researched: armors lv1, melee weapons lv 1, To Expert infantry, Shields lv1, Magic Armor lv1 and armors lv2, melee weapons lv 2, To Veteran infantry, Shields lv2, Magic armor lv2 (I may have skipped something, but only the idea is important here)

counterd all units and all of their upgrade or prop changer effects and all with the number they are used on units.
so if ONLY swordsmen are there than random which tech will be.
BUT
if there are archers too than the common eg. "armors lv1" will be chosen as more units need that.

It's not about stopping or making it less frequent (that is something reserved for weaker AI levels).
But to even utilise the logic here - we have to drop some techs from default production.
And apparently there were a lot of cases when blacksmith-like building (no units in any counters or defaults) have nothing to produce, so what if order of things would matter here greatly?

you can always put he "important" ones to be in the deafult production list - so those will be produced first.
(not sure i understood and answered properly.)

That was the idea to make additional weights only used in the case when tech was selected after having nothing in default production (so taking into account army composition).

i see so you want to influence order to chose when we are in auto determinning based on army case? i see but i feel not that impoartant which tech is first. but i can be wrong.

E.g. AoS new tech "Cultural Promotion" is meaningful only as a counter to enemy converters (and Onmyoji), but in case there is a lot of enemy converters it becomes very good.
So I'd set it to somehow be used as a counter to those (after already having some counter units).

if it is a "prop changer" tech than it will be among other prop changer techs to produce - if it is in a building that has no units in it to produce (!) - (so in early steps it did not give anything to produce it will be a possible invention)
i emphasize again: instead of the earlier logic that produced NOTHING this case.


also dont forget we dont want AI to be TOO clever - i already receove tons of complaints about too hard AI.

you can always put he "important" ones to be in the deafult production list - so those will be produced first.
(not sure i understood and answered properly.)

My point was exactly in case you cannot put it into default list, as e.g. it's useless without units affected and e.g. those affected units are only specific counters.

i see so you want to influence order to chose when we are in auto determinning based on army case?

Exactly.

if it is a "prop changer" tech than it will be among other prop changer techs to produce - if it is in a building that has no units in it to produce (!) - (so in early steps it did not give anything to produce it will be a possible invention)
i emphasize again: instead of the earlier logic that produced NOTHING this case.

I know it's better than nothing.
But it's gonna make AI weaker and weaker with more techs coming in without any priorities.

also dont forget we dont want AI to be TOO clever - i already receove tons of complaints about too hard AI.

I honestly doubt anybody is really nagging about AI being too clever.
Those tons of complains are mostly by novice players, which either or both:
- don't understand the mechanics yet
- don't have yet enough upgrades to use any meaningful strategy or lack counters against AI
- think it's unfair, that AI starts with some boosts (already researched techs)
- plays too hard campaigns (a lot of even easy ones are not very forgiving - one mistake is often enough to have to restart the map)

The second case is imho most important for new players, which I pointed repeatedly.
It's not very hard to see, that player with access to basic units (and e.g. few culture-specific ones) won't stand any chance against upgraded, blacksmith boosted army.
And also - the order of upgrades in shop is messy, which could make it harder to get most core upgrades.
As already mentioned in the forum somewhere - either total reordering or even additionally giving more important upgrades on the start.

And on the other hand - veteran players outplay AI like they want, so some of them might think AI is too stupid to even bother.

My point was exactly in case you cannot put it into default list, as e.g. it's useless without units affected and e.g. those affected units are only specific counters.

ok, i see.
the order could be i think the appearance order of the techs - is that good? (i mean the buildlist order)

And also - the order of upgrades in shop is messy, which could make it harder to get most core upgrades.

do u have any suggestion there? (dont write the whole plan but the "vision" if u have)

As already mentioned in the forum somewhere - either total reordering

if this is the same as provious point than i answered there.

or even additionally giving more important upgrades on the start.

we can, which u think should be?

Stratego (dev) wrote: ↑Fri Nov 04, 2022 12:55 pm

My point was exactly in case you cannot put it into default list, as e.g. it's useless without units affected and e.g. those affected units are only specific counters.

ok, i see.
the order could be i think the appearance order of the techs - is that good? (i mean the buildlist order)

Not sure if it's corresponding to importance - e.g. in AoS I tried to just put them close to similar ones and beside that - are mostly in order of introduction.
I can imagine, that right now the target of buildlist sorting is to present units in such way, that desired one can be found as easy as possible.
So it's not really in perfect accordance to the idea of how important they are.
But the same newly created list could be responsible for ordering shopitems in idea below.
Units missing in this proposed "importance list" could be all random on the bottom (after all in list).

And also - the order of upgrades in shop is messy, which could make it harder to get most core upgrades.

do u have any suggestion there? (dont write the whole plan but the "vision" if u have)

Main idea:
Upgrade shop as a list (similar to current unit help).
Player can filter by categories, race parameter.
Greying out upgrades unlocking units, that require any units with not-bought upgrades.
Sorted by some design leader given values.
With indicator for upgrade shop items unlocking more than one unit.

Quick solutions:
- making shop stop hiding upgrades in further rows
- rearrangement of shop items (well, quick code-wise, but not really total work-wise)

or even additionally giving more important upgrades on the start.

we can, which u think should be?

There is a topic, that already mentions some ideas:
http://www.ageofstrategy.net/viewtopic.php?f=6&t=12612

So it's not really in perfect accordance to the idea of how important they are.

ok, where should i place a list for that? buildlist json is good?

Upgrade shop as a list (similar to current unit help).
Player can filter by categories, race parameter.

this is not that i would like to, i like the mystery in that.

Greying out upgrades unlocking units, that require any units with not-bought upgrades.

i dont understand - can u tell me example?

Sorted by some design leader given values.

how do u mean?

With indicator for upgrade shop items unlocking more than one unit.

and example on this?

- making shop stop hiding upgrades in further rows

i want hiding mystery.

- rearrangement of shop items (well, quick code-wise, but not really total work-wise)

reordering can be naturally.

Stratego (dev) wrote: ↑Fri Nov 04, 2022 7:53 pm

So it's not really in perfect accordance to the idea of how important they are.

ok, where should i place a list for that? buildlist json is good?

Yeah, I think it's just the file for that, as other AI related settings are there.

Upgrade shop as a list (similar to current unit help).
Player can filter by categories, race parameter.

this is not that i would like to, i like the mystery in that.

- making shop stop hiding upgrades in further rows

i want hiding mystery.

OK.

Sorted by some design leader given values.

how do u mean?

It doesn't matter as it was related to list-like look.

Greying out upgrades unlocking units, that require any units with not-bought upgrades.

i dont understand - can u tell me example?

If none of builder has upgrade bought OR any of requirements have not bought upgrade.
For example - cavalry armors lv 1 could be greyed out until you have blacksmith (or any other visual indicator). Additionally on clicking it - different dialog could be shown, that reminds about missing requirements to use it.
It could still be bought, just to get to another shopitem below.

I know items can be placed under required one to solve it in some cases, but there is often not enough rows for all related.

With indicator for upgrade shop items unlocking more than one unit.

and example on this?

For example Fervor is bundled with Cathedral, so there could be some indicator like: Additionally "More..." button could cause another dialog choosing unit to show.
Or maybe dialog should be on initial tap?

I know items can be placed under required one to solve it in some cases, but there is often not enough rows for all related.

yes i wanted to suggest making them in one column statrting with the building.

is there any case when 5 is not enough?

A lot.
And there will be more.

I have been planning for some time already to set some common techs to decrease upgrade time.
Something akin to AoF, but on smaller scale.
Like there is few similar units, which all have 4-6 upgrade cost, so instead there would be common 3 turn tech and their upgrades would be 1-3 cost.

Endru1241 wrote: ↑Fri Nov 04, 2022 9:54 pm A lot.
And there will be more.

I have been planning for some time already to set some common techs to decrease upgrade time.
Something akin to AoF, but on smaller scale.
Like there is few similar units, which all have 4-6 upgrade cost, so instead there would be common 3 turn tech and their upgrades would be 1-3 cost.

we would need some better solution than, this is a little awky.

btw: are these many techs related to one building?
eg. blacksmith? maybe we can make blacksmith being free? -> that is another workaround to the problem.

Stratego (dev) wrote: ↑Sat Nov 12, 2022 6:21 pm

Endru1241 wrote: ↑Fri Nov 04, 2022 9:54 pm A lot.
And there will be more.

I have been planning for some time already to set some common techs to decrease upgrade time.
Something akin to AoF, but on smaller scale.
Like there is few similar units, which all have 4-6 upgrade cost, so instead there would be common 3 turn tech and their upgrades would be 1-3 cost.

we would need some better solution than, this is a little awky.

btw: are these many techs related to one building?
eg. blacksmith? maybe we can make blacksmith being free? -> that is another workaround to the problem.

I already started to think about that.
It would still leave the problem of many other buildings.
And quite a few techs.
Commonly new culture would never have only 5 total of units/techs. Always more.
It also applies to other variants even more. E.g. all subrace units in AoF.


But still - some units, techs and buildings should be made free anyway.
My propositions is :
Archery, Stables, Siege workshop, Blacksmith, Dock
Foot Knight, Light cavalry, Monk, Ram
Ranged Reach lv1 and Ranged Damage lv1 (it's actually one shopitem), Melee Damage lv1, Archer Armors lv1, Infantry Armors lv1, Cavalry Armors lv1
It's also with having "no upgrades" games in mind, but having new player being forced to choose specific upgrades or end with only infantry to train from factories is very limiting. First level of blacksmith techs is for having something in blacksmith building.

The question is - is it possible to recompensate players, that had specific shopitem id already bought?
Along with never recompensated Trebuchet (cost 8) it's a total of 71 gems.
If not, then I think I'd have to bring equal number of new things and assign them to old ids.

Endru1241 wrote: ↑Sat Nov 12, 2022 7:06 pm But still - some units, techs and buildings should be made free anyway.
My propositions is :
Archery, Stables, Siege workshop, Blacksmith, Dock
Foot Knight, Light cavalry, Monk, Ram
Ranged Reach lv1 and Ranged Damage lv1 (it's actually one shopitem), Melee Damage lv1, Archer Armors lv1, Infantry Armors lv1, Cavalry Armors lv1
It's also with having "no upgrades" games in mind, but having new player being forced to choose specific upgrades or end with only infantry to train from factories is very limiting. First level of blacksmith techs is for having something in blacksmith building.

I agree with this. It would also teach newer players how to use blacksmith a lot sooner. It's somewhat common for me to see new players not bothering with blacksmith techs for many matches and leaving their units a lot weaker due to that.

Archery, Stables, Siege workshop, Blacksmith, Dock
Foot Knight, Light cavalry, Monk, Ram

these surely can be (actually i thought most of these are already free lol )

Ranged Reach lv1 and Ranged Damage lv1 (it's actually one shopitem), Melee Damage lv1, Archer Armors lv1, Infantry Armors lv1, Cavalry Armors lv1

these can also be - but not sure needed but i can accept.

is it possible to recompensate players, that had specific shopitem id already bought?

we can do it that way we bring in new units/techs to shop and we use these existing ID-s.


so (if i counted right) we need 9 new techs/units for first pack.
and if we want the techs too we need extra 5.
and i jut saw you said trebuchet had an id also - so with this +1 extra
in total 15... quite many ...

i see we have a few new ones already in latest update (that is not yet published)
so we can use old ID-s on those as i count we have 5:
- Tlatoani
- Taxiarchos
- Noyan
- Cultural Assimilation
- Cultural Promotion