Build never, build often

[Steph]

Anyone knows the exact quantifiabl effect?

Normally, with "never build naval units", the AI should never build any sea unit.

But what is the exact effect of build often vs nohing?

Let say I have a swordsman offensive unit, a spearman defensive unit, and a galley naval unit.

What would the AI build if I have for instnace
- built often offensive
- nothing for defensive

or
- built often offesnive
- built often defensive

or
- nothing for offensive
- nothing fo defensive

 

[Bowsling]

I'd assume the AI would go with galleys if offensive and defensive were build never. In the case of both being build often, I'd say about 80-90% of the cities would be building spearmen and swordsmen, but there are also probably a couple of coastal cities building galleys. For often offensive never defensive, those 80-90% of cities would be building swordsmen.

That's just what I'm assuming, I've never tested it.

 

[Steph]

I'd like to have something more precise.

Like "build often = +75% chances" to build this type of unit.

 

[tom2050]

I am 90% pretty sure that Build Never does not apply to the very first turn. Whatever the building or unit is set at to build when game starts, Build Never is not taken into account, so an AI city may build it anyway, but then will not for the rest of the game.

If it is set for AI to Build Nothing for everything, will game crash? Will AI actually build nothing?

I can do some AI tests with just the build often flags for difference, and only 2 units, 1 offensive warrior and 1 defensive warrior. I should keep their stats the same to try to get a bottom line, perhaps just use 1.1.1.

I suspected, when doing tests in Ozy's thread, that the difference of build often compared to normal building was very small, or used in certain circumstances perhaps. So I will do the build tests with a large amount of units (500 or so each) to try to keep the chance variation factor to an absolute minimum.

It may be a bit later before I can finish.

Tom

 

[Bowsling]

I did a bit of a study on that
Rome was often for both, Egypt was never for both, and Greece was never for defensive and often for offensive.

             Egypt   Greece  Rome
Spearman           0       0       1
Swordsman         0       3       3
Galley                1.5     3       2
Building              1.5     0       0
Wealth (Turns)    10.5    1

Took a couple of tests to come up with this. On the first test, since spearman and swordsman were the only options ancient units, with me forgetting about Shift+F1 Rome went and destroyed me.
This was across 45 turns. Egypt ended up with 2 cities because they got a city from a hut. Rome declared war on Greece around turn 20.
Sorry it's messy.

 

[wolf_brother]

Bowsling, did you also change it so that those are the only units that those civs can build, and each civ builds the same unit? - ie, Rome couldn't produce Legionaries instead of Swordsman (a clearly superior unit), or take away the civ traits so that Rome isn't Militaristic (more units) and Egypt isn't Industrious (faster build times)?

 

[Bowsling]

Yes, Swordsman, Spearman, and Galley were the only options, hence why nobody built settlers. The traits stayed in, they wouldn't of had an effect on what the civs built. Plus each had seafaring, do that I could guarantee a coastal start.

 

[tom2050]

(Normal Aggresiveness, Chieftan Difficulty)
80 turns

Civs
----
Rome and Egypt, not at war, in seperate identical landmasses.
Each city produces 1 unit per turn. No techs can be researched in this time, and civ's stuck in Despotism. NOTHING but either the Warrior Defense or Warrior Offense can be built.

Units
-----
Warrior Defense 1.1.1
Warrior Offense 1.1.1
* Nothing else is flagged or checked besides Offense/Defense Strategy Flag, Load, Capture, and all Standard Orders, No Maintenance Required

Each civ has 9 cities, each starting with 3 Defensive flagged units (to not skew results, since AI will initially build all defenders); these are subtracted from final result below.

-----------------------------------------
NO BUILD OFTEN checked at all

Rome
353 Defense Units
367 Offense Units

Egypt
363 Defense Units
357 Offense Units

----------------------------------------

BUILD OFTEN OFFENSIVE LAND UNITS only

Rome
374 Defense Units
346 Offense Units

Egypt
366 Defense Units
354 Offense Units

----------------------------------------

BUILD OFTEN DEFENSIVE LAND UNITS only

Rome
355 Defense Units
365 Offense Units

Egypt
349 Defense Units
371 Offense Units

--------------------------------------------------------------------

Conclusion:
In these circumstances, no effect can be seen. The difference of unit builds is more likely from standard variation.

I will see if making offensive units 2.1.1 and defensive units 1.2.1 makes a difference. It is possible to conclude that the AI may build units with the same stats in a similar manner regardless of offensive/defensive flag designation.

Tom

 

[tom2050]

Same as above except:
Offensive Units 2.1.1
Defensive Units 1.2.1


NO BUILD OFTEN checked at all

Rome
352 Defense Units
368 Offense Units

Egypt
374 Defense Units
346 Offense Units

----------------------------------------

BUILD OFTEN OFFENSIVE LAND UNITS only

Rome
368 Defense Units
352 Offense Units

Egypt
357 Defense Units
363 Offense Units

----------------------------------------

BUILD OFTEN DEFENSIVE LAND UNITS only

Rome
363 Defense Units
357 Offense Units

Egypt
373 Defense Units
347 Offense Units

----------------------------------------------

Conclusion:
The difference of unit builds is still likely from standard variation. No difference is big enough from previous test to show that Build Often has any effect in this circumstance/situation.

Possibility: AI may use Build Often flags not on a global scale, but on each individual city. This would diminish any real change because it may depend on a limited number of hardcoded situations. E.G. If a city is under immediate threat, should it be more likely to build offense, defense, artillery, etc. Cities not under immediate threat, may ignore these flags ??

After all... units are treated individually, and not looked at globally, so it's pretty safe to assume the same for city build orders as well.

I'm not sure how else to test this... at war perhaps? Counting unit losses is too time consuming and likely for mistakes to be made.

Tom

 

[Steph]

Conclusion:
The difference of unit builds is still likely from standard variation. No difference is big enough from previous test to show that Build Often has any effect in this circumstance/situation.

Indeed. It seems to have the same effect as blowing in a violin.

Possibility: AI may use Build Often flags not on a global scale, but on each individual city. This would diminish any real change because it may depend on a limited number of hardcoded situations. E.G. If a city is under immediate threat, should it be more likely to build offense, defense, artillery, etc. Cities not under immediate threat, may ignore these flags ??

I think this may be also biased because you have no temple, granary, etc, and perhaps you need something else than units.

After all... units are treated individually, and not looked at globally, so it's pretty safe to assume the same for city build orders as well.

This is a rather bold assumption.... In the save format, you have for each civilization a list of units for reach strategy. So the AI knows that it currently has 20 offensive units, 10 defensive, 2 workers, 6 naval powers, etc.
It may completly ignore it... Or it may use it to know what to build next.
You could make a quick test. Place 3 spearmen in each city. Then place 50 spearmen in one more city.
And then see if the AI build only swordsmen until they balance the spearmen, or if it built half spearmen, half swordsmen.

I'm not sure how else to test this... at war perhaps? Counting unit losses is too time consuming and likely for mistakes to be made.
Tom

I'm wondering if the no maintenance is biasing it a bit. After all, if the AI doesn't have to pay for maintenance, it may just built units after units.
You should perhaps try to add a small maintenance. With something like 30 free units + 20 units/ town. To be sure the AI will built many units, but may stop at some point.