AI city placement and misc. suggestions

[civ4-advciv-oracle-bug]

This example is quite extreme too so i think it's a good idea to add it.
I had several examples of desperate defense but not so much of offense.
It is example 59.

Here AI insists to attack a city guarded by 15(!!!) swordsmen, with only.. 4 units.
The obvious move would obviously be to retreat, ideally to guard his cities.
But AI executes the attack anyway, despite having the movespeed advantage (attacks from his territory so he can move 2 tiles per turn unlike me) for a retreat.

I hope this can help too and i think it's quite extreme, i'm trying not to send additional examples unless they are particularly relevant or striking or not already covered.
Would be nice if this could be improved in advciv too if wished/possible i mean anyways.
Thanks,

edit: maybe he attacks because he considers each of these combats individually. Indeed, he has good odds for each of these combats. But if this is indeed (i don't know i mean anyways) how he sees it (plus possibly giving high value to taking back rome (the city he chooses to attack in this example)), then maybe a way to address it (if not already existing or not applied to this example case i mean anyways) would be to have cumulative strength of the stack vs cumulative strength of the other stack rather than individual combat (if that is how he sees it). It would be nice if the AI also had (if not already existing or applied in this case example too i mean anyways) an appreciation/estimation for defense rather than attack when circumstances are unfavourable (here me being stronger, and his remaining cities that would become much weaker if his attacking units dies (but this is something i already asked/covered in previous examples so better not develop it here again i mean anyways but defense and attack seem indeed to be linked and it would be nice if AI even if slightly (better if a lot i mean anyways) had a better evaluation/appreciation for it i mean anyways, thanks)

(note: i can't move else the city revolts so i am building a huge stack before i move. I tried focusing on culture with artists and music and such i mean anyways but it takes too long to have the revolt chance fade, so then Alexander came with a huge stack and kicked my ass xd i mean anyways, so now i am trying again from this position by building a bigger army before attacking the remaining cities, instead of focusing on culture or economy, else i am too weak militarily then die. More than being high in strength requirement, i think the bigger problem (i think i mean anyways) is that revolt chance takes too long to fade (relative to culture here) so since i can't get immediate culture enough, revolt chance forces me to ironically build more units (instead of focusing on economy or culture) to not die from other people or/and being able to keep my offensive. Delaying the attack is also not a good option (i think i mean anyways) as Rome AI is weak early in this map and he may be much stronger later, plus i need to expand and my economy would really benefit from the gold and new ressources i can use or trade or take back (pig and silver in this example). I would suggest again to make culture fade faster, but i also accept the challenge of starting from this position and trying another strategy i mean anyways, thanks)

(note 2: i finally found the trick or rather strategy that would work, instead of building culture or focusing on economy, skip culture (no code of laws, no artists, no music, no buildings) and keep building units, this way they provide enough strength to calm revolt while also making my rivals unwilling or less likely to attack me, and now i can finally focus on economy as i wanted to. I still think the strength fading time of revolt is a bit too high and i built as a result i mean anyways a bit more units than i would have liked to, i also think it's counterintuitive (in regards to how revolt mechanic works in advciv by needing culture to make it fade i mean anyways (which if i am not mistaken is similar in vanilla bts but felt less requirement/constraint of needed culture to offset the revolt i mean anyways)) to produce more units to calm revolt rather than more culture, but at least i found a way that works and which seems paradoxally more efficient at least in this example i mean anyways than producing culture. Another thing to note is that with the revolt chance being so high, it is better or more efficient for me to destroy the cities and rebuild them later at least in this map (unless they are especially good i mean anyways) rather than holding my army, and advance, which works here since i wanted to destroy antium anyways to replant cities on the coast for later game i mean anyways. If say strength requirement was the same initially but would fade faster, it would achieve the objective of slowing my chain warmongering if it is one of the intended purposes anyway, while also not incentivizing me to destroy cities (especially if my army is so-so, my culture production capacity still low at this stage of the game i mean anyways, and i don't want to be held back pushing on the front i mean anyways or rebuilding my economy (without spending too much on culture that would make me fall behind in military/buildings and be eventually attacked). Again this is just a suggestion i mean anyways and i am not sure of the best way to handle it, i just i feel i mean anyways it could fade a bit faster but it's just my perception/opinion i mean anyways)

(note 3: having played a few eras in the current map i'm playing, it seems revolt is much easier to handle at later stages of the game i mean anyways (edit 3: to meet strength requirement with later stronger units, and also easier to produce culture and have many happy ressources or/and buildings to compensate the resent of foreign culture in the new captured city, but not easier to make the strength requirement fade, in fact even in later eras that may still be slightly too high, as i still need 10+ units at 100-150 culture or so it seems approximately if i am not mistaken i mean anyways, so while it is much easier than early warmongering (even if just conquering one city early), it is still maybe a bit too high to fade (the strength requirement for revolt), the happiness part is easier to handle with the rule civic or other civics possibly, ressources or/and buildings/wonders i mean anyways, thanks), when many strong units are available to easily meet strength revolt offsetting requirements, plus there are many culture production capacities (music allows for culture, conquered cities have a large population which can all (that are available can) become artists, which fills the culture bar very fast, plus code of laws allows caste system so no limit on artists, and monarchy (the rule civic) allows to bring enough units so that the population becoems happy and is not lost every turn of producing culture. The combination of these factors (among possible others i forgot (edit 2: also ressources that may or not scale with buildings provide extra happiness and more of them are available later in the game, and less in the early game, that i partly forgot to mention and partly omitted to be honest it is my bad too partly in it i mean anyways) or don't know about or/and overlooked make it very easy in advciv to suppress revolt chance with current rules).
However, in the early game, units are weak (swordsman would be about the strongest, culture capacity low (no code of laws or low population cities so low artist count, lower number of units or no monarchy rule civic to calm unhappiness, or/and no music to produce culture, or/and low hammer count due to cities not developped enough or low pop to offset revolt chance. The result would be quick population loss (that is already low), not enough/artists culture, many of population unhappy and city stuck in revolt offset for a very long time, either halting offense or even just simply paralyzing the economy as a lot of ressources are spent just to stabilize the revolt chance (produce more units to offset strength requirement), resulting in high unit cost, less buildings in cities, and falling behind. I may not do this the most efficiently, but it seems there is significant advantage to attack (quite) early, when i have 10-15+ swordsmen extra available and the enemy is still weak, but with current rules of advciv i mean anyways it takes too long to stabilize back the economy, it's not so much about halting offense but more about weakening economy, making it better to simply destroy these (still quite low value) cities that would cost a lot to stabilize. It may not be the ideal but it seems to work best in current rules.
What i suggest is to change revolt mechanic per era somehow. Revolt suppression works fine (perhaps is very easy in later stages, when i ahve a bunch of cataphracts for example with justinian and solid economy and culture production capacity, perhaps make it slightly harder later in the game to suppress revolt, and much easier early. I don't know the exact formula, but if it's a flat strength rule that does not scale with era/turns of the game (but era is more likely to fit current units and production capacities rather than turn that may have different achievements unlocked based on difficulty or map local situation i mean anyways so maybe era or something similar may scale better?), then it favours too much later invasion and discourages too much early invasion. So maybe having this kind of turn number factor (easier to suppress early in the game, harder later), would make the strength revolt suppress mechanic more interesting and less frustrating? What do you think of this? Thanks,)

 

[civ4-advciv-oracle-bug]

There is another example i would like to add.
It is example 64 (i held back from sending other examples that may not be as urgent i mean anyways, plus this one is related to example 22 (edited the example 22 comment to add small info about this) so maybe they can be reviewed together).

Here Roosevelt AI asks for 405 gold for peace, however i am much stronger than him and conquered all his mainland cities (he only has a few cities remaining on a small island with relatvely low military numbers).
Why can't he accept a more even or even just an even peace treaty? Considering i won the war, i was expecting him to pay me actually, would that be possible too?
Would you consider improving this in advciv i mean anyways? Thanks,

edit:
after ending turn in example 64.1 he is much more reasonable suddenly and gives me gold as well as other things.
Would it be possible for him to adjust peace conditions based on my progress during the current turn?
Also, even before ending the turn, i think i had made enough war progression (i only captured one more city during the example 64.1 turn, but i won cities and lost much less units than him during the previous turns appended in example 64.2 i mean anyways) to have a much better peace treaty deal in the example 64.1 turn i mean anyways, so would you consider making the peace evaluation "fairer" at earlier stages of the war, here it would be him offering me gold sooner than when i capture his last city.
It may actually be a good incentive for me to declare peace with him sooner and which may benefit him or benefit in AI vs AI context war i mean anyways maybe (more gradual peace treaty conditions and not the sudden spike in one turn in an already lost war, to avoid war dragging too long due to harsh peace treaty conditions).
I have the save files of the turns before if needed, maybe this save file is enough to revisit the peace formula if wished/possible for advciv i mean anyways?

edit2:
added save files for all turns of the war (in example 67.2 archive. For each turn i sent first save file of the turn i have and last save file of the turn i have)
I hope these save files can help investigate better how the peace conditions evolve per turn as i keep conquering his cities and defeating his army too i mean anyways.
Ideally i would like in this example at least or a similar position example to have fairer peace treaty deals more early, which here would mean him giving me gold or/and cities the more advantage (military numbers remaining of each force ratio, number of cities he lost, number of units i killed vs lost and he killed vs he lost) i gain over him (or expecting me to give him peace bonuses if he won similarly i mean anyways in another map/context). Would you consider improving this i mean for advciv if wished/possible i mean anyways?

Thanks,

 

[civ4-advciv-oracle-bug]

I also added example 67.1 similar to examples 14 and 15 where a settler moves with escorting units out of cities during wartime (a German AI's settler group near Frankfurt in this case) instead of protecting the settler in city and/or protecting cities from the threat of my attacking troops in range of his cities and settler group. I added an edit in examples 14's and 15's post mentionning it.

I don't know if this example is helpful or redundant, perhaps it may help if you would want to test a fix on settler being too adventurous during wartime with this extra data, else it's easier for me to submit than to wonder i mean anyways so if it doesn't help maybe it is fine to ignore this example i would say i mean anyways.

note: German AI also behaves in example 67.1 similarly as in example 37: a huge stack gets baited out of its core city while i capture his other cities that are much less defended. In this case it's not as bad as in example 37 as they can retreat if i don't attack them, plus my nearby city is not as protected, but if it can serve as additional data for testing or observation perhaps i'm mentionning it else it may be ignored too i would say i mean anyways. Added an edit mentionning this in example 37's post too.

note 2: actually it's quite as bad as in example 37, after i spent one more turn gathering my nearby troops (cataphracts), he still didn't retreat (as i think he should) and got baited to attack a few more units if i'm not mistaken, as a result i had enough nearby troops to decimate all his incoming stack, and now his cities are defenseless almost i think (didn't look yet), if only he chose defense instead of offense (especially when enemy is stronger), he may have had a better chance, or perhaps he should have asked to become someone's vassal sooner maybe with mansa musa AI or willhem AI in this example i mean anyways.
I added example 67.2 (one turn after example 67.1 save file) where i gathered more troops and he still (as i think he should have done i mean anyways) didn't retreat, and example save file 67.3 with the result of my counterattack (his stack is decimated with minimal losses for me, ideally he should have seen this coming since i am stronger and just retreat or/and at least attempt somethign with his stack and not make it stay here on the hill i mean anyways or/and if attacking is too risky or most likely to be countered maybe instead ask to be someone else's vassal i mean anyways, thanks)

 

[civ4-advciv-oracle-bug]

I am aware i should not spam, but i found an example that expands the problems encountered in examples 34 and 31.
It is example 68 (i added the save file + screenshots for each case with its result, to reproduce trade during this turn with Willhem AI then do the actions explained below)

The AI's trading behaviour is inconsistent, and sometimes harms itself.

As shown in the screenshots :

Problem 1: AI is inconsistent + broken
- in example 68.1 Willhem AI is ok to accept 3 of my ressources against one of his, so far so good
- however in example 68.2.a, asking him what he wants for the same 3 ressources he agreed on, suddenly he says he is not ok. Same problem as in example 34 so i think it's interesting and relevant to add it if need additional testing/observation. He should be ok with things he agreed on if asked from my side rather than his.
edit: it's actually worse than i thought i mean anyways, asking in 68.2.b the same thing i asked in 68.1 (that he agreed on) but from the other side and what he wants for it now he does not agree, but undoing it and asking from the other side again after that the same that was asked in 68.1 he agrees to now

Problem 2: AI is not fair (AI does not agree to trade A vs X + possible bonuses but he agrees to trade A vs Y and also agrees to trade B vs X)
- now i want 2 of his ressources, so in example 68.3 i ask what he wants in exchange of 2 his ressources and he wants 2 of my ressources + gold per turn
- however in example 68.4, as 14 gold per turn is a bit too much and i'm trying to reduce it, i put 3 of my ressources he agreed on previously, against 2 of his ressources he was ok to trade in example 68.3 and what he wants to make this deal work, and suddenly he says it's impossible. However there's no reason for him to refuse, he agreed to want my 3 ressources (in 68.1), and he also agreed to trade 2 of his ressources (in 68.3), so what i expected and what i think should have happened is he would accept my 3 ressources plus some gold per turn but less than in 68.3 (so less than 14 gold per turn in this example). The value of the 3rd ressource i added as compared to in 68.3 was intended to reduce the price in gold per turn, however suddenly he doesn't want to trade anymore. This seems nonsensical to me as he had no problem to accept these items before? Does he want only gold in exchange of 2 ressources else it's not worth it to him? Seems weird and weak argument and more like AI is broken if i am not mistaken, but i ask the question if this is how AI is intended to work? Doesn't seem realistic, but what do you think of AI's behaviour, is it working as intended?
edit 2: if AI refuses out of being "considerate" (not wanting me to give more than i need to give to get (an) item(s) in a trade), and if while this is the case AI still insists to refuse the trade (due to being unfair for me) despite me wanting it, then perhaps a way to indicate it while signifying that AI is willing to accepting the trade may be "We don't mind accepting such a trade." or some formula that indicates that AI thinks i gave too much but if i insist and it benfits them they will accept it. For more realism AI may simply plain accept it witout any special consideration more often than not though, what would you think of adding such a formula for example but not rejecting a trade with a bonus if it costs nothing extra for the AI? Thanks,

Problem 3: AI is broken + needlessly reluctant to accept bonuses given to him for free (doesn't agree to what it agreed before plus extra bonuses for him for free), he should be ok to get freebies i mean anyways if it doesn't harm him and costs him nothing extra in exchange (this is not the same as asking a player too much and then the player would have to painfully bargain, no, it's just not refusing extra things given to him for free on top of what he already agreed to)
- now in example 68.5 i say screw this xd maybe if i input gold manually as in example 31 it would work? But no, giving manually same conditions that in 68.3 that he agreed on, but giving him on top of it a free ressource from me, then suddenly he is not ok?? AI seems really broken. This seems to fall back to example 31 problem where AI does not accept more if it's free for him and he has nothing to give more in exchange.

As this example 68 is interesting and overlaps with 2 cases in examples 34 and 31, plus a new situation in problem 2 (AI not agreeing to trade A vs X + possible bonuses if he agreed A vs Y and B vs X in the past). I added a note about these in specific examples posts.
Also, i'm aware these examples may not be reviewed due to the extra quantity i add that was already too much i mean anyways, but i think the AI is quite broken here and it would be nice if it could be improved or/and reviewed, but most importantly it's also because i'd rather post it and let you decide which examples you want or not and be at peace with that than wonder. I have a few extra more minor examples that i would keep for later or solve myself or in another post or at a later time possible i mean anyways, but if i think it's important enough i'd rather psot them at the risk of them being ignored due to quantity.
Thanks,

edit 3: there are other trading bugs, for now they are in total: examples 31, 34, 68, 105, 106, see their respective posts for details, thanks,

 

[civ4-advciv-oracle-bug]

I will quickly add a variant of other war declaration examples that i did not show in other examples.
It is example 69.
Mansa Musa AI declares war but while declaring war he does not invade at all unlike what he does in other examples, but instead he retreats further away from the border (closest would be examples 19, 25, 27.1 (added a mention in their respective posts)) i mean anyways.

At first during the turn before this save file (which is example 69.1), i hesitated but refused German AI's capitulation offer, following this i chose to continue my offensive against German AI and as a result lost a few more units to German AI and so did German AI defending against me. So next turn (after 1788 AD) alternative version (not this example but same date i mean anyways), Mansa Musa AI declared war and invaded me. I thought/felt it was an overlook and too greedy of me to not expect his invasion as he is stronger (plus some weird cavalry movements getting closer to me while he has no war rival) and instead keep pushing greedily (my mistake) against Germany AI.
So maybe he decided to declare war anyways, but since i stopped my offensive, he felt me + Germany combined were too strong for him to attack.

However, if that was the case, he should not have declared war.
I think it is correct and good judgement of him to not attack me while i am still strong enough (about as strong as him with effective offensive units i would say i mean anyways).
But if he is not going to start his offensive, it would be so much better if he actually delayed his war declaration until his offensive starts.

So this is a variant of the other war declaration then invade then retreat without city attack examples, hence i thought it was interesting maybe helpful to add here i mean anyways.
Would be nice if Mansa Musa AI timed his war declaration with the begining of his actual offensive.
He gains nothing and has no need to declare war so early, even if he intends to eventually.
Would you consider improving this in advciv i mean anyways? Thanks,

note: he is pleased but still declared war, i think this is nice because i was getting away with the game and close to domination victory, delaying anymore i would have cavalries and a bigger army, so the decision to declare war is good i think i mean anyways, but it's too soon, wait until offensive starts, else he gains nothing, plus surprise effect if there was is lost i mean anyways and possible other downsides i mean anyways.
Thanks,
edit 2: my bad he is cautious, but even at pleased (without making a vassal in alternative way i played (save file not provided here), he still declared war at pleased, and gifting him gold to increase his pleased by +1 did increase the pleased by +1 but did not prevent his war declaration either, which i think is good decision to declare war on me but too soon if his offensive doesn't actually start as said above i mean anyways.

edit: by uploading this, i'm aware it may be ignored as a result of it being too much data with the other examples, what am i trying to do is provide additional data and a variant that is quite different from the other examples, but it may be ignored if not relevant/wanted/needed/helpful i mean anyways and would be understandable i mean anyways. As for the rest i am also aware that it would take a lot of time to process all this data i uplaoded so i do not expect a reply anytime soon, i would be thankful enough if most of these examples were reviewed already it would be very nice, even nicer if actual fixes or/and changes regarding/related to them would be made in advciv that i could now play against again i mean anyways, but what i mean to say is that reviewing is something i am thankful for, even going through these examples i mean actually anyways. All this to say that i'm uploading this in awareness it may be ignored and thanks i mean anyways for the global consideration you provided, as for the rest i can only post it on my end and go with what happens as a result with it, and you handle it in whichever way suits you or you wish better i mean even though it's not my place to say i mean anyways, it may also be faster to review them directly rather than reply to me, and if so i don't mind (and would prefer actually) not to have a reply at all (at least not until all is finished), but of course it is only a suggestion, and if you prefer to reply to examples as you go i mean anwyays as one of you said i mean sure it is fine too, just that i wouldn't mind the no reply in favour of more review/focus (and would prefer it actually but is just my preference i mean anyways), thanks,

edit 3: due to his hesitation or rather here delay in preparing for/starting a war a few turns sooner, i won a domination victory just one turn later by using a great artist to mega expand frankfurt (save file 69.2 that i added too i mean anyways) and then ending turn in 1792 AD, plus planting a city too, so i think it's really too bad that he didn't declare war sooner (his army was stronger too so all he had to do was invade sooner or perhaps prepare for war sooner or/and backstabbing me while i expand on his friend Bismarck AI (especially considering i'm getting very close to a domination victory) or did something to prevent me win Perhaps when a player reaches 85% of a victory condition make the strongest AIs ultra weary and much mor aggressive if they are strong enough? Tbh even i didn't expect xd i mea anyways to have enough % to go from 50.x% to 55.x+% land just with the artist xd but all good i mean anyways thanks, so just if he could be more aggressive as i close-in on (>85% win condition) then maybe it would be harder to win/more competitive? Or maybe this is a bit extreme example but still would you consider fixing/changing/imrpoving this in advciv i mean anyways? Thanks,

edit 4: i added a screenshot (example 69.3) of what seems like a bug too, but i ask to be sure, from save file 69.2 if you end turn then you win a domination victory, and if you click "wait one more turn" (continue playing) then you don't get a new tech prompt (choosing which tech), also cities (nicomedia and cologne in this example) don't get a new production prompt and are stuck in no production while i can control and move my units freely i mean anyways, is this normal or a bug? Thanks

 

[civ4-advciv-oracle-bug]

I have also an issue i never reported (about worker actions) that i think would be very interesting to report.
Here in this example 70, workers chain loop building a mine then a windmill then a mine then a windmill etc non stop on the same tile every few turns.

It happens near the city of chicago of roosevelt AI, and the first time i noticed it was in 1710 AD due to city visibility gained by espionnage i mean anyways.
I have all the save files needed since 4000 BC (edit 2: about these saves files a news that is not bad, i accidentally deleted the alternative versions of where i tried another strategy however thankfully i still have almost all if not all save files from 4000 BC to last save i mean anyways) xd every turn i mean anyways if you'd want to investigate it more

I noticed it happen in 1710 AD then 1730 AD then 1750 AD. I did not look too deep into (and had no espionnage vision), but it may have happened in loop longer than that.
Also i noticed sometimes in advciv (but i don't remember if 1.11 or/and 1.12) the AI replacing farms with cottages in this map or previous one i played, could probably find it and also submit it if you'd be interested.

For now, i'll submit save files in the range where it happened with an extra margin if needed (see edit below) for now i added the save file for every turn (begining + another save file for my actions during the turn often i mean anyways). As i said, i have all the save files since 4000 BC every turn xd unless i forgot once or something, so if you'd want more data i could send it it was just not to overload, thanks,
(i also read quick this issue happen in recent advciv topic pages so i remembered i noticed it today xd so here i send this file).
I don't know if it helps but i do anyway i guess maybe, if not fine to ignore it, but if it helps all good, thanks,

edit: included a bit larger range (in archive example 70.3) in case it helps/is useful i mean anyways, from 1695 AD to 1760 AD, but as i said i first observed it in 1710 AD (may have happened before though) and last in 1750 AD (but did not look too deep may have happened again after that date i mean anyways). I have more save games before after these dates i mean anyways if needed, thanks,

edit 2 : added example 81 that may be related to this, here AI builds a fort on top of existing pasture sheep (could be linked with windmill being built on top of mine?), another possible could be that in all the worker examples ressource improvement problems mentionned, it seems to always be linked to hill tiles (could it be because of the defense bonus or other tile characteristic such as the yield or something?), see example 81 for details, thanks,

 

[civ4-advciv-oracle-bug]

edit: nevermind it solved itself after i opened the game again, i don't know if it took some delay to update itself or i just didn't notice but for now it seems to work.
I would still like to suggest the note i made in this post though but it's quite minor so i would not insist on that

edit 2: testing it from 80.0 again to be sure it seems the issue does not immediately solve itself, maybe it takes some time for the pathfinding to actualize during the turn after open borders are signed, in all cases as it's quite minor i do not insist too much on it but there is some bit of issue about it that is not too urgent i mean anyways.

Spoiler (example 80, nevermind it seems that the issue solves itself after waiting some time during the turn (i don't know how long though as i was writing this xd i mean anyways) :

Here is also another issue i never reported before.
This is more of a (quite minor but still i mean anyways) technical issue but i don't know if it can be useful or helpful so it is easier for me to send i mean anyways).
It is example 80 (with screenshots included too to see faster what i mean i mean anyways).

Right after signing open borders (during the same turn), auto pathfinding (i mean the auto suggested path to move from a point A to a point B) is not the fastest, it is slower than manual pathfinding.

You can load from example 80.0 save file (before i sign open borders with mongol AI) and sign open borders with mongol AI.
Or you can load from example 80.1 save file directly for the same result (i was not sure if there was some data useful maybe before signing open borders so i kept both save files just in case i mean anyways).

Trying to move the axeman as in screenshot 80.2, you can see the default path is slower (3 turns to move to karakoum) than the manual one in screenshot 80.3 (2 turns to move to karakoum)

Ideally pathfinding would get updated after open borders are signed, and most importantly the fastest path would be default (not a slower one at least i mean anyways).

(note: one thing that would be nice too is if 2 paths are equivalent in number of moves/turns, it would be so much nicer if AI stuck/chose the one that has the most roads, sometimes roads are built between turns or i change my mind and want to bring the unit back, but it is much slower than if the unit took a roaded path that had same movement cost than a non roaded one (move 2 tiles non roaded in 2 turns is same cost as 4 tiles roaded in 2 turns, but if origin and destination are the same, i would prefer if all things are equal that default would be the roaded path for pathfinding i mean anyways. I don't know if it would be possible to implement nor possible or/and wished for advciv but would be nice if possible as that situation happens quite often in my games i mean anyways. Thanks,)

 

[civ4-advciv-oracle-bug]

I finally found the problem with forts on ressources!!

Here in this example 81.
This example expands several already met examples (see edit below) so i think it's very interesting to look at it if wished/possible for advciv i mean anwyays.

From save file 81.1, end turn, then mongol AI builds a fort on top of an existing pasture sheep near karakoum city.
I added a screenshot 81.2 showing what happens if you do these actions, and screenshot 81.3 related to example 58 (see edit below).

What mongol AI does is a big big mistake i think i mean anyways because the pasture is already linked to his territory so he wastes workers moves for no benefit.
Plus, if he somehow takes the karakoum city back he would have to build a pasture again now.

The fort provides no advantage than what is already available, plus wastes worker move.
Would you consider making workers not build forts but instead ressources improvements (for example pasture for sheep, farm for wheat, etc), as it seems this confuses them and is very very inefficient.
If he wanted to sabotage it in case i capture his cities later (very bitter AI xd i mean anyways but very unlikely) then he could just build forts in his actualy remainig cities and defend harder there i mean anyways.

Would you consider improving this in advciv?
Thanks,

edit: worse, the next turn he brings/wastes one more worker to do this very inefficient worker "improvement" which is not in this already improved tile i mean anyways.
So this is linked to:
- example 41: ressources having a fort weirdly on them but no improvement
- example 58: using too many workers (as shown in screenshot 81.3) for one improvement can be inefficient (due to more walking to bring all workers instead of having each improve a separate area. Making workers work in separate areas or tiles may thus improve the AI efficiency i mean anyways),
- example 70: worker chain looping mine then windmill then mine then windmill on the same tile (i suspect it may be due to related to hills as they give defense so AI is more inclined to build a fort or improvements there somehow? All these examples happen on hills, i could be mistaken here though but it seems linked to this i mean anyways)
I added an edit in linked examples
- example 85: see edit 5 below

edit 2: added example save file 81.4 and screenshot 81.5 that fast forward to see the result (one turn before fort is completed in example 81.4, and when fort is completed in screenshot 81.5).
I want to insist that forts take longer to build than ressource improvements (for example a pasture) so this is even more inefficient, not only he rebuilds something already built, but he also wastes many workers to do that, and worse is even if there was no improvement on the sheep choosing the ressource specific improvement (here pasture) is still so much faster. Based on that, i think forts should really be deprioritized on ressources. To fix the issue of why AI wants to build a fort on an existing improved ressource is a separate issue, that perhaps if it is solved may also allow to reat/fix/examine the issue of AI focusing too much on forts if there are better and faster worker actions (roads, farms, cottages, mines, etc) that should almost always be better to improve the cities and AI strength. If enough of these are built, then maybe forts would be okay, but their benefits (when the AI choses to build them) are still questionable especially in regards to time spent on them. So i would suggest either do not allow AI forts until late in the game (does not fix the issue but losses over it/works around it) i mean anyways or find a way to better handle fort priority and relevance, or another approach maybe, i don't know the correct or possible best way to approach this so these are just my suggesitons i mean anyways but it would be very nice if AI could be improved regarding this forts on ressources problem (on top of forts giving little benefits especially in regard to time spent which is enough to deprioritize them a lot more i would say, on top fo fixing the reason for why AI needlessly builds some of them, but as this is a lot these are just my suggestions/remarks i mean anwyays and id on't know if it's wished/possible/easy enough i mean anwyays for in advciv i mean anyways, would be nice if possible though as i think AI would be quite a lot stronger as a result. Thanks,

edit 3: at the time the fort is completed, no mines are build in sight in the remaining mongol AI cities, instead of building this (needless and very inefficient i mean anyways) fort, would you consider deprioritzing forts further and instead for example giving higher priorities to mines on hills (especially hills grassland as they cost one less food for almost same hammer benefit than plains hills which s very strong/good early i think i would say i mean anyways) (or farms too as they are very strong to grow city or slave production i mean anyways)? 2 more mines on tiles where citizens can be affected can make a real and direct difference in gameplay/AI strength i mean anyways. Thanks,

edit 4: also, a problem i noticed in example 41 is that for example Japan AI spends worker turns to improve tiles where there is no city (the eastern part of japan territory), would you consider deprioritzing improving tiles (except roads and similar maybe and improving ressources) that are not in range of cities where citizens can be affected to them? This is linked to something similar i noted in example 82 i mean anwyays towards the end of my explanation (before edits if any edit happens further). I think it would make the AI quite a lot stronger especially early i mean anyways, thanks,

edit 3: linked to example 85 where an AI worker builds a cottage on unimproved banana instead of plantation or nothing, see example 85 for details, thanks,

 

[civ4-advciv-oracle-bug]

Here is also another example i didn't submit on how worker efficiency may be improved quite a lot.
It is example 82.

Here mongol AI worker (after ending turn in 82.1 save file) starts building a cottage on a plains tile.
However, there is a much better tile candidate (as shown in screenshot 82.2), the grassland tile (spot 1 in the screenshot) that is not improved too.

Therefore, i would suggest that for food type improvements (farm, cottage), start first with highest food terrain that is workable and uninmproved in city large ring (as one of you said the word ring i mean anyways) range (and in our cultural borders).
In this case, start with the unimproved grassland first if you want a cottage, not with the unimproved plains first, as this plains tile will be quite useless due to lower food terrain yield.

This could be linked with my edit 4 in example 81 where i suggest to last improve (if at all doing it, i really don't know if it's worth except connecting ressources but i digress i mean anyways it's not my main point) the tiles outside of workable large ring of cities, and first improve the tiles inside of the large ring of cities (that belongs to our culture not of another player) i mean anyways.

Would you consider improving this in advciv?
By making the AI focus first, for food type improvements (cottage, farm), on high food terrain tiles instead of lower food terrain tiles, AI may just with this change be quite a lot more competitive, especially early.
It would also be very pleasant to watch.
Thanks,

 

[civ4-advciv-oracle-bug]

Although i should not spam this is a big oversight/blunder from Willhem AI.
It is example 83.
I am aware it may be ignored as a result of too much data, but i think it would be a nice data point to improve the AI if wished/possible for advciv i mean anyways.
I also did not send a similar example if i am not mistaken so i hope it can help.

Here Willhem AI, one of the leading AIs, goes very far away to declare war on Roosevelt AI, however, there were closer rivals like mongol AI (or possibly Julius AI if Willhem feels confident enough to attack him i mean anyways, or me but i am quite strong militarily plus in quite good terms with Willhem AI so i may not be his first target but better than the farther away ones or target no one else if there is no other nearby target do not go for the furtest ones i mean anyways) that he could have targeted instead while staying close to his cities (making the expansion making much more sense and possibly profitable than capturing cities very far away).

The big problem is i have an about 20 stack of military units about to invade him, i don't know his military status in details but it seems his cities are very low guarded and i assume most of his army is far far away.
I am waiting a bit before i declare war so that he would not have the time to come back or/and injure himself in the process.
Willhem AI that was majorly ahead will now die easily from me.

It is quite realistic as i assume it happened in history (warmongering too far or too much then collapsing due to being too thin), however i feel/think it's a huge blunder of Willhem AI.

To improve this, make AI extra/more wary to declare war to a far away rival (for again very questionable if at all benefits), and more likely (if he really wants war) to target closer rivals (much more meaningul and likely to give benefits, plus if he fails or wants to withdraw (being backstabbed invaded or changing his mind or stalled war among reasons i mean anyways) his cities would be closer this way, and if he wins it's a nice and direct expansion from his land too (easier to manage/defend/spread his troops in reponse to events in the game with more dynamic reaction possible, expand from etc i mean anyways plus a bit less city cost from distance to palace).

Ideally this would result in this example in him attacking mongol AI or Augustus AI or me (but i am quite strong but quite friendly with him), but not faraway AIs like roosevelt AI (which make then Willhem AI a very easy target now, too bad as he was leading).

Would you consider improving this in advciv? Thanks,

edit:
maybe an exception should/could be i mean anyways if AI is boxed in (so no other rivals could backstab him or much less), or if all stronger rivals or most are busy fighting wars or have another target, then it may be worth to go for the far away and weaker or/and preferred rival to target rather than the closer one and maybe stronger i mean anyways, as shown in screenshot 208.1 where Boudica AI makes a profit in capturing a far away city (Nanjing city in this example i mean anyways) for quite low risk considering Boudica AI's only neighbour me in this game i mean anyways is busy fighting a war, and not too much stronger than her i mean anyways, thanks,

 

[f1rpo]

workers chain loop building a mine then a windmill then a mine then a windmill etc non stop on the same tile every few turns.

I've looked into this a bit. The city has deprioritized production when the Mine gets replaced the first time (from a multiplier of something like 154% down to 117%). Since these oscillation issue are especially tedious to investigate – and can't be fully weeded out – I've decided to abandon this one. You're conquering AI cities every few turns and the remaining cities are turning into some petty island kingdom. Not shocking if priorities shift rapidly in the process. Getting conquered also tends to leave the AI with too many workers, which makes such shifts more visible. Even from the perspective of a watching player, I don't find this particularly concerning. "I wish people didn't act so nonsensical as I kill them." Could also be that there is some major inconsistency in the calculation of the yield priorities, but I don't think it's strongly suggested.

Right after signing open borders (during the same turn), auto pathfinding (i mean the auto suggested path to move from a point A to a point B) is not the fastest, it is slower than manual pathfinding.

This was fairly straightforward; though not as easy as I thought. I had fixed a similar issue - upon declaring war -, but the old path wasn't cached in quite the same way. Anyway, I've pushed this to GitHub so that something can be said to have proceeded, even if it's the least important thing.

 

[civ4-advciv-oracle-bug]

Hehe thanks for the quick fix.
Yes, i'm aware it was minor but it's very cool that you have fixed it xd.

As for me, i am playing this map currently, so i would want to finish it with my current advciv version for consistency. If there is a new version or code by then all good, else i can just play another map and see ((even though a bit sad, i look at the future, i am a big grown person i mean anyways... maybe anyways...) and keep spamming examples... xd)

Like i said, i expect that some examples would not be taken into account, so it's good enough for me that you actually reviewed it i mean anyways.
To be more clear i mean anyways, what matters to me is the most changes or reviewing is done on my examples i sent (painstakingly.. no i kid xd i like to find new examples if not redundant ones anyways), so the more time you spend on examples the better it is for me. The more in depth you go into these the more i am likely to see changes related that hopefully would make the game more fun. If you review most examples and don't find it worth or doable/wishable as in this case, at least i submitted so it's all good xd or at least it's how it would have to be i would say.

So spend more time and more go in depth (if you want and like it i mean), this should be good for me and i am thankful for that. I expected while doing this process that some examples may not be taken into consideration, and that it would take time, so the more time the better in fact i mean anyways.
That being said i have a game to continue xd, but i'd rather that you take time and review it all or most rather than half ass it for lack of a better word. Whatever comes of it is the process. So thanks for that and now i play xd. But of course the more changes the better, just i don't expect it to be fast so all good on that, if it takes weeks or something, is not what matters most to me (even though the sooner is the better but) but that you go as in depth as possible and as you want i mean anyways in as many examples as possible, this is the most important to me even if it's not linear how you do it, just remember from time to time for my next maps or the next ones after and i'm all good.

Thanks,

 

[civ4-advciv-oracle-bug]

Also, i'm a bit of a dummy (at least in this case i mean anyways) i mean anyways and didn't notice xd i mean anyways, but about this chain loop mine windmill issue, i actually declared war in 1690 AD (just 1 turn before first save i sent), so i added now an extra 1685 AD save file (example 70.0.0) 1 turn before my war declaration (2 turns earlier than oldest save of this issue).

If you want to autoplay from there as they say this word autoplay i mean anyways, or maybe you manually play and skip turns without declaring war and watch what Roosevelt AI is doing near Chicago, you'd be able to see in a few turns, say 1710 AD or around it, if the same issue happens or not (i.e. if it is triggered by war behaviour/craziness or the issue is something else (i suspect their behaviour is not related to war but it's pure speculation/my intuition xd and maybe i am mistaken or not xd).

So if you'd still not review it further and abandon it i would say ok (what can i do xd i mean anyways but if it's done it's done anyways) i mean anyways, else there is this save file just one turn before war declaration where you can see in a few turns if same behaviour happens without war.

Thanks,

 

[civ4-advciv-oracle-bug]

I would like to add example 84 that i don't think i added before too.
Here wounded units of Willhem AI (that i assume are the remainders who survived from his example 83 initial attack on Roosevelt AI (but i didn't see his full offensive stack so i don't know for sure, maybe a part of it anyways)) accompanied by a bit few of (full health) catapults i mean anyways, are launching a desperate attack (or rather counter defense i mean anyways) against my outnumbering forces.

Unlike in previous examples where units are getting baited, the difference is some of these units are (quite badly) wounded (the horse archers).
It would be nice if AI could understand that wounded units should prioritize healing not fighting (unless they are numerous and healthy enough to think they can win and make a difference, but as a general rule having them retreat (so they can heal and fight again later when in full health or close to it i mean anyways) should always be safer or in the vast majority of cases be more fruitful for the AI i think i mean anyways).

Here it would mean retreating to the closest safe city which is utretch if i am not mistaken, it would mean AI makes a conscious choice (or what is the AI equivalent of it i mean anyways) to abandon the bottom cities as a result of him not having strong enough (wounded units below say 70% health/strength should be excluded from offensive force count and be individually assigned to retreat rather, not just to safer cities, but also away from the direction where any of my troops are (as they could get ambused and die which would defeat the purpose of them running away trying to heal and be useful later i mean anyways).

Would be very nice if wounded units focused extra more on retreat and avoiding combat (from all directions where there are enemy troops, if no city is safe to access heal in a tile safe enough away from the enemy troops i mean anyways), plus would be realistic and a nice feel and super cool to see the AI do if possible/wished for advciv i mean anyways..
What do you think of this?
Thanks,

edit: might be linked to example 77 below, see post for details, thanks,

 

[civ4-advciv-oracle-bug]

I also hesitated to add this example but now i think it's really relevant i mean anyways plus i don't think i sent this situation before i mean anyways.
It is example 77.
It might be linked with example 84 i mean anyways (added an edit note in example 84 about it).

Here, following mongol AI's failed offensive i mean anyways, after ending turn, a wounded axeman of mongol AI does not retreat (to karakoum in this example) like the rest of his units (a spearman), but stays very weirdly standing next to my units (not attacking them either).
AI broken or braindead?? xdd, i mean no offense or bad word, just this is quite weird why this wounded axeman does not follow the retreat, and stay instead in range of my outnumbering troops (a bit similarly than in example 84 but worse, here the unit stays idle and fails to retreat, while in 84 wounded units go out of their way to attack (counter defend me) both of which are not optimal for these wounded units, on top of the getting baited issue when losing and instead ideally focusing on defense as i highlighted i mean anyways in many other examples i mean anyways)

Ideally, mongol AI would understand this unit is wounded and make it retreat too, not standing there.
And also, more generally make all units when when is being lost and not expand on the losses needlessly as in other examples but instead focus on concentrating the remaining forces ond efense for better AI chances.
This lone unit standing there and not following the retreat seems strange, on top of not optimal, but i mean i wonder if something is not broken i mean anyways which is why i sent this example if you'd like to investigate it i mean anyways.

Would you consider improving this in advciv if wished/possible i mean anyways?
Thanks,

 

[civ4-advciv-oracle-bug]

It seems there's a quite serious problem with improving ressources xddd, i mean anyways but here in this example 85, mongol AI improves (after ending turn i mean anyways) banana with.. a cottage.
I added save file 85.1 and screenshot 85.2.
I also don't think i sent such an issue before unless i'm mistaken i mean anyways.

Looking from Foreign Advisor/Tech, it seems he can research currency so he would have mathematics, plus i don't see him not being able to receive calendar (plus he can receive my incense so if i'm not mistaken this would mean) so based on this i would assume if i'm not mistaken i mean anyways that he has calendar already.
So he could build an actual plantation on the banana (he might to build forts on bananas but i would greatly discourage this due to longer time to build them if i'm not mistaken and less flexibility (if another city is there anyways in other maps/circumstances i mean anyways)).

But he goes out of his way to build.. a cottage which is really nonsenical as it doesn't connect the ressource unless i'm mistaken or not knowledgeable about this, but even if it did more often than not in other maps or circumstances i mean anwyays this would result in inefficiences due to ressources (here banana) not giving their fully bonus and requiring a worker to rebuild a plantation later (very inefficient). If that's the case, why not build the correct improvement for each ressource tiles else no improvement at all? I feel/think it would be most useful for the majority of cases and to not decrease worker efficiency while being as flexible as possible.

It is linked to examples 41 and 81 (ressources having the wrong type of improvement built on them or no improvement at all (fort is not technically wrong but should be considered wrong i think i mean anyways when ressource is in city radius range and could benefit from the improved (especially considering it's cheaper to build than a fort) yields for the specific ressource improvement. If AI is trying to be hyper efficient by building a fort on ressources to gain them faster before tech is available, maybe change this approach as forts are expensive and long to build and AI could benefit more from other worker actions probably, plus when tech is discovered building specific ressource improvements should not be too long (making the fort questionable if it was the intention to anticipate the tech needed and having a fort to get quicker the ressource but i doubt it would be the main intention and even if it is it's way too inefficient for the AI i mean anyways in most cases so i'd suggest to simplify all this and never build forts in ressources in our cultural borders but instead only the specific ressource improvement when tech required for it is available i mean anyways. Considering AI has many workers generally this should probably work better for the AI, and solve or work around all the other fort issues and possibly others, as AI handles them quite poorly i mean anyway, not ideal but much much more simple and efficient resulting in quite a lot stronger AI than currently for changes that (i don't know about) but i assume that hopefully would not be too long (forbidding forts on ressources for AI, a bit drastic but should work i think i mean anyways but is just my suggestion (that i consider/think quite strongly is best suited but still just my suggestion/opinion i mean anyways)) which i will add edits to (and in example 81 too i mean anyways).
This example is also interesting because contrary to what i assumed the issue is not specific (and perhaps not linked) to hills, as it happens in flatland type tiles with ressources too if i'm not mistaken i mean anyways.

What do you think of this, would you consider improving this in advciv if wished/possible i mean anyways.
Thanks,

edit: i also found the same cottage on banana problem for Zulu AI (in same screenshot 85.2) so it's even worse xd, from that i tracked it down and i added save files 85.3.0 (3 turns before Zulu AI workers build cottage on banana tile if it helps at all with additional data or something i mean maybe i mean anyways) and 85.3.1 (end turn then zulu AI workers build banana cottage), and a screenshot 85.3.2 showing it i mean anyways. I don't know his tech status but i would say that regardless of whether he has required tech or not, do not build any other improvement on ressources than the ressource specific improvement (pasture for sheep, farm for wheat, plantation for banana, etc), i think it would probably on average and in most cases help a lot the AI increase its efficiency. A human may play it a bit differently (a farm in banana until plantation possibly) but as a general rule for the AI it would be easier to do it like this i would say, but of course if you could tell the AI to do only farms instead of cottages until plantation is discovered it would be even better (and do ressource specific instructions like don't bother with farms on say sheep since animal husbandry is so soon, nor mines on stone since quarry if such instructions already don't exist, but say banana it's not nonsensical to put a farm before calendar is discovered if calendar is far away, but cottages would be useless because their whole point is to grow and not be destroyed later, plus they don't provide food so better leave it unimproved if it's to do a cottage. The overall "do not improve tiles with ressource except with their ressource specific improvement" policy may be a lot easier but just my suggestion, farm early (assuming there is fresh water nearby (the civil service before calendar case may be too unlikely i mean anyways to cover it i would say i mean anyways, thanks) on banana would be nice if possible before calendar is discovered but it would be just a fancy bonus, i would suggest the simpler as a general rule maybe this is a bit easier, but if you'd want the fancier one all the better i mean anyways. Would you consider improving this in advciv if wished/possible? Thanks,

edit 2: added example 89 related to this where banana is still ignored and not improved by the AI, now the cottage is a hamlet, and AI goes for other improvements rather or overrides exisitng ones instead of building a much needed plantation if he (most likely i suspect already has calendar i mean anyways), see post of example 89 for details i mean anyways, thanks,

 

[civ4-advciv-oracle-bug]

I added also a (very/quite) minor issue related to example 21, but the opposite.
It is example 86.

Here from save file 86.1, worker capture chance may be not too low as in example 21 (1 in 7 if i remember correctly) i mean anyways, but (a bit) too high (5 in 8 workers captured is i think a bit too much).
To reproduce it, attack utretch with the strongest/healthiest i mean anyways swordsman i have near it and utretch is conquered.

While it may be fine to leave as is and say it is "chance", could it be made with another kind, less random kind of distribution (or i don't know the exact word/formula/technical detail either at all xd i mean anyways but what i mean is formula calculation), so that i get statistically more than 1 in 7 (if i remember correctly) workers in example 21, and less than 5 in 8 workers statistically in example 86?

While it could be luck, maybe make it more tend towards the average, or would it ruin the pleasure/fun of the game i mean anyways?
I added a note in example 21 about it.

What do you think of this? Thanks,

edit: i also added a save file 86.0 of the turn before 86.1 in case it changes something to the randomness if some changes are made i don't know? Thanks,

edit 2: i would suggest something closer to a guaranteed 2-3 workers in both of these examples may be more balanced or/and frustrating (example 21) or excessive (example 86), but it is just my opinion/suggestion (that i still think quite strongly even though issue may be a bit relatively more minor but still xd anyways but is just my opinion/suggestion i mean anyways) i mean anyways, thanks,

edit 3: i thought of a suggestion, what do you say of if number of workers is higher than 4, switch a to a formula capture chance calculation of chance = rounded-up(number of workers * 0,35) , what do you think of this suggestion or something similar? Perhaps add a random factor to it too to not ruin the fun? Something like 1 + rand()*number of workers * 0,4) for say > 4 workers on same tile for example maybe, else i mean if worker count <= 4 do something like 0,3 + 0,5*rand()*number of workers? Or keep current formula maybe i mean anyways? What do you think of this or something similar i mean anyways, thanks,

 

[civ4-advciv-oracle-bug]

i also found the same cottage on banana problem in zulu AI city, i provided a screenshot and save files 3 turns before and 1 turn before it is built, see example 85 post's edit for details, thanks,

 

[f1rpo]

about this chain loop mine windmill issue, i actually declared war in 1690 AD (just 1 turn before first save i sent), so i added now an extra 1685 AD save file (example 70.0.0) 1 turn before my war declaration (2 turns earlier than oldest save of this issue). [...] (i suspect their behaviour is not related to war but it's pure speculation/my intuition xd and maybe i am mistaken or not xd).

I think your intuition has turned out to be correct. Perhaps it was clear enough to you from the exact timing that your conquests weren't driving this process.

The persistent issue, rather than the production weight, seems to be the food weight. karadoc was aware that there's a circular logic at work where cities want food more badly when they have less. According to his comments, he didn't fully resolve the matter because of performance concerns and because he felt or hoped that a rough adjustment would counter the loop well enough. Maybe it does, perhaps my effort to get the AI to grow cities more aggressively made things worse. I've now just dialed up the K-Mod mechanism. I think this could be handled a good deal more reliably and without a significant performance penalty, but it doesn't seem worth the trouble to me considering also the potential for introducing new issues and that it's not clear that this will really calm cities down – there might still be more than enough pointless work for every worker's hand. I suspect that some 10-40% of AI cities (a very vague guess) constantly have some kind of bed bug that gives workers something to scratch – and that the main bottleneck for this kind of activity is the number of workers per landmass. In your example, it's 3 workers under Serfdom on a slender landmass with 2 coastal cities.

Through some added log output, within 3 turns I've spotted 4 AI cities in your game that replaced an improvement. Some of those may have shifted their food priorities for whatever reason. It's also not clear from the data I collected that they're going to immediately switch back to the old improvement. The data does show that they all like food a good deal more once they have less and vice versa, even with the K-Mod correction factor. E.g. Chicago, your example, thinks that food should have a (corrected) priority of 244% (pretty high this late in the game) while there is a Mine; two turns later, with a Windmill in place, the priority is only 208%. I've shrunk that difference to 236% vs. 215% by doubling the impact of the K-Mod adjustment. A little afraid to do more to avoid perhaps cities sometimes growing too inertly. Seems enough to stop the oscillation in this case. (Windmill wins out.)

Since you've also brought up several other examples of worker inefficiency: I don't think wastefulness of late game (or even midgame) worker turns should be a major concern. (Well, unless it's important for human automation maybe.) That's what the AI discounts are good for - to let the AI afford a few extra units. Important though to make sure that the AI produces enough workers, i.e. more than an experienced human player; there were too few in BtS and imo also in K-Mod. A flip side is that the AI often has time for tasks of dubious utility, and that doesn't look good.

I don't think there's much inertia built into improvement choices, for example (unsure; I haven't come across such code this time around, but I'm pretty sure it's somewhere and probably I've seen it before); if workers are idle enough, improvements will keep getting cycled around. Maybe more could be done - with reasonable implementation effort - toward doing nothing more often and perhaps also disbanding workers more readily. Rather than stopping inefficient tasks entirely, the AI could still perform them with a per-turn chance so that things that might turn out to be important once in a blue moon (e.g. long irrigation chains through unworked tiles) still get done eventually just not everywhere at once or a dozen times over. Something I'll keep in mind. Writing all this down also as a potential note to anyone who spots the occasional worker stuck in a loop.

Regarding worker capture chance (rather different subject): The most important part to me is really that it's 0 upon declaring war (i.e. no surprise attacks). And not 0 otherwise because that would remove any incentive to deal with orphaned/ cornered workers and would also seem rather vicious. Preferably significantly less than 100% too because I don't think players generally can occupy so many workers. But, in that regard, the AI could also simply delete workers when there is no hope of rescue. That could happen entirely behind the scenes, a humble AI change. But I do want the chance mechanism because it integrates the special rule upon declaring war better. I don't think anything other than a plain coin flip is going to be simple enough. It's not like this will be any credible simulation of taking civilian captives; don't want to give any such impression. The high variance aside, I don't think the expected numbers (0.5 per worker, obviously) are far off from what tends to play well. A large worker stack probably has been driven from city to city for a while, so that shouldn't be disproportionate and shouldn't require a special rule.

But maybe I'll see it somehow differently if and when I look at the respective savegames.

 

[civ4-advciv-oracle-bug]

Yes, about my intuition it was not something so complicated, it just felt weird like i am declaring war, and workers build a windmill then a mine, thing that they could have done without war, it just didn't seem intuitively related and 2 different separate/independent things which is probably/maybe why i thought this i don't know, even though i do not understand the in-depth things but hey cool to be right xd. No, i mean more importantly, thanks for digging into this. I looked at the commit changes too a while ago (and the nice and long notes detailed there too), so this extra message you added is nice to read too with all these details and all. Like i said many times, and i am very happy to read all these nice details, if you find it too much i am more than content with the commit notes (and the racking my head trying to look at the code (just from a general view), i find the whole process fun to follow the changes gradually (being able to give suggestions is even nicer to me i mean anyways), but just like i enjoy things my own way i mean anyways, i shouldn't meddle in saying this i think i mean anyways, but if you prefer to comment or detail or not you are more than welcome xd, i only worry about the energy you'd have left after this xd, but if looking at it from a long process perspective, it's nice i mean this gradual thing but it's just my opinion so i only wanted to give feedback on it and you handle as you see it best or what you like/want best i mean anyway i don't want to meddle xd so thanks for all that i mean anwyays, and of course gogogo at your pace too i mean bit by bit or lot by lot, i'll play with whatever version is ready later xd and thanks xd.

That being said, while it may be nice to fine tune the AI in things like workers efficiency and all, it may not be too easy as you'd go deeper into it (possibly i don't know but i assume it would be so quite naturally i mean anyways, as more and more things would/should (i assume) be untangled in a quite weird way due to how they were maybe? But what i mean is i would advocate for the most efficient fix. It's true that giving AI more workers would (and does) probably help a lot. Also, many small blunders are likely to persist. However, i think there are things that are maybe more relevant to focus on (about worker actions). So what i would advocate for if i may, but is just my suggestion, is to focus on the key core things rather than micro optimizations, i'm concerned about AI looking well and nice but also but performance and most importantly strength. When wasting many turns on forts instead of chops, it may not seem like a lot but added many times and with other improvable ideally things a bit of AI strength is lost. This is the kind of things that i would the most would like to be reviewed etc like all too i mean anyways, i mean they both look nice and also strong if done well by the AI ideally or at least better i mean anyways. Things like banana cottages (hamlets later in the game very weirdly still without plantation, even though AI is now building a windmill (i didn't update to your code yet as i want to finish with my current map first) but ignoring the cottage banana, i should add this example a bit later too but since i provided many data points about this banana cottage issue, maybe it would be redundant, if it helps i can add it for sure i mean anyways. As for the rest among worker efficiency issue, some may just be ignored, but if you look into them and it gives insights on other things or just this process of reviewing them in depth is nice i mean and what i want if i would say i mean anyways so i'm very happy xd about that i mean anyways.

There are a few positive things also i wanted to give feedback on (and kept spamming examples xd), but that i think would be nice to develop a bit. For example, i noticed in last few games of advciv 1.12 (same version i use since example 17+ from memory if i remember the example number correctly i mean anyways), which are city placement, it doesn't seem like a coincidence (i may be mistaken though but it seems less and less likely but i don't know i mean anyways) that it does seem quite better than in 1.11, i very rarely see them settling on metals for example. I think they ignore a bit too much food as pointed in other examples, but maybe a food/hammer balance works best so not putting too much emphasis on food maybe but a bit more? Also, they settle more aggressively to my borders which is very nice, in previous map Rome AI very "boldy" if i may say again settled near me and successfully stole my pig and silver which ruined my (already poor xd due to too many cities etc i mean anyways) economy prompting me if i may say this word i mean anyways even further to war, else i go stalled economy xd. In this current map i'm playing, Willhem AI very boldy and successfully stole the iron near my border, if i was not lucky that i could culture swap and get the copper from mongol AI (steal from him as they stole from me too xd), the game may have been very different for me, plus it's a nice production tile and freely improved from them but i digress i mean anyways, what i mean is they seem more bold, and to concentrate their cities a bit more (but may be just my impression or not), i think these should work quite well into making AI really more competitive early and i'm very satisfied with these).

As you maybe already saw or not yet maybe, there are a few war behaviours that i think are quite critical (example 37 xd among others), some that would be nice to improve, also some weird bugs and such. I wish i had classified them but maybe the linear approach is easier. Also, i did not want to reduce the exhaustiveness of my data (unless redundant or such examples), but also as you instructed in the very first messages in reply to me (and i follow instructions no problemo xd i mean anyways at least i try i mean anyways) i try to give the save point as close as possible to when the event happens and i hope this helps i mean but anyways. For now i'm trying to scratch my head/mind i mean anyways on my next move in my current map, i'll be very happy to follow the changes you make following this whether they are soon or not (ideally soon but i follow the process ((un)patiently but i kid a bit as i like to read your changes and all and understand even if a bit the whole things of why and what you change even if just the general idea i mean anwyays), i tried to backstab Julius AI while he was at war but he countered me and my army was a bit too thin, so maybe i'll change target reloading and go maybe for mongol AI as i had intended, this process is fun too. If i find significant examples i'll try to send them, else i'll keep them for my peace of mind as unsent numbered examples if want to revisit them later or something or maybe not, that's why some numbers may be missing as for example 77 that i added later after 84 due to being related to it i mean but anyways.

For the worker capture chance i have no strong preference on it, just it seemed a bit too high (and nice) for me to gain 5 workers in one attack, solving all my economy problems instantly (i should have built a few more) (edit: and also a bit too frustrating to gain only 1 in 7 prompting me to write this as an exmaple too i mean anyways as i thought it could be bit higher, but with this new example if it's maybe sometimes what i would consider to be too high, then maybe make it a bit less random to guarantee a bit, while not giving too much. Where i disagree a bit (a lot!!! but i kid i mean anyways) with you about workers is that they can make a lot of a difference and are never really enough as i heard some other players say quick xd, i don't mind having more workers xd, the more the better, i can always find things to do with them, else maybe gift them to someone else rather than delete (i just got this idea while writing this, for relationship modifier bonus, (edit 2: ) if i understand correctly how gifting units work and would affect relationship modifier i mean anyways as i didn't try it that much if at all i mean anyways but it's not my main point i mean anyways, just that if it works as i think then it would/may be a better option than delete them (but what if they pass workers around then xd) in the optimization idea but for the core rule i would say to make it a bit closer to average with some randomness but is just my suggestion i mean anyways).
(edit: i think the human player can always benefit from more workers as some other player is i heard say i mean anyways but i am repeating myself now xd so it was just a parenthesis i mean anyways that i think maybe a bit more sometimes and a bit less some other times with some randomness to keep it interesting may be a way to go about it but it's just my suggestion/opinion so it may or not be followed for advciv if you'd want/wish or not for technical or gameplay preference or other reasons i mean anyways i only submit the examples and my thoughts/suggestions on them i mean anyways and looking forward to whenever the feedback would come on them ideally i mean anyways but else is as is i mean anyways).

Also, a small note i wanted to remind you of with the open borders fix of pathfinding, there is an issue i often have which would be (very) nice quality of life if possible. I explained in more detail (very confusely xd maybe or not but my main point is) but the general idea is when i want to move from point A to point B, and there are 2 alternatives paths with same movement cost, one which is roaded, another that has tiles unroaded (but for same total movement cost from point A to point B), i would very very very very!! much want the auto pathfinder to default to the one with most roads. Reason is simple i mean anyways if i may say, sometimes i change my mind i mean about my units, and if i want to move them somewhere else, suddenly they are in an unroaded tile so i lose a lot of moves/turns to get them to my new desired destination, but if they were on a roaded path it would have been so much easier. It's not a critical issue but one that definitely may help my quality of life a lot if possible i mean or wished for advciv i mean anyways.

That being said, thanks again for all, and whether it be quick or long, i look forward to this journey as you go more in depth into my examples and such and others possibly comment on it or not, to which i may participate or not if i find it linked to what i want to say, else i would abstain i mean anyways but watch a bit i mean anyways, all this to say i mean anyways that i continue my game xd and thanks for this, i hope the AI can be even stronger smarter and better looking (edit 3: but i'm more concerned about strength xd, really i mean, i mean it's good and nice if it looks good and does nice things, but what i enjoy most is a strong/smart AI rather that is efficient, but if doing fancy as in refined things it's very nice but just i derive the most fun from a most optimal AI so that i can enjoy more each difficulty without cranking the bonuses too much, else it's really a different kind of game that i'm not as good at and/or that i find quite tedious nad less fun to me at least i mean anyways, but if AI does smart and strong efficient things (understanding some variety may be desired for me or all players i mean anyways) then i am more excited and find the AI competitive but it's just my preference i mean anyways and in general i like and am hapyp to watch these changes and play against them most importantly too i mean anyways but all is important maybe in this case or not i don't know how to say i mean anyways but i enjoy these things and thanks i mean anyways at the risk of repeating my self a bit overbearingly... or not maybe but thanks i mean anyways...) as in doing things less nonsensically even though there are some things that most likely will persist but ideally the critical things may be improved or looked at, i would be very happy with that, else if it's to be left as is at least i submitted it like i said xd, so that being said i am back to my games, and i watch/look gradually as you do them, and play my games too xd hoping more for my next games but adapting if there is not or not yet xd, so thanks for this etc etc i mean anyways and thanks for lack of a better way for me to quickly explain, but i really mean it i mean anyways (edit 4: as needless it is maybe of me to say or maybe not i mean anyways but thanks i mean anyways), going in depth in these things is very nice i mean anyways but i enjoy it too with these examples and all too i mean anyways and your commits and messages or not messages and all i mean anyways, thanks, and for now i continue to play, thanks,