JosEPh_II
TBS WarLord
This current game is the 1st time I've set stack limits. I jut have not noticed massive build ups....yet.
JosEPh
JosEPh
AI suggestions
- Thread starter Koshling
- Start date
Stack of 12. The AI seems to be able to play with it. Just comes with several stacks instead. Defend Only units don't count towards the limit either and most of the time the AI has a few healers along too, thus getting more than 12 in a stack.
Cheers
EDIT: Changed to 20 on Koshling's recommendation.
Cheers
Cheers
EDIT: Changed to 20 on Koshling's recommendation.
Cheers
Koskhing, we've been over AI and Civic Changes a few times. It might be slightly better but that's not saying a lot. It's still bad enough that playing on Deity the main reason I can catch up with the AIs early run-away is because they change Civics too often.
With the new Snail and Eternity Speeds this is even more so. However often the AI re-evaluates Civic choices it is way too often, and still throws them on switching "back" to earlier and worse Civics. It's not seldom I see AIs switching between the combos Prophets/Despotism and Divine Cult/Monarchy.
In current game 1280BC Carthage switched to Church/State Church/Despotism. Bad choice, and too little time between.
Before that it was 1306BC (26 turns earlier) that they switched to Open Borders/Divine Cult/Monarchy.
1332BC (26 turns earlier) they switched to State Church/Despotism. Again a bad choice, bad!
1403BC (71 turns before) they switched to Monarchy. From here on they shouldn't even have looked at anything earlier.
1439BC (36 turns) it was Interpreters. At least they didn't change this one...
It's been riddled with bad Civic changes the whole game from the 3 AIs I have contact with. Why go back to a "worse" civic, and multiple times?
The rest of the AIs must be experiencing similar Civic decisions. Sure, this game I might have limited the AI to 1 starting Tribe instead of 2, but I should still not have caught up with the AIs in the early Classical Era.
I know I was behind the AI (Most Advanced didn't even have me in it in the early Ancient Era) but now, already, I'm caught up and in second place in Most Advanced. I'm betting most of this is due to both bad Civic Choices as well as switching too often on the AIs' side of things.
I'll supply you with the save at current point (and later point depending on how much more I've played it) if you want it.
Cheers
EDIT: Just checked the GNP. All three have lots and lots of anarchy periods and in the last 700 years I've gone from roughly half of two of the AI's (equal to one) GNP's to equal to the highest one and roughly 120% higher than one and 4x higher than the struggling one.
Cheers
With the new Snail and Eternity Speeds this is even more so. However often the AI re-evaluates Civic choices it is way too often, and still throws them on switching "back" to earlier and worse Civics. It's not seldom I see AIs switching between the combos Prophets/Despotism and Divine Cult/Monarchy.
In current game 1280BC Carthage switched to Church/State Church/Despotism. Bad choice, and too little time between.
Before that it was 1306BC (26 turns earlier) that they switched to Open Borders/Divine Cult/Monarchy.
1332BC (26 turns earlier) they switched to State Church/Despotism. Again a bad choice, bad!
1403BC (71 turns before) they switched to Monarchy. From here on they shouldn't even have looked at anything earlier.
1439BC (36 turns) it was Interpreters. At least they didn't change this one...
It's been riddled with bad Civic changes the whole game from the 3 AIs I have contact with. Why go back to a "worse" civic, and multiple times?
The rest of the AIs must be experiencing similar Civic decisions. Sure, this game I might have limited the AI to 1 starting Tribe instead of 2, but I should still not have caught up with the AIs in the early Classical Era.
I know I was behind the AI (Most Advanced didn't even have me in it in the early Ancient Era) but now, already, I'm caught up and in second place in Most Advanced. I'm betting most of this is due to both bad Civic Choices as well as switching too often on the AIs' side of things.
I'll supply you with the save at current point (and later point depending on how much more I've played it) if you want it.
Cheers
EDIT: Just checked the GNP. All three have lots and lots of anarchy periods and in the last 700 years I've gone from roughly half of two of the AI's (equal to one) GNP's to equal to the highest one and roughly 120% higher than one and 4x higher than the struggling one.
Cheers
I will need a save from the turn in which bit makes such a switch on end turn please.
Double Post, I know, but do have another save game illustration a problem with the AI and evaluation where to keep Defensive units.
Where my bunch of units are now standing just inside Carthage's Capitol BFC I have just taken out, oh, probably 20-25 Town Watchmen, a 5-10 Archers, and 15-20 War Dogs.
Being able no, and not before, to see his Capitol and the units there I find that he has all of 1 Light Swordsman, 2 Javelineers, and 2 Town Watchmen.
I was counting on getting to the hard fight now, not have it behind me. More, and better, defense should have been stationed in the City itself, and not to mention what even 5 Town Watchmen extra would have done for his Crime Level there (apart from also having extra defense in cities).
Cheers
EDIT: He did move a number of units in for the next turn but without the extra 25% from fortifying it'll be an easier fight.
Cheers
Where my bunch of units are now standing just inside Carthage's Capitol BFC I have just taken out, oh, probably 20-25 Town Watchmen, a 5-10 Archers, and 15-20 War Dogs.
Being able no, and not before, to see his Capitol and the units there I find that he has all of 1 Light Swordsman, 2 Javelineers, and 2 Town Watchmen.
I was counting on getting to the hard fight now, not have it behind me. More, and better, defense should have been stationed in the City itself, and not to mention what even 5 Town Watchmen extra would have done for his Crime Level there (apart from also having extra defense in cities).
Cheers
EDIT: He did move a number of units in for the next turn but without the extra 25% from fortifying it'll be an easier fight.
Cheers
It's a tough balance. It very deliberately keeps units just outside the city on defensive terrain to prevent an attacker easily occupying good defensive spots adjacent to the city to attack from (and to a lesser extent to improve it's surveillance). The problem is that as it starts losing units there it tends to replenish them, keeping a percentage back in the city. Keep killing them and the percentage outside the city drops, so it replenishes it, until you eventually wind up with it reaching the point it feels it cannot move any more units out of the city.
This behaviour is reasonable, and good tactics in some cases, but it needs to be tweaked so that it only does it if the result is that your (visible to it at least) units cannot attack it's resulting stack on defensive terrain without expecting to suffer losses at least equal to (and probably significantly greater than) those of the defending stack. I'll see what I can do about said tweaking.
In regard to the anti crime units - the AI knows nothing at all about that mechanic yet. That is why I requested it was not put on until after the v22 release, because I knew i wasn't going to have time to add any AI for it for a while. Will be done sometime before v23 is about all I can promise.
*nods* Sound reasoning, in most cases. In this case I could attrition him down with a few units on the defensive hill (99+%) until his big units were gone and I could attack his units there en masse.
About -crime units: Since most, if not all, of them already have bonus in city defense and no where else can't a quick fix be to set them to AI City Defender tactics?
Cheers
About -crime units: Since most, if not all, of them already have bonus in city defense and no where else can't a quick fix be to set them to AI City Defender tactics?
Cheers
I think they are, but the city defense AI allows units beyond the bare minium the AI wants to alwasy keep IN the city to move into its surroundings (still leashed close to the city to be able to go back) and occupy defensive points (as you are seeing it do). Adding a little AI code to notice anti-crime and bias towards those staying in the cities will be easy - I just need to get around to it is all.
DRJ
Hedonist
Talking of AI behaviour on defending close to city tiles: I have a save where Ai has stationed 3 or 4 units in city and about 35 on a hill 2 tiles away (see persia in example).
They sit there and wait.. for nothing.
Aside 10 axes and some dogs, AI brought 23 slingers to defend the tile and kills itself with the costs of this (wasn't it tweaked so AI wouldn't build too many troops anymore to commit economical suicide?
In same savegame Mongolia - strangely - did not build too many troops while korea and persia and a lot of european nations did.
I also see that once AI tiles have been pillaged by barbs and there are still barb cities nearby they tend not to improve the tiles again... like forever... so instead of using their HUGE stackes to put 2/3 units on every pillaged tile (even if it has no good defense bonus) and then send gatherers/workers there they just keep stacks big on one tile and ignore the plots (to check that behaviour see the bison resource in korea in same save).
I mean what is AI thinking? They (korea, persia) have incredible stacks and the barb cities nearby only 3-5 units defending them.
Even if Persia or Korea would calculate to lose half of the stack they could easily attack and expand or lose and rebuild but have more gold for science in teh meantime as maintenance is gone.
Instead, they build troops again and again but never trigger the attack, as I suppose the first attack calculation is too low for them (do they consider that a dmaged enemy is easier to kill, so they could calculate the chances of first fight, remmeber the hitpoints the enemy would take, substract it from the value of the unit to simulate fight two etc. if four defenders were in there this would have to be done four times to get the chances for the secound round fighting the best defender with the new values) .
It also seems that Korea doesn't calculate to send it's magnificently promoted general to attack with the main stack? So there are a lot of synergies but sometimes AI can't knot the lose ends together to be effective (yet).
I think this save can actually help a lot to improve the AI a lot, if you play with some variables, like what value is needed that actually AI triggers attacks on nearby barb cities.
I know I already told you about the boat bug and you said you would look into it. So just a little follow-up: several hundred years later after my canoe blocked their coastal waters the Koreans still have not built a canoe or raft, so just to let you know the bug doesn't solve itself during the game.
They sit there and wait.. for nothing.
Aside 10 axes and some dogs, AI brought 23 slingers to defend the tile and kills itself with the costs of this (wasn't it tweaked so AI wouldn't build too many troops anymore to commit economical suicide?
In same savegame Mongolia - strangely - did not build too many troops while korea and persia and a lot of european nations did.
I also see that once AI tiles have been pillaged by barbs and there are still barb cities nearby they tend not to improve the tiles again... like forever... so instead of using their HUGE stackes to put 2/3 units on every pillaged tile (even if it has no good defense bonus) and then send gatherers/workers there they just keep stacks big on one tile and ignore the plots (to check that behaviour see the bison resource in korea in same save).
I mean what is AI thinking? They (korea, persia) have incredible stacks and the barb cities nearby only 3-5 units defending them.
Even if Persia or Korea would calculate to lose half of the stack they could easily attack and expand or lose and rebuild but have more gold for science in teh meantime as maintenance is gone.
Instead, they build troops again and again but never trigger the attack, as I suppose the first attack calculation is too low for them (do they consider that a dmaged enemy is easier to kill, so they could calculate the chances of first fight, remmeber the hitpoints the enemy would take, substract it from the value of the unit to simulate fight two etc. if four defenders were in there this would have to be done four times to get the chances for the secound round fighting the best defender with the new values) .
It also seems that Korea doesn't calculate to send it's magnificently promoted general to attack with the main stack? So there are a lot of synergies but sometimes AI can't knot the lose ends together to be effective (yet).
I think this save can actually help a lot to improve the AI a lot, if you play with some variables, like what value is needed that actually AI triggers attacks on nearby barb cities.
I know I already told you about the boat bug and you said you would look into it. So just a little follow-up: several hundred years later after my canoe blocked their coastal waters the Koreans still have not built a canoe or raft, so just to let you know the bug doesn't solve itself during the game.
http://forums.civfanatics.com/showthread.php?p=11312952#post11312952
Check out post 962 and further post by the same author on time reduction for AI turns etc.
Check out post 962 and further post by the same author on time reduction for AI turns etc.
Already looked at, and in practise made very little difference. The problem is that he has sped up some very basic routines to go twice as fast, but those routines were only 1% of the total time, so twice as fast was a net 0.5% speedup.
The best way to attack speed optimization is to profile the entire call tree in the contxt of actual save game turns to see what routines are consuming the most time, and then concentrate on those routines (and I've been doing this since v15 - the current DLL is around 5 times faster than the one we had then for the same game position at scale [large, populated map]). Part of the problem with this is that the Civ4 engine profiler isn't very good - one of the first things I did was write a replacement profiler (last June or so) to enable this approach.
Until VERY recently (literally a few days ago) AI unit path-generation calculations were the largest time consumers, but I have now got that down below a couple of other contributors (AI diplomacy and city-turn-calculations aboue joint top now), so I'll be moving on to the new hot-spots next (probably not until v23, as I want to do a bunch of AI work in the next few weeks).
kinda off topic but is it possible to be able to right click, move cursor here ( so i can see what a stack of units consists of horizontally across the bottom of the screen)
Because they are listed vertically I can never see it properly.
Is there a way to do this now maybe with the BUG option thingy
Because they are listed vertically I can never see it properly.
Is there a way to do this now maybe with the BUG option thingy
I don't think we can control the location/style of hover texts, only their content. AIAndy may know more on this one. If you can think of a more useful way to precis the information within the existing box style however, we an definately modify the content that goes into it.
BTW - we need a 'UI suggestions' thread.
can you have hover texts within hover texts? maybe just a simple list which can be expanded by individual hovers 
lol
Or maybe even like the shift hover when your looking at your survival odds of attack and it gets expanded by holding down the shift key.
lolOr maybe even like the shift hover when your looking at your survival odds of attack and it gets expanded by holding down the shift key.
can you have hover texts within hover texts? maybe just a simple list which can be expanded by individual hovers
Or maybe even like the shift hover when your looking at your survival odds of attack and it gets expanded by holding down the shift key.
It doesn't get expanded - it gets replaced. All that shift does (or other modifier keys) is change the generated hover text to go in the provided hover box. We can certainly use modifier keys, but the 'new' content cannot depend on where the cursor was in the old hover text in any way.
Have a very basic list of units and with shift it can be expanded to what it is currently.
Another suggestion might be to try and change the format of the list somehow like bunch it up alot more, not have each unit having its own line etc
I think it already says things like 'archer (25)' to indicate 25 archers after the list grows beyond a certain point. I'll look into it at some point and try to tune it a bit. Suggest you start an 'AI obsrvations & Requests' thread for this sort of thing so we have one place to look for outstanding work.
