February 15th, 2017, 09:00
Posts: 222
Threads: 2
Joined: Dec 2016
by the way i noticed on autocombat that casters in an already all-invisible stack tend to often cast mass invisibility, is this working as intended? (i can see the merits in having a second layer of invisibility, but maybe the priority should be reduced?)
February 15th, 2017, 09:51
Posts: 10,463
Threads: 394
Joined: Aug 2015
(February 15th, 2017, 09:00)Domon Wrote: by the way i noticed on autocombat that casters in an already all-invisible stack tend to often cast mass invisibility, is this working as intended? (i can see the merits in having a second layer of invisibility, but maybe the priority should be reduced?)
I considered adding this but when I saw how brutally effective combat summons can be if they come into play as already invisible, I decided it's worth wasting the 80 mana just for a chance of that. (invisible paladins, phantom beasts, earth elementals...)
Not only does it ensure these melee summons can make their way to your units without getting picked off by ranged attacks, but it even provides them protection in melee combat. (invisibility is -1 enemy to hit in melee)
All-invisible stacks are rare and usually small, but those benefit the most from being able to add invisible summoned units.
February 15th, 2017, 12:14
Posts: 10,463
Threads: 394
Joined: Aug 2015
I observed an AI settler target a very far away crappy tile (first settler in the game) for an unknown reason - a shore tile with 8 pop and coal near my capital. (there was adamant and gold and mithril with pop 12+ in between). Unfortunately I didn't have the earlier save file to reproduce it. Stopping the settler causes it to find the correct destination, so whatever triggered the bug was temporal for that one turn only when the destination was selected.
I have absolutely no idea why it went there, if anyone encounters the same situation and has the save file for the first turn (or at least the turn before the settler left the capital), let me have it.
February 15th, 2017, 13:21
Posts: 5,010
Threads: 17
Joined: Aug 2016
I'm not sure (and I don't have save files) but I've had a suspicion this has been happening in my games often enough to be noticeable. I'll keep an eye out. Actually my hero game night have one if I go back to the start...
I'll see if I can find it.
February 15th, 2017, 14:08
Posts: 5,648
Threads: 30
Joined: Mar 2014
(February 15th, 2017, 13:21)Nelphine Wrote: I'm not sure (and I don't have save files) but I've had a suspicion this has been happening in my games often enough to be noticeable. I'll keep an eye out. Actually my hero game night have one if I go back to the start...
I'll see if I can find it.
Same, I've reported this before too. I think it's related to why Lizardmen can be such an incredible pest in the early game, when you see them waltz past dozens of tiles just to settle 3 tiles from your capital.
Idea: I wonder if there's an indexing bug, that's causing the settler to look for good sites around player 0's capital (the human) instead of player i's?
A hack fix might be to always stop settlers once, so that they'll then attempt to find the correct site on their next turn?
February 15th, 2017, 14:40
Posts: 10,463
Threads: 394
Joined: Aug 2015
(February 15th, 2017, 14:08)GermanJoey Wrote: (February 15th, 2017, 13:21)Nelphine Wrote: I'm not sure (and I don't have save files) but I've had a suspicion this has been happening in my games often enough to be noticeable. I'll keep an eye out. Actually my hero game night have one if I go back to the start...
I'll see if I can find it.
Same, I've reported this before too. I think it's related to why Lizardmen can be such an incredible pest in the early game, when you see them waltz past dozens of tiles just to settle 3 tiles from your capital.
Idea: I wonder if there's an indexing bug, that's causing the settler to look for good sites around player 0's capital (the human) instead of player i's?
Nope, the settler is looking for good sites closest to themselves. I've double checked the variables used for coordinates and it seemed correct.
Funny thing is, I started about 10 games with the map revealed and couldn't spot a single settler with the bug, so it must be pretty rare - yet it seems to come up in games played for real anyway.
Quote:A hack fix might be to always stop settlers once, so that they'll then attempt to find the correct site on their next turn?
Now this is something worth considering. Settlers are the only exceptions to the "each unit reroutes to the best target every turn" rule, because the old settler movement system didn't support that. the new one however does (which is why it found the right spot after I stopped it). The only (pretty big) flaw I see is, if the tile they select is blocked, and the next best tile is in the opposing direction, they'll move back and forth between the two as long as there is enough traffic of random nonfriendly units on both tiles. If the player learns to pinpoint the best tile, they can put two spearmen on them and keep the settler from ever doing anything - if done to both settlers the AI can have on their starting continent, it can prevent making cities altogether for the entire game.
(note this does not work if the settler only stops right before entering the tile with your spearmen - the difference in distance and the generally good terrain around that one tile would make them pick an adjacent free tile unless none exists, or possible even build right where it stopped.)
February 15th, 2017, 21:28
Posts: 5,648
Threads: 30
Joined: Mar 2014
Maybe just stop and reroute on every fourth turn?
February 20th, 2017, 02:49
Posts: 6,457
Threads: 134
Joined: Aug 2004
Would it be possible to teach the AI to not send its doom stack against a single spearman that I put there to waste his turn? This is a very powerful strategy that costs nothing to execute and gives me one or two critical turns more to defend against an attack.
February 20th, 2017, 06:34
(This post was last modified: February 20th, 2017, 06:35 by Nelphine.)
Posts: 5,010
Threads: 17
Joined: Aug 2016
Catwalk, what's the line? For instance, in my hero game, I'm running around with one lone hero. So it can't be just stacks of one unit. Or my 'normal' strategy is halberdiers (I've killed an enemy fortress with 3 halberdiers, although they had 4 life buffs). So it can't be cost of the units.
You could make it literally spearmen, but then you just use swordsmen, etc.
I know for myself, I simply don't use the tactic (I generally avoid any bait tactics, nor mana abuse tactics like attacking with 9 lone spearmen against a death wizard with wave of despair before making my actual attack.)
February 20th, 2017, 07:09
Posts: 10,463
Threads: 394
Joined: Aug 2015
(February 20th, 2017, 02:49)Catwalk Wrote: Would it be possible to teach the AI to not send its doom stack against a single spearman that I put there to waste his turn? This is a very powerful strategy that costs nothing to execute and gives me one or two critical turns more to defend against an attack.
In theory, each stack picks the highest value reachable target the stack can beat. If it's going after the spearmen, it either means it cannot reach anything else, or that everything else is too strong to attack using that stack - distance is used in scaling value though, but it shouldn't be enough of a factor to outweight a city.
If that doesn't happen, it might be a bug, give me a save with specific details on which stack is attacking what instead of what else.
Note that each target can only be chosen by 1 land stack - unless there are no more targets and the distance is very low. If another doom stack is already heading towards your city, the other will attack the spearmen - regardless of which "doomstack" is weaker. You mentioned using Spermen to defend, your city might be targeted by a weak 2-3 unit stack as it has more power then the spearmen, while the 9 stack goes after whatever else it can find instead.
|