Killed First Enemy Settler!

Carazycool

Warlord
Joined
Feb 25, 2007
Messages
249
Location
MA
I am playing as Gandhi on a fully crowded standard map (King difficulty). I moved my warrior to find that Alexander's settler was only 4 hexes away from my settler. I moved my settler 1 hex toward Alexander and founded my city. Being only 3 hexes away, Alexander couldn't found his first city on the tile that his settler spawned on.

Strangely enough, Alexander didn't move his settler and found a city. He left his settler right where it was, doing nothing, and he didn't even guard his settler with his warrior. The next turn, I had a clear shot at his settler, so I declared war and took it. I then took out his warrior with my warrior. However, he wasn't knocked out of the game. Does he have another unit running around somewhere? Or did he never "lose" his capital and therefore will never be eliminated and always in the game to make diplomatic deals?

Has this ever happened to anyone before? I've posted the initial autosave if people are interested. I am playing on a mac.
 
I am playing as Gandhi on a fully crowded standard map (King difficulty). I moved my warrior to find that Alexander's settler was only 4 hexes away from my settler. I moved my settler 1 hex toward Alexander and founded my city. Being only 3 hexes away, Alexander couldn't found his first city on the tile that his settler spawned on.

Strangely enough, Alexander didn't move his settler and found a city. He left his settler right where it was, doing nothing, and he didn't even guard his settler with his warrior. The next turn, I had a clear shot at his settler, so I declared war and took it. I then took out his warrior with my warrior. However, he wasn't knocked out of the game. Does he have another unit running around somewhere? Or did he never "lose" his capital and therefore will never be eliminated and always in the game to make diplomatic deals?

Has this ever happened to anyone before? I've posted the initial autosave if people are interested. I am playing on a mac.

Haven't had it with a first settler, but I think people have noticed that the AI has trouble changing a fixed plan, and if something's in the way of its settler etc. it will just stop. I was at war with the Ottomans (their declaration) and had wiped out their units attacking Argos; they left their last unit, a settler that had previously had a pikeman with it, just sitting there for several turns for me to grab.
 
However, he wasn't knocked out of the game. Does he have another unit running around somewhere? Or did he never "lose" his capital and therefore will never be eliminated and always in the game to make diplomatic deals?

On King the AI gets a second warrior unit. You have to destroy it and then he´s definitely out.
 
Haven't had it with a first settler, but I think people have noticed that the AI has trouble changing a fixed plan, and if something's in the way of its settler etc. it will just stop. I was at war with the Ottomans (their declaration) and had wiped out their units attacking Argos; they left their last unit, a settler that had previously had a pikeman with it, just sitting there for several turns for me to grab.

Today's game, on Emperor - ran into the Venetian settler on my first turn (I didn't even know city-states started as settlers). When I logged off just now, several millennia later, it was still standing there (I'm waiting for someone to request the destruction of Venice - I don't know if it will happen when the city only exists as a settler, but the standard city-state contact info comes up when I click on it). I presume my Greek settler spawned in the spot it wanted to take, or otherwise its target was within my new borders.
 
The AI is preprogrammed to the very last detail (makes sense, AIs have no intelligence and are just a program).

All civilizations start with a unit...well, it can be anything, but for everyone it's defined as UNIT_SETTLER. (obviously, a settler). They also have a unit AI attached to that unit, and that unit will always follow that AI. In this case, it's UNITAI_SETTLE, so they'll settle in place. If they cannot settle in place, they have no other tasks but to keep trying to settle. That's why the AI didn't try to move away. They aren't programmed to move away. They're programmed to settle, but that option is disabled and they're stuck in an infinite loop, per se.

And yes, it has happened to me before (city-state) and it will eliminate them from the game if they have no settlers left.

Was "Complete Kills" active?
 
The AI is preprogrammed to the very last detail (makes sense, AIs have no intelligence and are just a program).

All civilizations start with a unit...well, it can be anything, but for everyone it's defined as UNIT_SETTLER. (obviously, a settler). They also have a unit AI attached to that unit, and that unit will always follow that AI. In this case, it's UNITAI_SETTLE, so they'll settle in place. If they cannot settle in place, they have no other tasks but to keep trying to settle. That's why the AI didn't try to move away. They aren't programmed to move away. They're programmed to settle, but that option is disabled and they're stuck in an infinite loop, per se.

And yes, it has happened to me before (city-state) and it will eliminate them from the game if they have no settlers left.

Was "Complete Kills" active?

I don't know the Complete Kills option - I only set the basic settings (Map: Earth, Size: Random, Civilization: Greece, Difficulty: Emperor). This settler actually had no associated Warrior.

And preprogramming seems a bad way in principle to design an AI. I understand most game AIs are programmed with logical operators - "If situation X, then perform action Y". I remain unclear, though, why learning algorithms can't be used; see for example the link on the Steam page to a study that scanned the manual for a learning-based AI to read, following which it was able to win with a 70-80% success rate on that difficulty level (albeit presumably against standard Civ AIs). Even a limited ability to learn from players' responses, as I believe the Starcraft AI has, would be a distinct improvement - if it sees, say, that the player has more success if he doesn't declare war on everyone simultaneously, if he doesn't stand Warriors around outside archer-garrisoned cities, or if he supports siege engines by placing melee infantry in front of them, it could learn to adapt. If it observes the player taking a city, for example, and it observes that the player does so with fewer losses and/or more quickly than the AI expects to, then it would adopt that superior strategy.

As for elimination, what I was asking was "Will a City-State issue a 'We want Venice exterminated' quest if Venice is just a settler?" No one's requested Venice's destruction yet.
 
The "complete kills" option was not selected. but i guess if the player never founds a city, then "complete kills" is in effect for them. after a while i guess someone killed alexander's lone remaining warrior, and he disappeared from the diplomacy screen.
 
PhilBowles said:
And preprogramming seems a bad way in principle to design an AI. I understand most game AIs are programmed with logical operators - "If situation X, then perform action Y". I remain unclear, though, why learning algorithms can't be used; see for example the link on the Steam page to a study that scanned the manual for a learning-based AI to read, following which it was able to win with a 70-80% success rate on that difficulty level (albeit presumably against standard Civ AIs). Even a limited ability to learn from players' responses, as I believe the Starcraft AI has, would be a distinct improvement - if it sees, say, that the player has more success if he doesn't declare war on everyone simultaneously, if he doesn't stand Warriors around outside archer-garrisoned cities, or if he supports siege engines by placing melee infantry in front of them, it could learn to adapt. If it observes the player taking a city, for example, and it observes that the player does so with fewer losses and/or more quickly than the AI expects to, then it would adopt that superior strategy.

I believe the AI chooses a tile in which to found a city before actually going to settle. They'll never change this tile after sending the Settler out. If the default tile is the civ starting point, then that's where the AI wants to settle. :)

You can't do learning AIs in games efficiently because (in the case you were describing) the AI took about 7-8 minutes to process a turn. Which would make this game unplayable. (50+ minute turn loads?) The processing required for this game and Starcraft is completely different (we have tiles, they have a free roaming environment IIRC).
 
... preprogramming seems a bad way in principle to design an AI. I understand most game AIs are programmed with logical operators - "If situation X, then perform action Y". I remain unclear, though, why learning algorithms can't be used; see for example the link on the Steam page to a study that scanned the manual for a learning-based AI to read, following which it was able to win with a 70-80% success rate on that difficulty level (albeit presumably against standard Civ AIs). Even a limited ability to learn from players' responses, as I believe the Starcraft AI has, would be a distinct improvement ....

I believe the AI chooses a tile in which to found a city before actually going to settle. They'll never change this tile after sending the Settler out. If the default tile is the civ starting point, then that's where the AI wants to settle. :)

You can't do learning AIs in games efficiently because (in the case you were describing) the AI took about 7-8 minutes to process a turn. Which would make this game unplayable. (50+ minute turn loads?) The processing required for this game and Starcraft is completely different (we have tiles, they have a free roaming environment IIRC).

While efficient learning is impossible for the reasons that Putmalk states, the situation with settlers should be much simpler than this. The programming comes down to asking the following questions for each unit at the beginning of its turn:

Do I have an assigned mission? (e.g., settle on a specific tile)
Is that mission impossible? (e.g., said tile is too close to another city)
If yes, then assign me another mission. (e.g., settle on a different tile)

These steps are programmable and do not really involve learning anything....
 
I believe the AI chooses a tile in which to found a city before actually going to settle. They'll never change this tile after sending the Settler out. If the default tile is the civ starting point, then that's where the AI wants to settle. :)

You can't do learning AIs in games efficiently because (in the case you were describing) the AI took about 7-8 minutes to process a turn. Which would make this game unplayable. (50+ minute turn loads?) The processing required for this game and Starcraft is completely different (we have tiles, they have a free roaming environment IIRC).

An alternative solution is not to use the learning AI in the game, but to use it in game development, in much the same way evolutionary neural nets have been used experimentally to refine AI performance.

So, for each difficulty level with preset AI opponents, you'd use a learning-based AI, show it the manual as in the MIT experiments, and have it play games. You could presumably then extract the 'evolved' code that results and use that as the game AI, without the learning algorithm itself. Similarly, you could train an AI to use superior strategies by playing it against humans who outperform it in, e.g., military success, allow it to learn the associated strategies, and then write a finalised, non-learning algorithm that incorporates those 'lessons learned'.
 
While efficient learning is impossible for the reasons that Putmalk states, the situation with settlers should be much simpler than this. The programming comes down to asking the following questions for each unit at the beginning of its turn:

Do I have an assigned mission? (e.g., settle on a specific tile)
Is that mission impossible? (e.g., said tile is too close to another city)
If yes, then assign me another mission. (e.g., settle on a different tile)

These steps are programmable and do not really involve learning anything....

Indeed. CIV IV's AI could do this, why can't V? I mean it's not like the code for this would be that different. Thing is while it's true that IV's AI is dumb as a rock, there are things it could pull of that V's can't.
 
As a comment to the original post, I've had several instances where I wiped out an AI before he could found his first city and in the victory conditions screen it does not show that he has lost his capital (presumably because he never had one) and thus I can never win a domination victory on those games.
 
Just went back to the game I mentioned after a break (taking another break now all-out war is on between me, Arabia, and the remaining city-states, which are split between both civs). Ragusa (not Venice after all) did, in fact, eventually move its settler, though only to settle on a small peninsula with no land - needless to say it couldn't build any units and I destroyed it as soon as Cape Town demanded it (much more to my chagrin, I lost Kuala Lumpur to a Japanese assault).
 
And yes, it has happened to me before (city-state) and it will eliminate them from the game if they have no settlers left.

If you lose your initial settler but still have your warrior you are not eliminated from the game. Seriously, try to explore with both units until a barb captures your settler. You aren't eliminated. This might be partly because you can recapture the settler still. I think the OP should try deleting that settler that became a worker that was captured and see if that takes them out.


Just went back to the game I mentioned after a break (taking another break now all-out war is on between me, Arabia, and the remaining city-states, which are split between both civs). Ragusa (not Venice after all) did, in fact, eventually move its settler, though only to settle on a small peninsula with no land - needless to say it couldn't build any units and I destroyed it as soon as Cape Town demanded it (much more to my chagrin, I lost Kuala Lumpur to a Japanese assault).


Was the settler forced to move because someone's territory expanded onto it?
 
I think the most important point that can be made regarding circumstances like this is that the game still needs fixes. People get excited about the next possible DLC civ or an expansion that will give them religion or espionage. I've never understood why more folks don't demand the flaws get patched out first. It's been 16 months since the game was released. I appreciate the fact that Firaxis has continued to work on the game, but too many bugs remain. There are still frequently occurring bugs that effect tile yields, map scripts, combat results and diplomacy. Even bugs that don't occur frequently, but significantly change the quality of the game (someone said they couldn't achieve a victory because one civ was entirely eliminated) are unacceptable.
 
Top Bottom