112 Comments

[Txurce]

After a layoff and two straight GotVEM games on King, I just played my first game on Immortal using v112. I used Montezuma for a Science victory.

The AI expands rapidly enough that an NC start doesn't seem to be a good idea (as it is on Emperor). I don't have a problem with this, but wonder if scaling the AI production bonus a la the recent Science scaling would make the game start with less of a jolt, and provide more play options on higher levels.

As intended, AI cities are not metastasizing.

The Science buff is not giving the AI a noticeable gain. (It didn't in GotVEM, either.) It's worth considering buffing it some more.

AI pillaging is an undoubted success - one of the best changes VEM has made. It forces a player to have a real army to avoid a potential major economic setback, even if intending to play a peaceful game.

 

[EsoEs]

I'm waiting for the Friday release before I update, 112 seems to have a few kinks to work out. Its interesting that you didn't observe a noticeable science gain, I would have thought a high percentage like that would have been much more noticeable! I aqgree and love the pillage mechanics, it really forces the human player to engage the AI; rather than sitting back and picking them apart from range for a large number of turns, the human is much more encouraged to engage in melee combat in order to protect/reclaim lost luxuries and improvements.

Did you find yourself building more "strategic" roads around your cities to protect from pillagers, or is that something you have always done, or did you not do that at all?

 

[Txurce]

Its interesting that you didn't observe a noticeable science gain, I would have thought a high percentage like that would have been much more noticeable!

Did you find yourself building more "strategic" roads around your cities to protect from pillagers, or is that something you have always done, or did you not do that at all?

I was surprised by the lack of a noticeable tech gain by the AI, but it was based on three games at two levels, so it seems to be the case. I'm looking forward to Thal's view on this.

More strategic roads is probably what I should be doing, but I have a bad habit of applying the Japanese "just in time" production model to my entire Civ game. It blows up in my face almost every game here or there.

 

[Gamewizard]

Playing GotVEM 2 myself, also like the pillaging.

Did you notice if the AI built walls in your Immortal game? They are lacking in the GotVEM and thus I am rolling a city roughly every 3 turns.

 

[Txurce]

Big point - I forgot to mention that. The AI does seem to have slacked off on wall-building, especially in non-capitals. Combined with the weaker city defenses, it seems too easy to take the smaller AI cities now. (I've mentioned that already.)

In my Immortal game I only took two AI cities (playing for Science) but only noticed walls in a CS. There may have been more, of course.

 

[Thalassicus]

I believe the production scaling similar to science scaling made it into v112, though I might not have added that until afterwards. You can check in WWGD/HandicapInfos.xml. Keep in mind the AI gets more science but less growth, which have counteractive effects in terms of research pace (but affect other things in the game significantly).

On Monday or Tuesday I increased the AI starting bonuses slightly, and reduced the long-term acceleration, to provide somewhat more of a benefit in the early game. I've also been working on lots of AI decision-making improvements for the next beta.

In v112 I reduced the FLAVOR_CITY_DEFENSE of AIs from 5 to 4, which had a much bigger effect than intended. In the next version, among other changes I've included these for the Leader Personalities:

FLAVOR_CITY_DEFENSE =

• 4 : "Conqueror"
• 5 : "Coalition"
• 6 : "Expansionist"
• 7 : "Cultural"

This should hopefully provide a nice balance between defending with units (militaristic leaders) vs defending with walls (peaceful leaders). I'm doing testing to see how this works out.

 

[Seek]

The AI's ability to defend with units is questionable, though. How many times have we all been in a situation like this: the AI dows you, sends the vast majority of it's army at you, you defeat it and move in to take it's cities. Keeping the warmonger civs from building defensive buildings will make the last stage easier, and seems like it'll be a shot in the foot for what you're trying to achieve to me. (Actually, I'm not sure what your goal is here; I've found the AI's defensive building rate sufficient in the past.)

 

[Txurce]

I believe the production scaling similar to science scaling made it into v112, though I might not have added that until afterwards. You can check in WWGD/HandicapInfos.xml. Keep in mind the AI gets more science but less growth, which have counteractive effects in terms of research pace (but affect other things in the game significantly).

On Monday or Tuesday I increased the AI starting bonuses slightly, and reduced the long-term acceleration, to provide somewhat more of a benefit in the early game. I've also been working on lots of AI decision-making improvements for the next beta.

I started a v114 game, which presumably has all this. I agree that at least the initial goal is to rebalance (rather than buff) the AI. But why would you pump up their opening, if there's a consensus that the AI is most out of whack at the start, but then falters in the end game?

In v112 I reduced the FLAVOR_CITY_DEFENSE of AIs from 5 to 4, which had a much bigger effect than intended. In the next version, among other changes I've included these for the Leader Personalities:

FLAVOR_CITY_DEFENSE =

• 4 : "Conqueror"
• 5 : "Coalition"
• 6 : "Expansionist"
• 7 : "Cultural"

This should hopefully provide a nice balance between defending with units (militaristic leaders) vs defending with walls (peaceful leaders). I'm doing testing to see how this works out.

In v114 I have noticed that cities are harder to take, which was a welcome surprise. I like the thought behind the variation in city defenses, even if Seek proves to be right and you wind up adjusting it. There's no doubt the AI in general isn't building as many walls as it should.

 

[Thalassicus]

the AI dows you, sends the vast majority of it's army at you

We can control it with this variable:

AI_STRATEGY_MILITARY_RESERVE_PERCENTAGE

In vanilla it's 35%, changed to 20% in WWGD. I've been testing it out back at 35% for the next beta, and could adjust it further up or down.

@Txurce
The problem was 0% early AI bonuses made the game much too easy if we have a strong start. From testing out Game 2 I feel I swung the pendulum too far the other way, so now I've shifted to a midpoint. The goal is for human progress to roughly match AI progress.

 

[Txurce]

The problem was 0% early AI bonuses made the game much too easy if we have a strong start. From testing out Game 2 I feel I swung the pendulum too far the other way, so now I've shifted to a midpoint. The goal is for human progress to roughly match AI progress.

Agreed - and I think we are really tackling it the right way, in all these areas (including adjusting the strategic military reserve). The latter may be another way to further individualize the AI, by the way. Consider that while you don't want patsies, there's something to be said for different levels of AI performance (which is inherently true anyway, given the different civ traits).

On a more specific note, I don't think the AI leaves itself wide open to other AI to a damaging degree by throwing most of its units against one target. The human player is more likely to take advantage of this, but I think that's pretty fair. Bottom line, I'm not sure that reserve adjustment is needed.

 

[Ahriman]

I would tend to lean more towards the throwing its army against one target.
I think the biggest AI combat problem is that it doesn't have *quite* enough force to take a city, but loses most of its army in attempting to do so.

The downside of being slightly more vulnerable seems low relative to the potential downside of not being able to prosecute a successful offensive.

 

[Txurce]

I would tend to lean more towards the throwing its army against one target.
I think the biggest AI combat problem is that it doesn't have *quite* enough force to take a city, but loses most of its army in attempting to do so.

The downside of being slightly more vulnerable seems low relative to the potential downside of not being able to prosecute a successful offensive.

This is a gray area, but I completely agree. The AI's biggest problem is hesitation. An excess of units are usually needed for the AI to take a city.

 

[Ahriman]

I agree that it is a gray area; if the AI's invasion is not successful, then just a few ranged units in its cities will have a big impact on its ability to fend off a counterattack.

Is there any way to affect the composition of the reserves? Eg to set a reserve amount to be small but limited to ranged units?

 

[Txurce]

City defenses update: playing on v114, Washington built Pioneer Forts in his second and third cities before I declared early war on him (with Germany).

 

[Thalassicus]

Garrisons are typically chosen from ranged units. I don't remember the name of the variable, but did see that somewhere. It appears to match AI behavior.