New Beta Version - November 9th (11-9)

[Aristos]

I'm honestly not sure why any of those changes were made. It was my impression that we were in late days for development for this mod. overhauling flanking/siege damage/fort and citadel mechanics all strike me as arbitrary changes that don't necessarily fix anything that wasn't working just fine in its own way before.

As G might say, these are all changes for changes' sake. They're not necessarily worse, once we iron out the kinks, but I just have no idea why anyone felt the need to overhaul these systems.

Yeap... I fear we are threading on scope creep territory now, I hope I am wrong...

 

[pineappledan]

They were changes for debug sake.
Flanking was overly bugged and needed to be rewritten anyway. And someone said something like "By the way, flanking is the only bonus that isn't additive, but multiplicative, is there any reason to keep this distinction? No?" and someone else in the thread "I'm changing that. Done!"

Aight, well even so, changes to combat require a re-assessment of the relative power of these mechanics in comparison to the old system.

Flanking is stronger now, as G put it, but the reality is that flanking is entirely different, and stronger/weaker remains to be seen. This new system deducts bonuses for “counter-flanking”, which, to me, sounds like heavier emphasis is being put on raw numbers of units. You need warm bodies occupying tiles for both offence and Defense. I’m sure humans will do just fine with this, but it’s also very likely that the civ most geared towards “death-carpets”, Zulu, who already gets a lot of extra bonus from flanks, could have been buffed even more than we thought.

Even after we get the needed fix for siege weapons, how does this change to ignoring defenses stack with city assault promotion? Volley? Bombers? Does the CS/RCS of siege weapons need to be re-evaluated since all their numbers were calibrated based on a separate combat mechanic?

I honestly can’t even make heads or tails of the fort/citadel change. The bonuses being granted to horses isn’t a big deal to me, except it crowds in on melee unit’s niche pretty badly. The fact that siege units ignore forts sounds pretty dumb, especially considering the fact that the citadel tile is literally a star fort - a fort design specifically invented to minimize the effectiveness of artillery.

Even if the changes are good, it’s still pulling the temple down on our heads. It’s a crapload if work to reassess combat, and it’s going to take a ton of testing to confirm this hasn’t made a chain of trickle-down balance tweaks to a large swathe of combat mechanics

 

[Gazebo]

Flanking: It was causing major problems for the AI. I consider that a bug fix.

Ignore Building Defense: I agree, and I think the change should be reverted. I wasn’t aware it was in the merge profile other wise I would have left it out.

G

 

[Omen of Peace]

Regarding Flanking, I also remember being utterly baffled by it when starting to play VP. For instance I thought the "+X % while flanking" promotions would actually give me +X %, fool that I was! So the new system is more newbie-friendly (though the "delta = adj. allies - adj. enemies" equation will need to be somewhere where it can be found - I haven't checked Civilopedia).

 

[Stalker0]

Flanking: It was causing major problems for the AI. I consider that a bug fix.

Unless I am mistaken, flanking now accounts for your unit numbers as well as your enemies....whereas before it only accounted for your numbers. So its beyond a bug fix, it changes how flanking works.

 

[Txurce]

Unless I am mistaken, flanking now accounts for your unit numbers as well as your enemies....whereas before it only accounted for your numbers. So its beyond a bug fix, it changes how flanking works.

To me, that's semantics. The work has already been done, with the aim of making the AI more effective. That's all I need to hear, unless the change ruins the game's playability. In my current game, I'm not having a problem with it.

 

[Stalker0]

To me, that's semantics. The work has already been done, with the aim of making the AI more effective. That's all I need to hear, unless the change ruins the game's playability. In my current game, I'm not having a problem with it.

If we are fixing the coding so that the AI can use it, that's a bug fix. If we fix the coding....and modify how flanking works....that's a feature change. Its two separate things

 

[Moi Magnus]

Unless I am mistaken, flanking now accounts for your unit numbers as well as your enemies....whereas before it only accounted for your numbers. So its beyond a bug fix, it changes how flanking works.

Partially wrong. To quote ilteroi on the situation before the change:

you would think more people notice that, but flanking currently is a mess with about 4 or 5 bugs thrown in. most of them i fixed but some are more conceptual issues ...

1. apparently the flanking bonus applies both to the attacker AND the defender. does that make sense?
   
   
2. additionally, there's a flank attack modifier but it's applied in both cases (in c++, not in lua). that definitely seems wrong.

So you had a flanking bonus for the attacker and the defender, which is mostly equivalent to having a smaller flanking bonus for the attacker and nothing for the defender.

 

[SpankmyMetroid]

Re: Building defense, it seems to me like city state defenses are not affected by the bug.

Spoiler :

Compare that to Morocco one turn before and one turn after building an arsenal, respectively. Damage inflicted is the same but the city just has more HP.

Spoiler :

Morocco constantly has a garrison, so that's not the difference. Is there some kind of flag for the city state that makes the ignore defense bug not happen?

 

[nekokon]

I think we're getting confused over which was the bug fix for AI that G mentioned.
I think a few of us thought the bug was "there were both flanking bonus AND flank attack modifier in game, and the fix was to simply change all flank attack modifier into flanking bonus (adding straight onto cs bonus, instead of multiplying). The new flanking mechanic (delta value) doesn't touch the issue above".
Ex: Before, shock I gives 25% flank attack modifier, and the game adds all modifiers together before multiplies it with base flanking bonus (10% from each ally) to get the value shown in game (equal to cs bonus)
After, shock I gives flat 2.5% flanking bonus, and all flanking bonus got added straight into the base flanking bonus (5% delta currently).

However, I don't think it would affect the AI much if this was the case. The new flanking mechanic has taken care of another issue which I think are more relevant to how the AI makes decision, which is removing flanking bonus from defender.
Ex: Old flanking, when 1 unit attacking while standing next to 2 enemies, it gets "flanked" instead, despite being the attacker (pic attached). This caused wonky interaction as the attacking order would matter a lot more, thus pushing the AI into making the wrong decision (they only "think" once at the start of the turns, and does not change decision after each move)

New flanking, with the same condition, both side simply get no flanking bonus (sorry no picture here, I flunked my last game). With this attacking order matters less, which helps the AI.

However, I think the new flanking can be quite illogical as far as reality goes, as an unit facing 2 adjacent enemies can still get the bonus (or neutralize enemy's flanking bonus) with another unit at their rear even if it doesn't come into contact with neither enemies. This should be called "supporting" rather than "flanking". Personally I still prefer the old mechanic, as unit placement are more emphasized than sheer number for flanking bonus. I assume there's no simple way to remove only flanking bonus from the defender with the old system ?

 

[YukiN]

New flanking is straightforward and usable. It helps the AI too. It has already been implemented. I think we are good there.

@Gazebo
Since you mentioned being away, I take it we should not expect a hotfix until next week?

 

[tu_79]

Just to be sure. (Assuming we are attacking)

Old flanking:
Positive for friendly units adjacent to the defender.
Negative for enemy units adjacent to the defender.
Defense is easier.

New flanking:
Positive for friendly units adjacent to the attacker.
Negative for enemy units adjacent to the defender.
This rewards having more units in clusters, with no positioning requirements. AI friendly?

Optional flanking:
Positive for friendly units adjacent to the defender.
Negative for enemy units adjacent to the attacker.
This limits heavily how many units can flank and rewards positioning. Not very AI friendly.

Edit. The first method has a weird situation where one unit attacking an unit that has support behind, gets flanked by the defense, which makes no sense. In other words, defenders can flank more often with enough units.

In the second method, units that are directly behind the units in figth are actually taking part in the flanking, even units behind the unit that is moving, which has no sense either. But AI usually has more units and knows less about positioning than the human, so this favors AI.

In the optional method, units behind the combat line take no part in flanking. Only side by side units account. This is more realistic but on the other hand, AI is at disadvantage.

 

[Questdog]

Why should the :"Optional Flanking" above be less AI Friendly than the others? They are ALL about positioning your units... And even if it is, it is the only method that makes any sense at all....

The two biggest things that make AI fight worse than human: 1) Spreading its forces across the entirety of the map instead of concentrating them together before any war starts. 2) Seeing any attack where its unit can deal more damage than it takes in the attack as a good move.

 

[Grassland Farm]

Happiness looks much easier to manage in this version! Maybe even too much, I don't even want to work Specialists because I'm always so high on happiness I'd rather let my cities grow even higher.
I'll see how it turns out in later eras (just hit Industrial). Also, it seems defended cities are much tougher to capture, and healing in enemy lands was halved - I like those changes.

 

[nekokon]

Just to be sure. (Assuming we are attacking)

Old flanking:
Positive for friendly units adjacent to the defender.
Negative for enemy units adjacent to the defender.
Defense is easier.

It's negative for enemy units adjacent to the attacker.

New flanking:
Positive for friendly units adjacent to the attacker.
Negative for enemy units adjacent to the defender.
This rewards having more units in clusters, with no positioning requirements. AI friendly?

Positive for friendly units adjacent to the defender.

Optional flanking:
Positive for friendly units adjacent to the defender.
Negative for enemy units adjacent to the attacker.
This limits heavily how many units can flank and rewards positioning. Not very AI friendly.

This is the actual old flanking (which is nice and working, except for the AI).

Why should the :"Optional Flanking" above be less AI Friendly than the others? They are ALL about positioning your units... And even if it is, it is the only method that makes any sense at all....

The two biggest things that make AI fight worse than human: 1) Spreading its forces across the entirety of the map instead of concentrating them together before any war starts. 2) Seeing any attack where its unit can deal more damage than it takes in the attack as a good move.

I've already explained why the old flanking (3rd method) is bad for the AI in my previous post. Human player can take advantage of the fact that when you're locked in melee (2v2 or more), it's better to use the unit not being flanked to attack first, potentially killing 1 of the enemies (thus removing the flank bonus for enemy when your 2nd unit - previously flanked, now free - attack).
It's the same with moving your unit to surround the target before attacking, so that even the 1st attack can benefit from all flanking bonus possible.

The AI only make decision right at the start of the turn, and will not change its moves if something unexpected happened, thus always pick the safe move which can be non optimal (or sometime making outright dumb mistake because of FoW)
New flanking won't fix the root of the problem (AI can't make decision on the fly, costing too much computing resource), but it helps with some situations. Favoring number is just a by product of this new mechanic (which also benefits the AI with their unit spam).

 

[Questdog]

Having a mechanic that is less than optimal for the AI is better than having a mechanic that makes no sense... And I still do not buy that the new way is any better for the AI, but it is definitely easier for the player; you do not need to flank to get a flank....

 

[tu_79]

It's negative for enemy units adjacent to the attacker.



Positive for friendly units adjacent to the defender.



This is the actual old flanking (which is nice and working, except for the AI).



I've already explained why the old flanking (3rd method) is bad for the AI in my previous post. Human player can take advantage of the fact that when you're locked in melee (2v2 or more), it's better to use the unit not being flanked to attack first, potentially killing 1 of the enemies (thus removing the flank bonus for enemy when your 2nd unit - previously flanked, now free - attack).
It's the same with moving your unit to surround the target before attacking, so that even the 1st attack can benefit from all flanking bonus possible.

The AI only make decision right at the start of the turn, and will not change its moves if something unexpected happened, thus always pick the safe move which can be non optimal (or sometime making outright dumb mistake because of FoW)
New flanking won't fix the root of the problem (AI can't make decision on the fly, costing too much computing resource), but it helps with some situations. Favoring number is just a by product of this new mechanic (which also benefits the AI with their unit spam).

Thanks. I wasn't sure of which was the actual mechanic.

The old one seems to be the right one, even if it favors the attacker. Hit first, hit twice.
The reason for AI not knowing how to plan for it, I'm not that convinced. AI has handicaps for a reason. When it's too hard for an algorithm to replicate human behavior, we give G handicaps to AI. But for human vs human fights, the mechanic seems wrong now. Having units that are not adjacent to where the figth happens, providing flanking has little sense.

Surely, as Questdog says, AI can be told to favor positioning in the front line (form a melee wall).

 

[Gazebo]

New flanking is straightforward and usable. It helps the AI too. It has already been implemented. I think we are good there.

@Gazebo
Since you mentioned being away, I take it we should not expect a hotfix until next week?

DIY hotfixes have already been shared elsewhere in this thread. It’s just a matter of copy paste.

G

 

[Moi Magnus]

As a flanking bonus, it should reward having units multiple attacking units adjacent to the defender.
However, I'm not decided on how it should work for the defender.
Should we reward units adjacent to each others, or should we have "defensive flanks".

Solution 1:
Positive for allied units adjacent to the target.
Negative for ennemy units adjacent to the target.
=> Significantly easier for the AI, since he don't have to try to guess about "how the player will place its troups"

Solution 2:
Positive for allied units adjacent to the target.
Negative for ennemy units adjacent to the attacker.
=> The old way. (Or the way the old way should have worked if it was not bugged)

 

[nekokon]

Thanks. I wasn't sure of which was the actual mechanic.

The old one seems to be the right one, even if it favors the attacker. Hit first, hit twice.
The reason for AI not knowing how to plan for it, I'm not that convinced. AI has handicaps for a reason. When it's too hard for an algorithm to replicate human behavior, we give G handicaps to AI. But for human vs human fights, the mechanic seems wrong now. Having units that are not adjacent to where the figth happens, providing flanking has little sense.

Surely, as Questdog says, AI can be told to favor positioning in the front line (form a melee wall).

Yea, the old mechanic makes more sense and encourage better tactical play. I'd take it over new flanking even with a bit weaker tactical AI. Imho, if you have to dumb down some gameplay element to help the AI, I'd rather give them advantage on other things to counter balance instead. It's gamer's satisfaction to overcome raw power difference with creative solution (glorified term for exploit ).

To clarify, old flanking doesn't favor attacker, it favor attacking order (which can only be utilized by the attacker). Attacking in the wrong order can favor the defender too.

The root of this issue (AI can't change decision on the fly) was discussed somewhere in here a few months ago, and ilteroi confirmed it was how the AI works (either hard coded or take up too much computing resource). We can only try to go around it.

If there's anyway to add a check "if flanking bonus for attacker is smaller than flanking bonus for defender, negative all flanking bonus" to the old flanking, it'd take care of the issue when defender has higher flanking bonus, which I think was the main reason for the new flanking. Of course it'd increase computing resource, but I can't think of a more elegant solution (well, except for renaming the new flanking "combat support" instead).