Better unit iPowers

[MartinHarper]

The standard unit iPowers have some flaws. The main one is that they don't fully account for the benefit of high relative strength. I'm not sure whether jdog intends to fix this in the xml or in the dll. I suppose fixing it in the dll has the advantage of backwards-compatability for mod-makers.

There are also some power assignments that I regard as dubious in other ways, aside from the scaling. Some are over-rated for power:

• Dog Soldier : 6; should be 5 - axes are better, if you have copper
• Paratrooper : 30; should be 25 - the ability to paradrop justifies +1 power, but not +6.
• Marine: 28; should be 26 - the attack bonuses and free promotion justifies +2 power, but not +4. Navy SEALs are ok on 28.
• Tank: 30; should be 28 - the lack of defence bonuses cancels out the movement and blitz. Panzers are ok on 30.

Most siege units are under-rated for power, with the possible exception of Trebuchet:

• Mobile Artillery: 26; should be 40 (like modern armor)
• Artillery: 22; should be 26 (like gunships)
• Cannon: 12; should be 15 (like cavalry)
• Hwacha: 5; should be 7 (better than axes)
• Catapult: 5; should be 6 (like axes)
• Chokonu: 7; should be 8 (better than crossbows)

Other under-rated units for power include:

• Exporer: 2; should be 4 (like chariots - though not worth fixing unless ^1.5 code is added)
• Landsknecht: 6; should be 7 (better than axes and pikes)
• Conquistador: 12; should be 13 (better than Cuirassiers)
• Mobile SAM : 22; should be 25 - the ability to intercept aircraft and the bonus against helicopters is worth about +3.
• SAM Infantry : 18; should be 20 - the ability to intercept aircraft and the bonus against helicopters is worth about +2.
• Anti-Tank : 14; should be 16 - the ability to intercept aircraft and the bonus against tanks is worth about +2.
• Impi: 4; should be 5 (better than spears)
• War Chariot: 4; should be 5 (better than chariots)
• Machine Gun: 14; should be 16 (immunity to collateral mostly makes up for the inability to attack)

 

[PieceOfMind]

Dog Soldier : 6; should be 5 - axes are better, if you have copper

Careful. Axes are not always better. To see this, try attacking a dog solider on flat ground using an axe. Then put an enemy axe on a forest and see whether you have better chances of killing it with a dog soldier or an axe.

The point is, which is better is situational. The details of all the whens and whys I won't go into. My gut feeling would be that having the same ipower as an axe is appropriate.

I agree with most of your other suggestions though.

 

[MartinHarper]

The point is, which is better is situational.

This is true for all units. Axes are better when in combat with non-melee units. Even against melee units, axes are better when they have a 51% or better bonus due to defence bonuses or city raider. To me that justifies axes having an extra point of iPower.

 

[Yakk]

[*]Dog Soldier : 6; should be 5 - axes are better, if you have copper

4 +100%?

In any case, you'll note that the Dog Soldier being crappy is not intended. And this is a 17% difference.

[*]Paratrooper : 30; should be 25 - the ability to paradrop justifies +1 power, but not +6.

20%difference.

[*]Marine: 28; should be 26 - the attack bonuses and free promotion justifies +2 power, but not +4. Navy SEALs are ok on 28.

8% difference.

[*]Tank: 30; should be 28 - the lack of defence bonuses cancels out the movement and blitz.

No, the lack of defences bonuses doesn't cancel out the movement and blitz. Tanks really change the game when they arrive -- you go from a fight between infantry to a completely new kind of unit.

Look at knights -- Tanks are super-knights, because the anti-Tank technology is behind the Tank and not as large an edge as Pikemen where.

Also of importance is this is an 8% difference.

[*]Mobile Artillery: 26; should be 40 (like modern armor)
[*]Artillery: 22; should be 26 (like gunships)
[*]Cannon: 12; should be 15 (like cavalry)
[*]Hwacha: 5; should be 7 (better than axes)
[*]Catapult: 5; should be 6 (like axes)

Sure; but barring Mobile Artillery, none of these changes are worth doing and maintaining an XML file delta.

[*]Chokonu: 7; should be 8 (better than crossbows)

Ditto; ratio is too small for it to be worth maintaining and the headache of merging mods.

[*]Exporer: 2; should be 4 (like chariots - though not worth fixing unless ^1.5 code is added)

Explorers cannot attack.

Most of the rest are in the 'tiny difference' area.

[*]Machine Gun: 14; should be 16 (immunity to collateral mostly makes up for the inability to attack)

Inability to attack means you cannot attack to take out the collateral attacking units.

And this is a tiny difference.

In essence, the largest differences you proposed are on the order of 17% for one culture's unique unit (that happens to be good at defeating the best ancient era attack unit), and paratroopers.

Everything else is in the single-digit percent range.

 

[MartinHarper]

20%difference

I'm assuming a ^1.5 modifier being applied on top of that by the dll, so for Paratroopers:
25^1.5 = 125
30^1.5 = 164
30% difference.

For Hwacha: 5^1.5 = 11.2 ; 7^1.5 = 18.5 -> 65% difference.
For Mobile Artillery: 26^1.5 = 133 ; 40^1.5 = 253 -> 90% difference.

none of these changes are worth doing and maintaining an XML file delta... and the headache of merging mods.

CIV4UnitInfos.xml is already changed in Better BTS AI (from the unofficial patch), but obviously it's not my call as to whether it's worth changing further.

 

[Yakk]

Ya -- but that civ4unitinfos should be changed to a modular XML loading based 'patch' file that just changes the single unit in question. (yes, I did a diff)

As an aside, modular XML loading with standard XML files in the same mod 'just works' -- the changed units in the modular part are applied after the standard XML is loaded.

 

[TheMeInTeam]

Flat opinions only hold so much water.

For example, I could make a very powerful argument that 2 move units, used well, can account for WAY more than an 8% difference. Such units in human hands can probably be considered 10-20% stronger (despite no defensive bonuses) than the same unit with 1 move...maybe even more because of the fewer units faced.

Now, the AI doesn't use them so well, but nevertheless the mobility and 2x attack are enough to weight it above its str value.

A lot of the others are questionable too, but not in terms of whether the argument is bad, but moreso because we have a degree of subjectivity here no matter what, and you're suggestion changes of a relatively small magnitude.

I can't tell the difference of a unit's opportunity cost of less than 10% variance through casual observation, and I assert the OP can't either. Do we have an OBJECTIVE, CONSISTENT measure to justify these changes? Maybe if the AI can't handle a certain kind of unit very well we should weight it weaker than its true value when it comes to building them?

 

[MartinHarper]

Do we have an OBJECTIVE, CONSISTENT measure to justify these changes?

Well, I have a spreadsheet where I have plugged the unit attributes into a simple formula, and I used this post to highlight the units that are furthest away from the spreadsheet value. Where possible, I tried to be objective. For example, to value defence I assumed that:
1) Each combat involves an attacker and a defender, so attack and defence are equally valuable.
2) The average defence bonus is 25%.
3) 1.5 exponent between strength and value.

So, take a unit with strength 100 that gets defensive bonuses. When it is attacking we calculate its value as 100^1.5 = 1,000. When it is defending we calculate its value as 125^1.5 = 1,398. So its average value is 1199 - around a 20% increase in value. This maps to a 13% increase in iPower with the current DLL.

However, I wasn't able to value everything objectively, with mobility being the clearest example of something that is always going to be a subjective judgement. How much would you value going from move 1 to move 2 if all else stays the same?

Maybe if the AI can't handle a certain kind of unit very well we should weight it weaker than its true value when it comes to building them?

I believe this is a question for iAIWeight, not iPower.

 

[Roland Johansen]

The value of higher mobility is tactical and that makes it very hard to value. The highest value for the AI is probably that it allows it to attack with these units before the enemy has time to react. This is even more true for helicopters with their 4 moves. Slow moving infantry stacks can often be intercepted by defending stacks with collateral damage inflicting units.

But it's of course hard to put a value on that. I'd say some 15% power difference (so base 10%) for 2 move units and 30% (base 19%) for 4 move units. But that's very subjective.

I don't value blitz very highly as it's mainly useful against much weaker units. I don't see the AI using this well regularly.

It's pretty much impossible to value a unit individually without looking at its opponents. Knights for instance should be valued higher when pikemen didn't exist.

 

[PieceOfMind]

The question is, should the power for each unit depend on whether the player is AI or human? An army of mounted units might indeed be more useful in the hands of a human player but should the AIs determining of power ratings always be based on the assumption the player is another AI?

Note when I use the word "player" I include AIs, in case it wasn't already obvious .

 

[MartinHarper]

It's pretty much impossible to value a unit individually without looking at its opponents. Knights for instance should be valued higher when pikemen didn't exist.

Yes - I've tried to give each unit type a rough modifier based on how many counters it has and what promotions it has available, but this is subjective.

The question is, should the power for each unit depend on whether the player is AI or human?

Ideally, some average of the two.

 

[Roland Johansen]

The question is, should the power for each unit depend on whether the player is AI or human? An army of mounted units might indeed be more useful in the hands of a human player but should the AIs determining of power ratings always be based on the assumption the player is another AI?

Note when I use the word "player" I include AIs, in case it wasn't already obvious .

Ideally, some average of the two.

I don't know if you realise, but there exist quite a lot of players who are worse at combat than the AI. On this forum, they're usually called newbies and they're likely a lot more prevalent among those who don't frequent this website.

There are players who will attack with units piecemeal getting them all slaughtered against city defenders and then complain that 10 chariots could not even kill a single archer. There are players who lose a stack against the AI because the AI used siege units against their stack while they don't use the slow weak (in base strength) siege units. There are players who don't understand the concept of unit counters and thus lose with uniform stacks of units and lose their macemen against supposedly weaker crossbowmen.

Most of the civilisations on the map and most battles on the map are AI controlled (in most games that are played) and the increased or decreased value of a unit in human hands is extremely vague. So I suggest to just base the iPower on the AI's capabilities with the unit.

 

[Poppis]

Well, people who are worse at combat than the AI, probably aren't the core audience of BetterAI.

But anyway I do agree with this:

Most of the civilisations on the map and most battles on the map are AI controlled (in most games that are played) and the increased or decreased value of a unit in human hands is extremely vague.

 

[PieceOfMind]

BBAIAOAI
Better AI Against Other AIs.

I'm not cynical - I think I agree with RJ's suggestion that the focus here be on power considerations for units controlled by AIs.

 

[Roland Johansen]

BBAIAOAI
Better AI Against Other AIs.

Thanks for that.