The tonnage modules are such a bad deal for most races. I still haven't looked at the stats for all of the races, but the $350 price tag for 25 space or the $1000 for 50(?) is hard to swallow. It makes it worse when you use the %weight modules (power, armor, etc), because those modules grow in size. If you have 50% of your ship taken up by %weight modules, you'll only get effectively half of the tonnage module...



The Power modules drive me nuts too. The first 3 power modules all modify individual ship stats, but the 4-6 modules effect the fleet. You'd like to keep the 1-3 versions on your dreadnoughts and the 4-6 versions on smaller ships. The price tag of the level 6 module is also obscene.



Did you add the siege module change I proposed earlier in the thread? It makes their siege ships MUCH better because they choose the highest level module instead of the inferior ones.



I also noticed that some of your ship designs don't use the full special slots.



Know anything about Point Defense and how it interacts with Fighters & Bombers?

I don't use a generic template for all races. I too do look up what a specific hull benefits from and try to incoroporate that. It's just not so in-depth.

I think that for most races a kinda-generic approach is okay but the Harmony are really an exception where you made me thinking. For them CP-effectiveness does not matter, since they don't have upkeep.

It's a pity I'm so lazy. I really wanna be ready and start testing. ^^

My take on the UE v1.0.

• Corvettes are speciality ships. Power/Repair/Siege.
• Destroyers are pretty generic.
• Cruisers are weapon module ships.
• Battleships are carriers
• Dreadnoughts are carriers, heavier cruisers, and double armor weapon ships.
• Kept offence modules above 50% of weight.
  
  

ShipDesignTerran.zip

Vaulter Ships designs are just a copy of the UE ship designs. The ship traits are the same too. There was a note about changing them in the file, but they are still the same. A copy of the Terran ship design would also work if we are being lazy, although they won't get the 40% health bonus from Strong Alloys, so fewer armor modules may be justified.

My take on the Cravers.
ShipDesignSwarm.zip

Thanks for both the UE and the Cravers!

So package them all together and post them so we can try them out.

Well the Vaulters were still missing. But your post kinda set me under pressure so I quickly made their file aswell.



Unlike you I haven't bothered with the Invasion-ships at all, so I'm definately going to manually pick your two races to see if this makes a big difference in addition to the changes to the combat-designs.



The more people test, constructively criticize the designs or supply a greater variety of design-templates themselves, the better.



Enough-words, here's the file.



Installation-Instructions:

Unzip into (my documents)\Endless Space\Disharmony\Modding. This will create a new directory (my documents)\Endless Space\Disharmony\Modding\Small containing a file index.xml and several subdirectories.

In the main game menu, click the MODS button, find the mod "AISHIPS", select it and click load.

To keep the mod loaded: find "Endless Space" in the Steam Library, right click, select properties, and click the "set launch options" button. Add "+mod AISHIPS" to the game launch command line.
AISHIPS.ZIP

Well, your United Empire is bugged.







Reason is pretty simple:



You did this:



ModuleCategory="LongRange" Weight="0.325"/>



Correct would've been this:







Your Craver-File does not appear to have this bug.



I'm gonna fix it...

Here's the version with the fixed UE-ships.
AISHIPS.ZIP

Okay, I definitely need to make changes to the targetting. Having a 6 vs. 7 battle of my glass-cannons vs. the AI's glass cannons and them using focus-fire is kinda stupid.

Not sure if it'll work but my plan now for this is that:









































The idea is: If the opponent has more weapons than defenses use Spread-Fire, if he has more defenses than weapons use focus-fire and if it's roughly balanced use Guillotine.

Oops.



Thanks for catching that.



Invasion ships are pretty easy.

• 1 all siege design.
• 1 max elite troops + siege design
• 1 max bombers type 3 + siege design
  
  

Nothing else worth having on them. Bombs may be the exception, but I don't know enough about AI bomb use to trust them (even if they can use them).



re: targeting

You know where I stand on the issue.

I'd like to propose a challenge, if you would be interested. See how far you can make it just using spread fire. And which battles that not using it make a difference. (make the AI do it too).

Spread fire works very well against the AI because you'll rarely see them use any Dread-Only-Fleets. If my change works as it should, than it should use Spread-Fire almost exclusively in most circumstances but not all. That's what I gotta test. Slept till now but now will continue testing by repeating the same fight where the Sheredyn stupidly didn't use spread-fire with the changes that should cause them to indeed do it.



As for the invasion-ships: No doubt that that design is good. I'm more concerned that the AI will build too few invasion ships, if there's only 3 designs to choose from. That's what I gotta test. If it works, then I don't see why I wouldn't copy&paste your changes over to all AIs.

Too bad. My change doesn't work as expected.

Apparently it doesn't check the tonnage/mp but just the existance of weapons/defenses. It played Guillotine because I had something of everything. So it killed 4 ships instead of 6 as it should.

Guillotine would've been okay with beams/Kinetics in this case. But they had missiles.

I will ad a fat positive modifier for missiles on spread fire and a negative one on the others. For missiles you really will almost never want to use something else because the lack of mid-round-retargetting.



Edit: something is weird. I'm pretty sure they should be using Spread-Fire by now but it's always that 2 more ships survive than they should.



I will now force spread-fire to see if it doesn't have to do with some kind of dodging-mechanic (2 of their ships seem to have missile level 2, but apparently less missiles, so probably those are missing). If that turns out to be the case, then that sheds a light on missiles that make them extra-bad in comparison to other stuff!



Edit 2: Assumption turned out to be correct:

They actually used spread-fire all the time but apparently 2 of the ships just miss due to the dodge-mechanics. So it's more of a balancing-issue than an AI-issue.

I think you've been right about the tonnage module all the time. For a medium design it increases production-cost from ~400 to ~700. 25% more CP-efficiency certainly doesn't justify a 75% higher price.

It maybe becomes worthwhile at weapon-tier 3 and thus a general increase in cost or with a lot of Hexaferrum available due to the price-reduction... oh, wow, I already have 3 of them and it's still so expensive!

Using a generic template for all the races would be fine. But you know how I can't resist over complicating things.

Ail wrote:


Edit 2: Assumption turned out to be correct:

They actually used spread-fire all the time but apparently 2 of the ships just miss due to the dodge-mechanics. So it's more of a balancing-issue than an AI-issue.

Did they miss or did they get intercepted by the flakk modules?

I am not sure what actually happened game-mechanics-wise.



All I can say is:



It was 6 AI-Ships who all had LR-Missiles as their loadout. 2 of them obviously used the Tier 2 missiles as they roughly had twice the MP in missiles.

I had 7 ships with 1 anti-missile-module each.

All AI ships did only 1 shot at the beginning of the long-range-phase.



My expectancy:

I kill all 6 ships of them and they kill 6/7 of my ships.



Result:

They killed 4 of my 7 ships.



My first thought was: They used Guillotine instead of Spread-Fire.

But I ruled that one out by making sure they use Spread-fire. Also in that case the expectancy would be 3/7 ships killed, not 4.



So second thought was: The Tier-2 missiles must've missed, probably because they are fewer in number and due to the weird dodging-formula.



But as I said: I don't know enough about the mechanics to be sure what happened here.

Were your ships that survived damaged at all?

Any idea what range type the missile modules were?



So, riddle me this.

If the formula for Flak Interception is:

#Flak Interception >= Missile Evasion * (1+ Turn before Reach) * Random]0;1

Then, you might calculate the expected defense success against a set of missiles.



Assuming level 1 flak modules we have InterceptionAccuracy="10".

Assuming level 1 missile modules we have InterceptionEvasion="28" (LRM), 21 (MRM), or 14 (SRM).

(Note: because all missiles were reduced to a salvo rate of 1, this means that SRMs are even worse than we had previously thought. They are twice as likely to die to flak modules as LRMs of the same rank).



Now, I'm not sure how many attempts each flak module gets, I'm guessing 4? Given our formula, I'm further guessing the value in the brackets would then be 4, 3, 2, & 1.

Plugging these into our formula:



10 = 28 * ([4,3,2,1])*Random[0,1]

10/[4,3,2,1] = 28 * Random[0,1]

[2.53.3510] = 28*Random[0,1]

[2.53.3510]/28 = Random[0,1]

[0.089285710.119047620.178571430.35714286] = Random[0,1]



So, how many numbers between 0 & 1 are less than or equal to 0.0892857? 0.0892857.

So to get the average number of missiles blocked in this situation, we just add up the numbers, which gives us 0.744 missiles intercepted per flak module. So a fleet of 7 ships with 7 level 1 flak modules, should, on average intercept 5.208 missiles (assuming 4 rounds of fire). However, if you are lucky, it could destroy 28 missiles.

thuvian wrote:
Were your ships that survived damaged at all?

Oh, that might've been a super-important information for that matter.

No they weren't. All 3 survivors on my side where completely unharmed.



The range was long-range for the enemy-missiles. Pretty much to be expected since I made all small AI-templates wield long-range.

Some notes from my current tests.

• Terran Templates aren't working very well, not enough weapons and too much wasted space.
• 5% hullmass on destroyers is too small to effectively place anything, leaving lots of empty space.
• If the AI used spread fire reliably all the time, I wouldn't stand a chance. Changing the weights again to see if that helps.
• Possibly the others could be used, but not until late game.
• What do you think about having mono- or dual- defences only on destroyers?
• Any ideas about how/when the AI uses the -1 weights? Many of the designs that use fighters/bombers aren't getting the fighters and bombers, possibly because the weight is used up earlier. 30 tons for a fighter is really rough to budget on a destroyer.
• Checking the AI templates, I see 5+ Scout Design variants and a large portion of fleet designs are made of scouts. I wonder if we need to do so something about them. Incidentally, the Scout AIFleet weight is 0.0 compared to 0.3 for most (all?) others. Even though the devs said it was deprecated, I wonder if something from that is sneaking through.
• Siege concentrations are fine with just 2 ship types.
• The Harmony AI and I are throwing 6 CP worth of ships for the last 50 turns (started at turn 30). I'm barely making headway, but only because he doesn't use Spread Fire. The 200/200/200 defence 0/0/300 offence cruiser is a brutal design to fight against. Thank goodness it is a bit expensive for the AI. The Craver AI is walking all over the UE, which suggests that the UE designs aren't working, the Cravers are better, or the AI plays with the strengths of the Cravers more easily. The UE is wallowing in fleet debt due to having tons of ships on planets and not engaging in fights. Didn't they remove the AI's requirement to pay fleet upkeep?
• Missiles are having a hard time hitting opponents. Against the 200^3 monster, you need 12 LRMs to hit with any, and only in the Long Range Phase. It is essentially immune to LRMs in the other phases.
• Higher level bombers/fighters use multiple special slots whereas early ones use only one. That may cause problems with the hard weighting creating illegal ships.
  
  

I think you guys are doing a really good job with this. I have read the whole thread but by the time that I had done this you had already done all the races.



DarkMeph

I'm trying to figure out how ship templates are used. I've designed some ships that should provide evidence for these mechanisms.

• 
  
  

Ends up with a 110/110 (perfect fill) ship. with 1 Long Range Missile Rank 2 (14), 1 Long Range Missile Rank 1 (12), 1 Flak Module Rank 2 (6), 1 Power (22), 1 Engine (6), 1 Scout (30), and 1 Bomber (20). This is impossible for players to make in Disharmony, but legalish. We only get the first type of Special Module and only 1, which bodes well for more complicated designs at higher tech levels that use different numbers of special module slots per module. I have no idea how the weight is proportioned out. We should have equal numbers of each, instead we have 3 LRMs and 1 Flak. It is possible it added 1 LRM, 1 Flak, and then filled in the rest with something that would fit. However, in many other cases (undocumented), ships are left with large amounts of empty space.

• 
  
  

Creates a 110/110 weight ship. With 2 Kinetic Long Range Rank 2 (10*2=24), 1 Deflector Rank 2 (10=10), 2 Long Range Beams (12*2=24), 2 Shields Rank 2 (8*2=16), 2 Long Range Missile Rank 2 (14*1=14), 3 Flak Rank 2 (6*3=18), and 1 Kinetic Long Range Rank 1 (8*1=8). Again, this ship is impossible for a player to make due to the presence of 2 different ranks of a single weapon type, but legalish.

We should be seeing 100/6 = 18 tons per module type. Curiously, the weapons are taking more weight and the defenses less weight, even though the defenses were higher in the ship template. Perhaps weapons are placed on the ship first?

• 
  
  

Virtually the same design as the previous ship, I just reversed the ordering of the Weapon and Defense modules. I'm abbreviating module names now to make this easier.

We end up with a 108/110 (pretty close) weight ship. It has 2 LRK2 (20), 2 D2 (20), 1 LRB2 (12), 3 S2 (24), 1 LRM2 (14), and 3 F2 (18).

A legal design. I would have expected there to be 2 S2 (16) which is only 2 short of the 18 alloted weight. Instead it went with 3 (24 weight), which is 6 more. Perhaps the Shield was placed first? Or when backfilling the ship after the initial allocation, shield defenses had had priority. Suggests, perhaps, that items at the bottom of the list have priority over items at the top?



Some conclusions

• In some cases I was having problems where allocating sum weights greater than 1 was giving me empty space on the ship. Here we have evidence that this is not always the case.
• Special Module Slots are allocated while adhering to ship limits.
• Hard weighted modules (-x) have priority. But sometimes they don't appear at all, even when the modules are available. Perhaps a typo in the template?
• hard weighted modules are filled from top to bottom.
  
  

I think that the changes I made should infact realize just that: Use spread-fire almost always unless you encounter really well defended ships.

I put your changes in. I'll see how it works.

*Update* I tried them. They didn't spread fire against glass cannons reliably.



What do you think about having mono- or dual- defences only on destroyers?

Well, not very much. Omni-weapon-designs would kill them easily. I really don't think of destroyers as ships that should wield much defenses at all. Reason being their bad hull-weakness makes it very inefficient.



True. However, just a little bit of defense helps the Kinetics to last that extra round which is important for kinetics. Spreading out to 3 defenses covers all your bases, but I'm not sure its enough to make a defense. Going with the dual-defenses may be enough to help 2/3 times. Going with mono-defense would be safer, but only 1/3.



I think that's because you violated the scout-design suggestion. The AI maybe identifies the scouts by their default-loadout which is Engine+Sensor+Nothing.

Your scouts were stuffed with weapons in addition to that so that maybe changed it. At least I cannot remember scout-spam from other races.

The ships from ShipDesigns.xml also show up. The events seem to use those templates too.



Didn't they remove the AI's requirement to pay fleet upkeep?

No, fleet-upkeep has been the bane of the AI throughout vanilla and only got better due to the improvements to FIDS/pop in Disharmony.



That's unfortunate. However, out of the immediate scope of the current issue.



Missiles are having a hard time hitting opponents. Against the 200^3 monster, you need 12 LRMs to hit with any, and only in the Long Range Phase. It is essentially immune to LRMs in the other phases.

That's what I found aswell. Not only can missiles not swap targets, no they also have a high chance to not even hit.

I think it is more that Flak is stronger than it would seem. I've some thoughts, but need to test them.

After quite some time I revisted ES while showing it to a friend.

I used the Ship-Designs-Mod for this and saw that in the early-game the altered designs worked pretty well.



The medium-sized hulls of the AI, however weren't as good as I thought. He designed his medium-sized hulls similar to the small ones with almost no defenses and mostly weapons and could destroy most AI-ships in Long-Range-Phase.



This apparently means that for medium-sized hulls a smiliar approach to the small-sized ships is very applicable.



One thing that worked well against that was Fire-and-Forget-Carriers that were in company of at least one tanky ship. The Fighters and Bombers dealt massive amounts of damage while their ships were long gone.



He eventually lost. An interesting sight was, that the AI-invasions always used invasion-troops to insta-invade, since this mod used those on their designs.



I now think the best approach would be:



All small and medium-Ships should be cheap glass-cannon-like long-ranged-weapon-wielders, which also utilize bombers when researched.

Large ships should have a lot of defenses and fighters but very little weapons. This would probably be the best approach for the AI that tends to mix its ships anyways. Some ships with more standing-power allow the bombers of the other ships to unfold their full potential.



Invasion-Ships should only be the smallest design and always contain invasion-troop-pods to speed up invasion.



I would also say that the AI probably never should use the +25/+50-tonnage modules. The increase of ship-efficiency simply is not worth the increase in ship-cost, since most ships become like 70% more expensive.



The AI is actually pretty good with swarming tactics and having many cheap ships would help them more than a few very strong ones.

The lack of A.I troop invasions is a big flaw in vanilla.

Well, it is the way it was in vanilla and back then I never considered it to be a flaw.



I actually liked the simplicity of fleet-management back then.

Sorry I ment no invasions in vanilla disharmony.

It's not exactly "no invasions". It is just so rare because there's like only 1 out of a total of 12 or so invasion-ship-templates use them. The AI is capable of performing the invasion but can't because their ships are can not.

thuvian wrote:
Retreat Rules

AI Fleet Design

• I tried out a experimental game with the start of my modified Harmony Fleet Designs. I never had a chance.
  
  

Not really sure if you mean what I think you mean.

So you modified the AI-Harmony-Ship-templates and then played against them and got owned by your creation?

If that's the case, then that's exactly the thing that I had in mind to happen with this. Improving one of the AI's weaknesses to make the game harder.

What was the difficulty-level you used for it?



Could you please post an example of your Templates so I can compare them with my changes to see if I go in the right direction?

Found out the first one myself:



Works like that:

Hahaha, yea if you don't know, I definitely don't know

I've done it for 2 races but it gets really tedious. I also don't want to simply copy&paste since parts of the designs also consider special-race-boni.

Is someone else in the mood of doing this for the greater good. I'm pretty sure this could be one of the things we could convince Meedoc to include in a patch, if it makes the AI better.



In my opinion it's one of the greatest weaknesses of the AI right now: They lose all those battles they should have won or drawn because their ship-designs are so terrible.



If not, I'll make some work-schedule for that and do 1-2 races each day from now on. ^^



Update: 4/10 done

Can you post up an example of what you've done? I can take a swing at the Harmony if you haven't filled it out yet.

Sure, I'll make the fifth race now and then post some of it...



I've filled out up to Hissho now. Harmony being before Hissho in alphabetic order.



Here's an example:



Before:

















After:



















Often you'll find Repair-Mods on Small ships. I always replace them with Power-Mods.



In case you don't know yet how those templates work:



Negative numbers behind a module mean: Put exactly this number on the ship.



Example:





This means: Always put 2 Fighters on (of course only if you have researched them)



Positive numbers mean: Distribute the remaining space like: this number/all positive numbers



Example:











This means from the remaining space take:

0.15/(0.15+0.25+0.35+0.4)=13% for least usefull defense-module

0.25/(0.15+0.25+0.35+0.4)=22% for medium usefull defense-module

0.35/(0.15+0.25+0.35+0.4)=30% for most usefull defense-module

0.4/(0.15+0.25+0.35+0.4)=35% for most usefull weapon-module



Of course it won't always fit perfectly but it'll try to get as close as possible.



For weapons you can add this:

Weight="0.8"/>



to force long-ranged. Which is what I did for all small ships but not for Medium and Large ones since they usually are not glass-cannons and thus can take quite a beating.

• Did you figure out how exactly "FleetWeight" works?
• Does the AI refit using money?
• How does the AI calculate defense and offense usefulness?
• Do the AIs always use the base Races?
• What are you using as the AI targetting plan? What can I do to convince you (and Meedoc) to make it spreadfire 100%?
• How does ShipDesign.xml fit into things? Is it just another list of ships? I noticed the AI was building them.
  
  

I think that ModuleSupportInvasion strength is being miscalculated by the AI. More precisely, from SupportModule.xml:











I was looking at the ships for a late game Endless Craver AI. I'm no longer impressed with my ability to match him him militarily.

FleetWeight is not used at the moment. It's a deprecated property...

Yep, the AI refit with money as the player. But the difficulty grant some bonus on the refit cost.

We compare the effectiveness of the weapon/defense during the several last battle. We try to find some weak point in the enemy defense and try to counter his weapon with appropriate defense. So if you use a lot a missile, the ai will try to use a lot of anti missile.

We compute the long/medium/short range stuff by the following rule (define in the registry.xml around line 434):

-> Long is good against short, average against long, weak against medium. (best weapon = missile, average = laser, weak = kinetic)

-> Medium is good against long, average against medium, weak against short. (best weapon = laser, average = missile, weak = kinetic)

-> short is good against medium, average against short, weak against long. (best weapon = kinetic, average = laser, weak = missile)



I will let Meedoc answer about targetting plan ^^.

Thanks for the information!

Two more if I may?

• Is there a way to dictate ship construction type composition? The original thought was FleetWeight, but since that is deprecated, what took its place?
• Currently, defense fleets get filled up with siege ships. Is there a way to modify which ships get put into which types of fleets (Commando/Defense/Etc)?
  
  

Nope, currently, we use a "Random" to choose the ship. And I agree that we should use something else...

Odd... normally, siege ships are not used by the "military" part... I define the ship utility by comparing the Military power to the invasion power. If the invasion power is greater than 0.5f (not sure if it 0.5 or 0.8) then I don't use it as military...

Did you have a save with can describe this odd behavior?

Perhaps the siege fleet was just "gathering" on the system when you attack it... so it was not a "defense" but an "attack" which has been intercepted...

"

Did you figure out how exactly "FleetWeight" works?"

Would it probably help to make copies of the designs of which I want the AI to use more, so that the random picking has a higher chance of picking one of them?



Say I have 6 Small designs but 10 Medium and 15 Large, would that probably lead to the AI building more large designs?



Much better would be if FleetWeight would just work.



"

What are you using as the AI targetting plan? What can I do to convince you (and Meedoc) to make it spreadfire 100%?"

Well, I disagree with Spreadfire 100%...

It should really depend on what you are up against.

One can't make it overly-complicated but some often-applicable rules of thumb might suffice.

Maybe something like this:

if(Enemy fleet has small ships in it)

use Spread-Fire

else if(my fleet has more ships than enemy)

use Guillotine

else

use Nosebreaker



I even think there already are rules that are something like this. Didn't look at it very deeply since I consider the bad designs more of an issue than wrong-targetting-option.



"I was looking at the ships for a late game Endless Craver AI. I'm no longer impressed with my ability to match him him militarily."

In Vanilla the AI for a long time had massive troubles to work out their economy. Very often they kinda killed themselves by overtaxing their people into rebellion and then having to sell improvements. This had been fixed over the time.

Now in Disharmony, where emphasizis has shifted towards more important decision-making during and before combat, it turns out that the weakest spot of the AI now lies here.

That's why I think that creating competative AI-Ship-Design-Templates is really important to keep good players at it.



For me the last few games (Serios/Impossible) all started like this: The AI looks really fearsome. Fast expansion, big fleets with high MP... But when you engage those fleets in combat you can wipe them out with much less MP simply because they are so bad!



Even the percentage-based bonusses on the higher-difficulty-levels don't cut it if the underlying base is so bad.



Making AI-Ship-design-templates is no rocket-science. So every modder should easily grasp how it's done. I'd think it would be great if different people would make different races with different approaches that are known to be better than what the AI does. This would even add some flavour to races when you know that it's different stuff they like most.



At the end we can throw them all together, have Meedoc/ThorTillas/ourselves review them for potential mistakes and get them into a patch... Now that sounds like what Games2Gether is all about.

ThorTillas wrote:
Nope, currently, we use a "Random" to choose the ship. And I agree that we should use something else...

Odd... normally, siege ships are not used by the "military" part... I define the ship utility by comparing the Military power to the invasion power. If the invasion power is greater than 0.5f (not sure if it 0.5 or 0.8) then I don't use it as military...

Did you have a save with can describe this odd behavior?

Perhaps the siege fleet was just "gathering" on the system when you attack it... so it was not a "defense" but an "attack" which has been intercepted...

Sure.

Here are 2 pictures from my last game of the AI Craver separated by 4 turns (also attached are the saves Save.zip). You can see he has massive amounts of Siege in his defense fleets. It might be that he ships were indeed gathering at the system prior to be assigned to another fleet. I've not worked through the fleet AI manager, so I'm guessing here. In the save, you can also see that almost half of his design pre-dreadnaught are seige designs (that use siege module 3 instead of 4).





More speculation from my part: If he is building randomly, then he would have a high number of siege ships. Then if he sends his combat ships off to die, but keeps his siege ships at home, you'd gradually see the population of siege ships overwhelm the population of non-siege ships, because nothing is killing the siege ships. Then the "defense" fleets would look as they do, but not by design, just a emergent behavior.

Ail wrote:
"

• Did you figure out how exactly "FleetWeight" works?"
  
  

Would it probably help to make copies of the designs of which I want the AI to use more, so that the random picking has a higher chance of picking one of them?

That would work, although some sort of frequency weighting would be better. What we have now leads to the situation with siege ships I describe above.

Say I have 6 Small designs but 10 Medium and 15 Large, would that probably lead to the AI building more large designs?

Yes. But I think it should be the other way around (You know how I feel about ship design). I see the AI building 1 medium $600 ships and 1 small $100 ship. That means you are spending 6/7 of your economy on a big useless ship. I think a 6:1 small to medium would be a much better way to do it.

Much better would be if FleetWeight would just work.

Yes. It would be a start in a better direction.

"

What are you using as the AI targetting plan? What can I do to convince you (and Meedoc) to make it spreadfire 100%?"

Well, I disagree with Spreadfire 100%...

It should really depend on what you are up against.

One can't make it overly-complicated but some often-applicable rules of thumb might suffice.

Maybe something like this:

if(Enemy fleet has small ships in it)

use Spread-Fire

else if(my fleet has more ships than enemy)

use Guillotine

else

use Nosebreaker

Marcelo_0 wrote:

• Added variety in the AI attack targeting
• The AI select the best formation & targeting; not anymore randomly one from 3
• Lowered weight for Retreat and Offensive Retreat
• Updated weight on formation & targeting for the AI to choose
  
  

It was: 33% for each Spreadfire, Guillotine, or Nosebreaker. I now believe it does the following:



From AIParametersBattleCard_Xp1_locales.xml we have:

Making AI-Ship-design-templates is no rocket-science. So every modder should easily grasp how it's done. I'd think it would be great if different people would make different races with different approaches that are known to be better than what the AI does. This would even add some flavour to races when you know that it's different stuff they like most.

Well, I actually find ship optimization to rather challenging. Creating templates that are universal enough to work for all tech levels, economic developments, and enemy fleet designs is complicated. Engines Modules are a good example. Who needs engines? Power Modules is another example, under what circumstances do you use power modules?

At the end we can throw them all together, have Meedoc/ThorTillas/ourselves review them for potential mistakes and get them into a patch... Now that sounds like what Games2Gether is all about.

That's why I'm trying to help.

thuvian wrote:
Well, I actually find ship optimization to rather challenging. Creating templates that are universal enough to work for all tech levels, economic developments, and enemy fleet designs is complicated. Engines Modules are a good example. Who needs engines? Power Modules is another example, under what circumstances do you use power modules?

They don't have to be all perfect.

I think even simple things can improve it a lot.



Like:

Small ships only Long-Ranged

no Repair-Mod on small ships



As for Power-Modules: The high-end ones get good on small ships, since they have a stacking boost for all ships. The low-end ones, however, are pretty much useless.



As for engines: I only put them in the templates for the ships that have this -75% on engine-tonnage.



Gonna make another race now. In a few days I shall be done. Hope trying it out in the end will be fun.



@Meedoc/ThorTillas:

I also am thinking about how to teach the AI the Retreat after long-range-trick.



Would something like this work:



















The problem is that I would need several of this things work together to trigger the retreat. Let's say 3 of these conditions fit, so I already get a value of 1.2 and then the fleet would retreat for no reason.

Retreat Rules

• I think you are missing the negative weights to the things that would make the fleet stay (e.g., Short Range should be -0.4).
  
  

AI Fleet Design

• I tried out a experimental game with the start of my modified Harmony Fleet Designs. I never had a chance.
  
  

Some Suggested Guidelines for Retreats

• Long Range
• • BASIC - You are outgunned.
  • BASIC - You ships are low on health.
  • MODERATE - Enemy Fleet has few Long Range weapons relative to Short & Medium Range Weapons
  • MODERATE - You do not have many Long Range weapons and the enemy has many Medium and Short Range Weapons.
  • ADVANCED - You can survive running away.
• Medium Range
• • BASIC - You are outgunned.
  • BASIC - You ships are low on health.
  • MODERATE - Enemy Fleet has few Medium Range weapons relative to Short Range Weapons
  • MODERATE - You do not have many Medium Range weapons and the enemy has many Short Range Weapons.
  • ADVANCED - You can survive running away.
  • ADVANCED - You have many Long Range weapons, the enemy is still strong, he does not have many Medium Range weapons, and he has many Short Range Weapons.
• Short Range
• • BASIC - You probably shouldn't.
  
  
  

It might look strange that I ask these things since you'd probably think that I'm one of those who knows this stuff already but I have a few questions about the ShipDesignRacename.xml's:



How exactly does "FleetWeight" work? I once fiddled around with it but it seemed to change absolutely nothing.

My expectation would be that FleetWeight determines how often a specific design is made as in:

Build Probability=FleetWeight/Sum of All FleetWeights



Apparently with a line like this:



you can force a specific module-type. But is it also possible to force a specific weapon-range?



My two goals are:



Make the AI use small ships as LR-Glass-cannons since defenses work poorly on small designs with their massive Hull-weakness anyways.

Make the AI use more big ships in comparison to small ones when bigger ships are available since their much greater durability renders the small ones kinda obsolete.

Sorry for being ambiguous. Yes. The AI performed very well.



Some observations

• The AI SuperDefender class ships start around 600 Industry. That's a massive speed bump if it builds any.
• ShipDesign.xml needs to be update in addition to the individual AI profiles.
• The Ships in ShipDesign.xml are hardcoded and leave much of the ship empty.
• The AI ship designs feature engines, power modules, scout modules, repair, and tonnage modules. Very rarely are they good choices.
• Just fixing these things lets the Harmony AI out expand the other AIs.
• The siege module weights need to be fixed as I mentioned previously in the thread.
  
  

My thoughts on build guidelines

• No Engines, Power, Scout, Repair, or tonnage modules on small or medium ships.
• Maybe use Engines, Power, Scout, Repair, or tonnage modules on large ships, but I doubt it.
• I don't have good ideas about large ship design, perhaps 1 power: 1 repair: 1 armor: 25 defenses: 50 weapons?
• Put enough weapons on a large ship to kill a medium/light ship in each combat round per range phase.
• Special Slots are a trap. I'm not sure they are useful for combat ships.
• Fighters/Bombers should belong on small ships, so you can maximize percentage bonuses on large ships.
• Except for the Harmony2 hull which gets -75% fighter/bomber weight, which is an okay deal.
• Combat ship Fighters/Bombers should always be battle designs.
• Three designs of siege ships, 1 with elite troops, 1 with siege bombers, and 1 with only siege modules.
• Limit the number of possible siege designs, otherwise they will end up dominating fleet composition. This results in game states where there are tons of siege ships, and none with guns. Potentially just design hybrid ships for the AI.
• With an equal chance of building any ship and equal numbers of variations of each ship class, the AI will end up dominating its economic fleet composition with large ships. Example: If you have 1 of each class of small, medium and large, you'll end up with a fleet consisting of one of each. However, that small ships cost 200, the medium 500, and the large 3000. So the majority of your AIs economy is nested in the big ships. This means that you either need to work on weighting the ships better, or designing really good big ships.
• Hard coding kinetic long range modules and laser medium range modules, sounds like a good idea, but then AI can't switch them to something else if the player builds his ships to account for this. I suppose we could design the three different versions of glass cannons and then the ai will build equal amounts of each. That might work.
• I need to setup a system so I can watch the AI develop after changing templates.
• Not sure what to do with Scouts (which currently have minimal weapons) or Scouts & Colony ships (which have engines).
  
  

[code]








































































































































ShipDesignName="%ShipClassHarmonySmall2Title">

With just the few changes I made, the AI is much harder. After my attempts at Endless met with disaster, I've been trying on the second hardest difficulty.

If anyone wants to try it out, here are the updates I made for the Harmony. It isn't in a mod form, you'll need to drop them into your Disharmony install for it to work. Backup your files before you do it. This is a work in progress. On the bright side, you can also load a base Harmony game and if you have the Harmony AI it will change to reflect these new build plans.



Simulation/ShipDesign.xml - The default ships

Simulation/SupportModule.xml - Attempt to fix the weighting of the various siege modules

AI/ShipDesignHarmony.xml - The Harmony upgrade ship plans

AI/parameters/AIParametersBattleCard_Xp1_Locales.xml - Changing targeting rules to heavily favor spreadfire.



The ship designs aren't the perfect yet. I'm still working on refining the math behind which ship designs and fleet composition weighting. However, it still puts up much more of a fight and I think you'll be pleased at the difference.



To uninstall write over the files with the ones you backed up, or just validate and repair using Steam.



HarmonyShipRedesignv1.zip

Had a look at your changes and there's some points that I found questionable.



In Battle-Cards you kinda forced to use spread-fire all the time.

I can imagine that it works very well against you because you said you only use small ships all the time and consider the big ones a waste of ressources. Against a 4-Dreadnough-fleet with high defenses and repair-modules this would be a totally awful choice!

It should use a sophisticated logic that probably is not doable with XML.



I'd say:

DefaultFleetTargeting should get these modifieres:

myWeaponMissile 1 (since you can't reatarget)

BiggerFleet 1 (if I'm bigger I'll probably eliminate the opposing fleet anyways)



FocusFireFleetTargeting should get:

myWeaponKinetics 1 (I can retarget often enough to switch target after each destroyed vessel)

SmallerFleet 1 (if I'm smaller I want to try to kill at least one of the enemies)



StrongFireFleetTargeting I would leave it blank so it can randomly be taken when none of the other conditions meet (which will rarely happen)



In ShipDesignHarmony.xml you took out the Tonnage-Modules of all ships. This might be okay for small glass-cannons in order to make them cheaper. But on the medium and big-ships I would try to get as much stuff on as possible and definately not leave out tonnage-modules.