An extensive combat analysis of ship hulls

[slv]

Hi there, another longpost from me

This time I will try to give an overview of combat capabilities of different hulls, finding out what's good what's bad currently. Also if the result isn't close to hat you expect maybe we need to nerf/buff smth?

TLDR: Shields suck until lategame (namely until Titanic/Solar/Scattered Asteroid), Most hulls are balanced, Quantum Energy OP.

Preliminaries
Before we start I need to specify how exactly will we compare combat strength of ships. One of a good ways would be to run simulations, even though it's quite tedious. Turns out that this isn't very needed.
First of all in most cases ship hp would be substantially greater than a single weapon damage, as a result (due to the law of large numbers) if our fleet is large enough at the end of every combat round your ships will be damaged in a similar fashion (Possibly you have noticed that your ships often end the battle with similar hp). In that case in fleet battles consisting of only one ship design the most important thing is the product of Damage*Structure. If you have two fleets and total damage*structure of them are equal they are likely to either both survive the battle or both be destroyed. If your damage*structure is greater than in some cases (if damage/structure ratio for both of you is too high both fleets will live) it's possible for your fleet to survive and "win" the battle.

Of course the production cost of the ship also contributes a lot to its combat efficiency, if your ship is twice cheaper than opponent's you can build two more ships, so your damage*structure will be four times greater.

This makes it easy to have a rough comparison of ship designs, it's the Damage*Structure/(Production cost)^2 of design what usually matter. The reason I say "usually" is that if ships are invisible combat math is completely different, also this doesn't take into account fleet upkeep (it favours larger ships). And also sometimes if your Damage is too low you can't destroy enemy fleet even if your product of Damage*Structure is over the roof. Shields also mess up the math a lot. Despite this we will analyze shields and "Raiders" (invisible ships) in this post too.

Currently ingame designer tries to take into account the fleet upkeep by displaying you (Damage*Structure)^(0.6)/(Production cost) figure in design menu. It makes some sense, the main problem with this is that 0.6 is a self-made parameter and until very late in the game fleet upkeep doesn't change math that much.

Shields will be omitted from comparison until the very ending of the post.

Early/Midgame/Lategame separation is made based on ship parts, laser+zortrium for early, plasma+diamond for midgame. Hulls are chosen so that they have reasonable RP cost comparing to parts. Some of hulls are harder to get eve if they are compared in the same category.

1.Early game
It is not that easy to tell what exactly "early game" is. In this case we will assume you have access to Laser and Zortrium armor. A neat fact is that it doesn't matter what level of laser do you have, all laser designs are improved by the same amount when you research the next level, so I will use lvl1 number from now on.

{I left an optimal configuration, if one changes the slot distribution the rating will drop, sometimes only by a little bit}.

Static Multicellular hull
Laser, double Zortrium, speed 100
5*38/ (70^2) = 0.0387755102

Robotic Hull speed 80
double Laser, double Zortrium, speed 75
10*47/(112^2)= 0.03746811224

Asteroid Hull speed 60
Laser, triple Zortrium
5*63 / (68^2) = 0.06812283737

Asteroid Hull + drive (speed 80)
5*63 / (88^2) = 0.04067665289

Organic Hull speed 90
Laser, double Zortrium
Initial 5*27 / (56^2) = 0.04304846938
Final[25 turns] 5*32 / (56^2) = 0.05102040816

Symbiotic Hull speed 10
Laser, Zortrium
Initial 5*21/(48^2) = 0.04557291666
25 turns 5*26/(48^2) = 0.05642361111
Final (50 turns = never, don't take this into account ) 5*31/(48^2) = 0.06727430555

Asteroid is the best and for a good reason since it is deeper in the tech tree. Note that if speed is a concern than organic ships are better than asteroid, drives are really expensive on small ships.

Static Multicellular is less efficient than organic and has no growth/repair. 100 speed vs 90 is a substantial increase, but I am not sure if Static Multicellular is a good hull. Was amused by the strength of symbiotics, wasn't using them before. Played a couple of games using symbiotics as a main vessel, Laser+Standard armour plating symbiotic has a combat rating of 0.0378, does not require zortrium research and develops nicely into Protoplasmic/Bioadapter/Endosymbiotic stealth armada.


I don't like how Static Multicellular is worse than a straight organic. One may say that 100 speed vs 90 speed matters (and I agree),but turns out that it's worse than symbiotic which also has better stealth and regeneration. As we will later see that's a problem with whole "dead" organic line. "Alive" organic ships have mediocre stats too, but the quirk of stealth makes using them a reasonable approach, zombie hulls does not offer that.


This section needs to be taken with a grain of salt since Damage*structure mnemonics doesn't work particularly well since structure/Weapon damage ratio haven't grown enough yet. Having 34 structure doesn't differ from having 42 structure if you are facing Laser IV, but formulae will show a substantial increase of efficiency of 42-structure ship. On the other hand if your enemies vary and you face Mass Driver IV enemy then Laser II enemy and then Laser IV enemy the damage*structure approach works well and the range of enemy weapons smoothes things.


Midgame

At this point I will assume you have access to Plasma and Diamond Armor

Static Multicellular speed 100
Plasma+2Diamond
9*52/(74^2) =0.08546384222

Endomorphic speed 100
Plasma+3Diamond
Initial 9*59/(87^2) = 0.07015457788
Final[30] 9*74/(87^2) = 0.08799048751

Energy Frigate
It's a joke
1 Plasma 3 Diamond
9*69/(104^2) = 0.05741494082

Heavy Asteroid Hull
2 Plasma 3Rock, 3 Engines, speed 90
18*122/(174^2) = 0.07253269916
2 engines instead (speed 80)
18*122/(164^2) = 0.08164782867

Ravenous Hull Speed 100
1 Plasma, 4 Diamond
Initial 9*77/(97^2) = 0.07365288553
20-turn 9*87/(97^2) = 0.08321819534
Final[40] 9*97/(97^2) = 0.09278350515

Quantum Speed 120
2 Plasma 5 Diamond
18*140/(150^2) = 0.112

Fractal speed 120 (It should be obvious that a result is strictly worse than quantum, 2 quantum ships have exactly the same PP cost and slot number as fractal, but have 60 more structure)
4 plasma 10 diamond
36*220/(300^2) = 0.88

Nano-robotic Hull speed 80
2Plasma 5 Diamond
18*120/(162^2) = 0.08230452674
+N-dimensional engine, speed 100
18*120/(182^2) = 0.06520951575

Self-gravitating
3 Plasma 3 Diamond; Speed 80
27*154/(204^2) = 0.0999134948
+Engines, speed 100
27*154/(224^2) = 0.08286830357

Scattered Asteroid
5 Plasma 10 Diamond, 4 Engines, Speed 100
45*320/(440)^2 = 0.07438016528 (without taking shields into account)
Against enemy with plasma (15 damage)
=0.09297520661
With deflector shield (speed 90)
45*320/(480^2) = 0.0625 (without taking shields into account)
against plasma 4 opponent (18 damage, shields strength 8 )
0.1125

Bioadaptive (Please don't use this in combat , more on bioadaptive later)
Single Plasma Double Diamond
{I mean really, don't use this, play with stealth bioadapters instead}
Initial 9*51/77^2 = 0.0774160904
20turns 9*61/77^2 = 0.09259571597
Final [50] 9*76/77^2 = 0.11536515432

Sentient Speed 100
Double Plasma, 4 Diamond
Initial 18*84/(172^2) = 0.0511087074
Final [45] 18*129/(172^2) = 0.07848837209

Amusingly most hulls seem to have similar range of combat efficiency. Static Multicellular seems to be wrong here, it's an early game hull which have comparable stats even to late game ones. The real-game fleet upkeep thing mitigates that so it's not an issue I guess. Fact that Ravenous/Endomorphic/Sentient are weaker than Multicellular is not fine. At the moment I see no reason to build the Zombie-hulls. Bioadaptive is a completely different beast since its strength is a stealth-raider route.

This is something I am really concerned about right now. Speed advantage does not mitigate that since placing drives on self-gravitating will eliminate the speed difference while still maintaining the combat lead. In my experience ravenous/bioadaptives can't be obtained earlier than Self-Gravitating and if the later is also better combatwise something is wrong.

The reason for this is a common misconception (one I had until I did this calculations) that the most important thing for ship hull strength is the ratio of external slot/production cost. Turns out that initial structure plays high role until the very end of the game, even with diamond armor (which is quite deep into armor tree) most efficient ship hulls are those that start with high starting structure. AFAIR organic hull ships had high starting structure too (changed into growth), but appear to be overnerfed.

Scattered asteroid hull is a neat one since it's the only midgame hull to take advantage of shields (more on that later), that's because it's not a midgame hull but a lategame hull you get in midgame.
Another notable feature is that most "optimal" designs have low damage output, they can't kill themself in a fight. Whenever you are fighting a weaker enemy (as you usually do against AI) there are some merits in building worse design but with higher damage. I played a few games using these "optimal" designs and it turned out that they were still ok in killing AI fleets.

3. Lategame
I lategame people finally get the chance to play with 30-slot capitalships and it changes the combat math completely. To the point I am not able to analyze it.

The lategame ships is a wonderful land which differs substantially from what we saw before. Differences include
1)Efficiency of shield
As soon as your ships start to have huge PP cost shields become good. The reason it makes computations hard is that now great pilots bonus starts mattering. When ships are shieldless relative power doesn't depend on whether a player has or hasn't aquired Mu Ursh, with shields it depends.
2)Efficiency of engines
Since it's more efficient to put engines on large ships it becomes really cheap to make your ships quick. Te choice of your fleet speed starts to matter and efficiency will depend on different factors.
3)Significantly different research cost makes it possible to have different weapon arrangements, people with more easily researched hulls will have better weaponry.
4)Spinal Antimatter Cannon makes things complicated.
Previous things are manageable, this one is not.
First of all the cannon makes Damage*Structure/PP_cost^2 model not adequate (hostile ). If you mindlessly try to apply it you will get that Sentient Hull with just an Antimatter Cannon and Neutroniums is the best hull by far. In reality this ship Will lose to any fleet of small unarmoured ships. 100 damage is a lot, so assumptions turn out to be incorrect, 101-structure ships is twice stronger than 100-structure ship. Fleet of Transpatial Hulls with a cannon technically beats everything while losing to a typical early-game fleet.



Moreover, blackshield isn't efficient against the cannon. But in case if your opponent doesn't have blackshied then the cannon itself is less efficient than deathray. What we end up is the complicated rock-paper-scissor-lizard-spock scheme with much more typical cases. I really think that's intentional and I like such kind of balance.

Roughly you have following
QE hull beats everything without blackshield
Transpatial hull with cannon beats everything except QE hull or deathray-heavy ships. (also is slow as hell)
Sentient without Deathrays and blackshield beats all blackshield hulls and QE hull, loses to transpatial
Titan/Solar/Scattered beat different things depending on number of blackshields or Deathrays

If you ask what's the best hull assuming nobody has cannon?
Scattered/Solar : around 0.19
Titan/QE hull: around 0.15
Once again, hulls are pretty balanced here. Easier to get titanic has worse combat rating, rating of scattered is a bit inflated because of its shield bonus, optimally you want to have a fleet of solars and one scatter in it. QE hull has a really good rating, but at this point fleet upkeep rally starts to matter, we are talking about 2-3 Quantums against 1 other ratios, for reasonable number of ships rating drops to 0.14, so it's still worth it to continue researching better hulls even post quantum especially considering quantum can't carry antimatter cannon.


Below we assumed we had generic ships, does Mu Ursh change the outcome? It does. Turns out that as soon as you get Good Pilots (Etty/Hhohh) Antimatter cannon becomes less efficient than deathrays. For Etty/Hhohh cannon is still good with the largest hulls making a rockpaper-scissor picture still valid. But for Mu-Ursh/Eaxxaw (and even so for Misiorla) cannon becomes completely bad.

As a by product this makes shields significantly worse, skyrocketing the QE hull efficiency to 0.18 (for Eaxaw) and 0.2 for Misiorla.

[Ok, I am cheating here, in reality QE hull rating is the same but other drop by the same amount, what happens is that blackshields gives you 2x effective hp against generic species but only (45/30)=1.5 effective hp against Misiorla.

Also playing with Misiorlas makes scattered hull a bit better]

Of course the fleet upkeep still makes other hulls better, but not by much. As a conclusion, if you have Eaxaw/Mu Ursh don't worry if you are stuck with quantums or Fractals, they are still competitive even against very lategame ships.


4.Raiders (BioAdapters&Energy Corvettes)

Earlier I've told you that damage*structure/PP_cost^2 combat rating doesn't apply to stealth ships. Given that it's a valid tactic I will try to analyze them by other means.

Designs we are interested in are Compressed energy hull + Laser 4 (against early-game designs) and Bioadaptive Hull Plasma IV + Adsorption Field (against midgame designs).

There are few complication in raider analysis. First of all while there was a efficiency measure in doomstack vs doomstack battles it doesn't coincide with an efficiency against raiders, 1 plasma + 6 Diamond Quantum-energy ship is better against raiders despite being almost 50% worse against normal ships. We will assume that one player uses raider while the other uses "optimal" ships from previous sections.

Other complication is that it depends on which design a player uses, once again, military robotic hull is substantially stronger against raiders than it is against real ships.

We will compare raiders with (arbitrarily chosen) early game organic hull and robotic hull and midgame quantum hull designs.

Energy raider (laser IV) vs Organic Laser IV +2 Zortrium

Energy Raider has 11 damage and 10 hp, 35pp cost organic has 11 damage (if it was laser 3 raiders would have been twice more efficient) and 27-32 hp (happily the number of hp doesn't matter) 56 pp cost.

In a battle when 56 Raiders fights 35 organic ships odds are very much i raider's favour. In round 1&2 raiders deal a total of 112 shots which is enough to destroy all organic ships, and at least 21 raider will survive (some of organic ships will die in round 1, law of large numbers doesn't apply when organic ship die from 3 hits).

Energy raider (laser IV) vs Robotic 2x Laser IV +2x Zortrium

Energy Raider has 11 damage and 10 hp, 35pp cost robotic has 22 damage and 47 structure 112 pp cost.

Here you see a weakness of raiders, Robotic ship require 5 hits to be destroyed (it will have just 3 hp after 5 hits) and has just enough damage to destroy a raider.

112 raiders vs 35 robotic ships
After 2 rounds raiders would be able to destroy 35 ships, so at the end Raiders will win.

Conclusion: Energy raiders are reasonable, possibly too good, although fleet upkeep makes them worse. Please also note that they are obtained farther down the tech tree.

BioAdaptive + Adsorption field + 3 Plasma IV vs Quantum Energy 2 Plasma 5 Diamond

Raider costs 146PP with 54 attack and 15 structure (grows up to 40)
Quantum Energy ship costs 150 PP with 18*2=36 attack and 140 structure.

Bioadapers lose unless they have a numeric advantage (~50%) so that they kill QE ship in two rounds.

Conclusion is something which should be expected, Bioadapters are not efficient in doomstack vs doomstack fight. On the other hand they are not THAT inefficient,
if they manage to beat 25% of enemy fleet before it gathers they will win.

Shields
So here I will say a few words about something I was mostly ignoring before, shields.

In context of doomstack vs doomstackbattles the most important thing is Damage*Structure/Cost^2, so how do shields change that?

Shield increases cost and increases it a lot, doesn't change damage, but changes effective Structure. If you fight with a Plasma IV opponent and have a plasma shield your ship behaves in the same way as a ship with twice that may HP behaves.

So when it's worth it to place a shield? If shield makes your ship twice better then you need it to make the cost less than sqrt(2) larger.
If the shield makes ship p times better ship has to cost at least Cost_of_shield / (sqrt(p)-1) to be worthy.

For Defense grid against MD ship needs to have >70 PP cost.
For defense grid vs lasers a ship needs to have >173 PP cost
For Deflector against lasers a ship needs to cost more than >141pp
For Deflector against Plasma >523
For Plasma shield against Plasma >217
For Plasma shield against Death Rays >460
For Blackshield against Death Rays >362

And if you were patient you saw that most "optimal" ships have very low PP cost, i.e. Deathray+Neutronium Quantum Energy ship will cost 270. Plasma+Diamond will cost 150. I also have't even mentioned that shields technology cost RP and time to research. Numbers above just show that even if you have researched shield it's still not worth it to put it on your ships.

Ok, there are few exceptions.
First of all, if you are fighting AI with an out-of-date technologies than equiping your QE ships with blackshield against lasers is indeed efficient.
Second exception is Scattered Asteroid Hull and sometimes Acirema ships. The way shields stak sometimes makes it reasonable to put a shield on your ship (especially for scattered Hull). Usually by a tiny amount.

Third exception is indeed a reasoable one. At the moment you start building Titanic Hull/SOlar Hulls your ship will >500 of PP. For these huge hulls Blackshield and sometimes even Plasma is worth it.

6.Balance comments

So after I did few calculations I feel less happy about combat than I was before . Energy Frigate is still miles worse than anything. Organic heavy hulls turned out to be bad, it's not good if Static Multicellular (and Organic too) is comparable in power to late game hulls. Actually with Reinforcement Hull tech (it improves weak ships more) Organic is more efficient than Endomorphic/Ravenous!

I feel like both shields and organic hulls are overnerfed. Shields being bad until very late in the game is something I don't like. It does't feel good that even though I researched few of them I still choose not include them because of their liability.

As a suggestion I think increasing the growth speed of Ravenous hull to 1, Endomorphic hull to 0.75 is a good idea. These hulls will still be worse than Contra-gravitational (even after full growth!), but at least they will not be eclipsed by early-game robotic/static/organic/asteroid hulls.

The difference between organic hulls is so high that even adding a slot barely fixes them.

Fully grown Endomorphic with an extra slot will possess a rating of 0.092
Fully grown Ravenous with an extra slot has a rating of 0.094

It's not energy frigate levels, but still. At least when one gets energy frigate (which is awful) they are closer to quantum (which is the best ship in the game).

Speaking of quantums, how about making quantum/fractal more distinct? Maybe make quantums worse but give a bonus if it has a shield?

For shields balancing is complicated, since AI doesn't look into tech-advances and any chance to shields will make it overbuild them, need to think about it more. But some change is definitely necessary, it's really wrong that rushing Quantums and not-putting shields on it seem to be an optimal strategy for this calculations. It is still optimal even if you assume you get shields earlier than you supposed to, having Plasma Shields vs Plasma opponents still doesn't make it efficient to put it on your quantums.

[MatGB]

See, um, it's a lot of work, but your comparisons are based on single weapon hulls at almost all points.

Against the AI, a fleet of BioAdaptives with 2 plasma guns, one diamond armour, engines and absorption field is lethal and I use it fairly constantly.

If building a Robotic hull, I use either 2 guns, 2 armour and engines or 3 guns, one armour and shield. There is no way, at all, I'm going to put 3 out of 4 external slots for armour, ships are mobile gun platforms. Incredibly high structure ships are only good for making raiders bounce off you and if I ever see an enemy building that sort of thing I switch priorities.

Your maths says an organic is better than a ravenous. Well, if you're only putting one gun on a 5 slot hull, then of course it is. The advantage of more slots is more firepower, if you don't use the extra firepower then cheaper hulls are going to be better.

As for the Energy Frigate? Yeah, it's underpowered in straight up fights. But it's also the fastest 'cheap' warship without engines, and that extra turn of speed can be very useful in the early to mid game at taking systems behind/around enemy chokepoints. If you spread your forces out to force the AI to split up then you don't need a big stack fight that often.

I want to do more work to balance hulls at some point, but this analysis is so far against what I've had reports from from so many different players that it's hard to use, generally feedback I get is I use too much armour, I use way less than you do.

[slv]

See, um, it's a lot of work, but your comparisons are based on single weapon hulls at almost all points.

Against the AI, a fleet of BioAdaptives with 2 plasma guns, one diamond armour, engines and absorption field is lethal and I use it fairly constantly.

If building a Robotic hull, I use either 2 guns, 2 armour and engines or 3 guns, one armour and shield. There is no way, at all, I'm going to put 3 out of 4 external slots for armour, ships are mobile gun platforms. Incredibly high structure ships are only good for making raiders bounce off you and if I ever see an enemy building that sort of thing I switch priorities.

Your maths says an organic is better than a ravenous. Well, if you're only putting one gun on a 5 slot hull, then of course it is. The advantage of more slots is more firepower, if you don't use the extra firepower then cheaper hulls are going to be better.

As for the Energy Frigate? Yeah, it's underpowered in straight up fights. But it's also the fastest 'cheap' warship without engines, and that extra turn of speed can be very useful in the early to mid game at taking systems behind/around enemy chokepoints. If you spread your forces out to force the AI to split up then you don't need a big stack fight that often.

I want to do more work to balance hulls at some point, but this analysis is so far against what I've had reports from from so many different players that it's hard to use, generally feedback I get is I use too much armour, I use way less than you do.

Well, It wasn't exactly my preference, just that weapon/armour slot ratio turned out to be "optimal" for respective sense of being optimal (damage*structure).

Let's take your robotic hull example.
1 Laser IV Triple zortrium ship has 11 damage 58 structure and costs 88 PP.
3 Laser IV single zortrium has 33 damage 36 structure and costs 136 PP
So for the cost of two 3laser/armour ship I can build three 1 laser/3 armor ships.
So when you will have a 66 damage 72 structure fleet I will have 33 damage 174 structure fleet which will destroy your fleet which will have somewhere around 50% chance of destroying your fleet without losing a ship (in most cases both fleets are destroyed).
If it's 4 ships vs 6 you almost always lose a fleet.

For double laser and double zortrium robotic hull ship you turned out to be correct. I can't actually tell how did it happen, but it's indeed better. I made an adjustment in initial post.

For Ravenous, 1 Plasma IV with 4 armours has 18 damage 77 structure and 97 PP
If you put 4 Plasma guns and one armour you get 54 damage 41 structure ship for 161 PP

Again, after we build them in 5/3 ratio we will see a 90 damage 380 structure fleet against a 162 damage 123 structure fleet. The second fleet will almost always lose a battle.

It is possible that I didn't make it clear but all designs I used were not just "Design I play in game" but ones which make combat rating the largest possible. As soon as I decided that damage*structure was the way to compare designs I was bound to display only those design which make the better damage*structure fleets.

For Bioadapters it's completely different story, if you were attentive you saw that in my BioAdaptive analysis I considered 3 Plasma guns Adsorption field design, so I have to agree that you put too much armour on your ships

If you really really are convinced that it's better to have way more guns than listed designs use feel free to suggest a specific design.

For the record I am actually using more weapons in my ships since it's a bit better against planets and I don't like it when battles last more than one round (the last can happen if your weapon/armour ration is as high as "optimal" designs have)

UPD: Whoops, I found a mistake in calculations, 2 laser double zortrium is indeed better than one laser triple zortrium. Can't realize how come did it happen, but after a quick recheck it should be the only mistake.

UPD2: Also if you don't believe me that say 1 Plasma gun 4 Diamond is optimal for ravenous, feel free to open FreeOrion make it, look on its "Estimated Combat Rating per PP", add a laser instead of one Plate and see that it dropped. That thing displays (Damage*Structure)^0.6 / Production_Cost which isn't that far from a thing I was estimating (I am having 0.5 as a power instead).

[mem359]

Single weapon ships are "optimal" since armor is significantly cheaper than a comparable weapon.
I'll have to think if it makes sense to build "tanks" instead of "cannons".
(Obviously, that changes if you have Etty or Mu Ursh for that extra weapon rank.)

Maybe I'm remembering wrong, but didn't ship costs go up as more ships were in service?
Did this go away?
If so, then that makes organic hulls a lot more practical than what they were in the past.
(It also means I shouldn't ever be scrapping ships. )

[MatGB]

Single weapon ships are "optimal" since armor is significantly cheaper than a comparable weapon.

This is in fact something I want to address in a future balancing pass, armour is distinctly undercosted at the moment and that will need to change, not sure by how much, it'll take testing.

Maybe I'm remembering wrong, but didn't ship costs go up as more ships were in service?
Did this go away?
If so, then that makes organic hulls a lot more practical than what they were in the past.
(It also means I shouldn't ever be scrapping ships. )

It hasn't changed, so yes, the maths gets more and more wrong over time, and you should still be scrapping ships.

A change is in the making, have a look at the "Influence" threads, at some point hulls will each have an influence cost and influence will be produced via a focus setting in the same way that production and research is, at which point different metrics will apply and I will be intentionally balancing things to make smaller ships and mixed fleets more common/optimal.

I'm busy, but note my optimal design is 2 guns/2 armour without shield or 3 guns/1 armour with shield (which is also what the AI algorithm comes up with at current costs), and in game given the AI uses a mix of shielded and unshielded ships using my preferred designs I've tested them intensely.

For what it's worth, one of the reasons I dislike algorithmic analysis at this sort of level is one you already state, you don't use these designs, this is what your maths says is optimal. If playing experience says the 'optimal' design isn't, then the maths is wrong.

[mem359]

Okay, I see that the ship costs do go up.
I had forgotten the original cost, and the numbers are already "baked in" when I'm designing.

Suggestion 1: Have a note in one of the windows (production? ship build?) about how much ship costs have gone up.
Maybe "With 35 ships in service, the ship building costs have gone up by 34.5%".
It would give me some idea whether it is worth it to scrap a fleet, or keep it around a little longer for protection.

That completely changes the math for slv's analysis.
A single ship will be more effective than twice as many ships with half as many slots.

------

My other suggestion has to do with drydocks.
If I am producing ships near the front line, I will build a drydock (for instant repairs), regardless of the type of ships I plan on building.
That means I am essentially getting Robotic hulls for "free", compared to other hulls that require other shipyard additions. I'd need some kind of benefit for building an Organic or Asteroid hull, instead of Robotic.

Suggestion 2: Have drydocks only instant repair for base hulls and robotic-type hulls.
Give a repair bonus for other types of hulls, but not full repair. (25% of total value? +5 repair?) This should be in addition to any repair technology.

Suggestion 2a: Have organic incubators do instant repair for organic-type hulls, and a repair bonus for other types of hulls (but not full repair).
Suggestion 2b: Have asteroid processors do instant repair for asteroid-type hulls, and a repair bonus for other types of hulls (but not full repair).
Suggestion 2c: Have energy compressors do instant repair for energy-type hulls, and a repair bonus for other types of hulls (but not full repair).

(The repair benefits should not stack, just the best one in the system should apply for a particular ship.)

If this sounds like too much coding work, then at least have org incub. / ast. proc. / energy comp. also have the "full repair" abilities of the drydock.

[UrshMost]

A few thoughts:

I always use rock armour instead of diamond. It's cheaper to research, cheaper to build, and has the same defence strength. You do need to find an asteroid belt and build the two needed buildings there though. My armour progression is usually zortium -> rock -> neutronium.

Scattered Asteroid hulls are nice, but you have to get through the 1600 RP to research monomelecular armour first, besides all the other research prereqs. I would rather research titan hulls which have more external slots and more structure for only a small per hull additional cost. (In a really long game I will research scattered asteroid hulls in the late game to have a few ships to give the shield bonuses to my main fleets.)

Primarily I think you are overlooking the value of the survivability of a ship. If I have a few big hulls against a fleet of smaller ships, most of those smaller ships are going to die, and dead ships can't be repaired. If I have a titan hull with only 5 hp left it's still a zero cost repair away from being back at full strength.

I use the same logic for shields. Yes they are more expensive initially, but are more likely to survive and be repaired, allowing them to fight in more battles, destroying more enemy ships.

Imagine a cost vs destroyed enemy ships. Initially say my robotic shielded ship costs twice as much as a defender, but survives a fight, is repaired for free, and destroys another defender, etc. After 5 battles, my 2x cost ship has destroyed 5 1x cost enemy ships.

[slv]

Single weapon ships are "optimal" since armor is significantly cheaper than a comparable weapon.

This is in fact something I want to address in a future balancing pass, armour is distinctly undercosted at the moment and that will need to change, not sure by how much, it'll take testing.

Fixing this will make weapon-heavy designs better, if armour costed the same as weapons "optimal" designs would use smth close to a half/half split (and more weapons the higher initial structure is).

Maybe I'm remembering wrong, but didn't ship costs go up as more ships were in service?
Did this go away?
If so, then that makes organic hulls a lot more practical than what they were in the past.
(It also means I shouldn't ever be scrapping ships. )

It hasn't changed, so yes, the maths gets more and more wrong over time, and you should still be scrapping ships.

It doesn't go wrong as fast as you might think. Say if you compare fractal vs quantum 0.112 vs 0.88 fractals become better only if you have more than 15 of them (15 fractals vs 30 quantums). If you take into account that you have non-combat vessels too you need even more. If you have 10 non-combat ships fractals aren't better until you have 20 fractals against 40 quantums.

This effect is really hard to estimate, so I decided that it's better to have combat rating not taking it into account. I still can remember the existence of an effect and make adjustments.

I'm busy, but note my optimal design is 2 guns/2 armour without shield or 3 guns/1 armour with shield (which is also what the AI algorithm comes up with at current costs), and in game given the AI uses a mix of shielded and unshielded ships using my preferred designs I've tested them intensely.

Well, the first one is optimal indeed (Due to a mistake I had 1 gun/3 weapons as an optimal listed initially). For Asteroid Hulls 1 weapon/3 armor is slightly better, but the difference isn't noticeable.

3 weapons 1 gun 1 shield is actually bad. Of course if your opponent is wielding Mass drivers it may be good, but rarely so.
Triple Laser single Zortrium with Energy Field is worse than 2/2 shieldless unless your opponent has 6 attack weapon or less. So even Laser 2 makes it obsolete, MD4 is close to a break-even with a 4% improvement for a shielded design.
Triple Laser single Zortrium with Deflector is worse than 2/2 shieldless unless your opponent has 8 attack weapon or less. So even Laser 3 makes it obsolete.

That was one of main points of my analysis, AI seems to overvalue shields. Players seem to overvalue shields too. They still have some place, in case you are rushing for Quantum Networking and you happen to have shields while most your neighbours have just Mass Drivers they are ok but note that for Energy Field against MD4 the improvement is very tiny, and the loss when your opponent will get Laser II or III is much larger.

I guess in AI vs AI fight AI thinks opponent will be on MD 4 for a long time and goes for 4% improvement, in that case it shouldn't go for it if it sees a laser or an Etty neighbour.

Shields are really good in case your opponent lags behind in tech, is that their purpose? My "shields are bad" argument applies only to equal-technology opponents (deflector vs plasma, energy grid vs laser). I guess there is a tiny gap when shields are good when you have energy grid against mass drivers, but other shields have large enough RP cost so that at the moment you are likely to research them they are already outdated.

For what it's worth, one of the reasons I dislike algorithmic analysis at this sort of level is one you already state, you don't use these designs, this is what your maths says is optimal. If playing experience says the 'optimal' design isn't, then the maths is wrong.

Well, math isn't wrong per se. If original model didn't take into account the fleet upkeep it's reasonable that it (model/math) thinks that 0.089999 ship than twice smaller ship with a rating of 0.90000, while that's clearly not the case.

Let's say I have a choice between 2x plasma 5x diamond 0.11 quantum ship and a 3x plasma 4x diamond 0.10 ship. Math says the first one is better in head-to-head combat and it indeed is. On the other hand the 10% difference isn't large enough to matter until I wield fleets of dozens of them and I may think that using inferior ship can be better because of factors outside of analysis (fleet upkeep/planetary defence favouring high laser ships/having just enough damage to kill Sentinel in one turn). Weighting these factors is subjective and I am sure I am doing it wrong in many games, that's why I decided to have an objective parameter ("best" ship for a given hull). If you add additional weapon for most designs it doesn't decrease rating much and neither does it make one hulls better than others.

If you change "optimal" ships for a "real" ships than 1)all values will be decreased in a similar rate, not changing overall tendencies 2)I will have troubles determining what a "real" ship is, as you said some people play with more armour, some play with less. Personally I rarely change more than 1-2 slots from "optimal".

I am not very fond of relying on player experience in a vacuum. People are often wrong, and given it's really easy to win over AI it's often not easy to spot what is efficient and what's not. Until I did the calculation I was frequently using 3 Plasma double Diamond Ravenous ships, imagine how shocked I was when I realized it's twice worse than an optimal one. I was winning with them, I am winning with better ships now, sometimes it matters.

A few thoughts:

I always use rock armour instead of diamond. It's cheaper to research, cheaper to build, and has the same defence strength. You do need to find an asteroid belt and build the two needed buildings there though. My armour progression is usually zortium -> rock -> neutronium.

This is something I have to try. Although it doesn't change most ratings since the armour PP cost is negligible even for Diamond.

Scattered Asteroid hulls are nice, but you have to get through the 1600 RP to research monomelecular armour first, besides all the other research prereqs. I would rather research titan hulls which have more external slots and more structure for only a small per hull additional cost. (In a really long game I will research scattered asteroid hulls in the late game to have a few ships to give the shield bonuses to my main fleets.)

The real reason I included Scattered hulls was to find the reason to use shields against an equally teched enemy. Both titan and Scattered fit into "lategame" category better, they are efficient with shields and shield math is somewhat complicated, weapon levels start to matter, speed of ship starts to matter, antimatter cannon makes things horrible, etc.

Primarily I think you are overlooking the value of the survivability of a ship. If I have a few big hulls against a fleet of smaller ships, most of those smaller ships are going to die, and dead ships can't be repaired. If I have a titan hull with only 5 hp left it's still a zero cost repair away from being back at full strength.

I use the same logic for shields. Yes they are more expensive initially, but are more likely to survive and be repaired, allowing them to fight in more battles, destroying more enemy ships.

Imagine a cost vs destroyed enemy ships. Initially say my robotic shielded ship costs twice as much as a defender, but survives a fight, is repaired for free, and destroys another defender, etc. After 5 battles, my 2x cost ship has destroyed 5 1x cost enemy ships.

This wasn't intentional and is something which is really hard to take into account (it's a second-order effect in a sense). But it's not that big as you think it is. If I tweak your example into a more real-world one where a Titan ship survives with 20% hp turns out that a fleet of quantums (140hp) ships will also most likely survive a fight with ~20% hp on each ship. Each ship takes 8 hits to die (assuming Plasma IV) and 8 is large enough to distribute hits more or less uniformly. If Titan ship just barely survives the battle than it's indeed noticeably better than a fleet of quantums, but I am not sure it happens often

For a Robotic ship example nothing changes if you substitute "robotic shielded ship" for "3 nonshielded organics". They will win first fight without suffering casualties (although each ship will be damaged), get repaired for free, destroy another defender, etc. If your fleet is, say, twice stronger and your ships die from, say 6 hits you will lose just 5% of your ships. If your ships survive more hits the damage distribution will be even more uniform. Of course if a fleet barely survives these fight robotic is better. And I have to note that nonshielded ships are noticeably more efficient than shielded ones, so if a shielded ship barely survives a fight then a fleet of shielded ones will survive it non-barely.

Also I need to note that there exists an advantage for small ships I am not taking into account too. You can split the fleet or leave one ship to block a planet. Let's hope these two effects cancel, so the answer is close to reality


P.S. I added a bit on lategame in the initial post, turned out that if players have Great Pilot species it's possible to find out what happens (the reason is that Great Pilots bonus makes death ray better than Antimatter cannon, eliminating blackshield/cannon/deathray rock-paper-scissors complications). Result is close to what you would have expected,
Solar/Scattered>Titan>Quantum, Titan + one scattered > fleet of scattered, Solar + one scattered > everything else.
What wasn't obvious is that Quantums are competitive, and in case of Misiorla Ultimate Pilots bonus become stronger than Titanic hulls, though it is quickly reversed when fleet upkeeps starts ticking, somewhere around 10 Quantums they are inferior again like they should be.

[Bluehand]

Thanks for doing this analysis, there's some good information here.

[Morlic]

That was one of main points of my analysis, AI seems to overvalue shields.

Well, the AI currently assesses his design options based on some estimate of what enemy it will encounter. And since it does not predict any kind of tech advances, it will simply look around and see what the enemy ship strength is. If there are many ships with Mass Drivers or many monsters with weak attack, then it is likely to build shields.

So no, the rating algorithm does (i.e. should) not overvalue shields but yes, it uses inadequate information / no foresight which may render the designs useless quickly.


As the AI algorithm output was mentioned as "optimum" design: Please note that the AI tries to design so that a single ship has optimum attack*structure/cost. Note the linear cost weighting instead of the quadratic (which you get since you try to design for an optimum fleet).
Reasons for that include but are not limited to fleet upkeep considerations (which you would need to plan for with your rating scheme), the fact that the AI may have troubles combining its fleets and the fact that the AI generally sucks at retreating from a battle.

Also, there are weird combinations that are local maxima at some point of the game (given the AI research path) such as Spatial Flux hulls with Mass Drivers, Small Asteroid Hulls being better than normal Asteroid Hulls, and Robotic Hull is supposed to use 1 Mass Driver and 1 Laser each (at least at some point in tech...).
These are combinations that require additional boundary conditions to explain why they suck (which they obviously do, e.g. because of fleet upkeep and extremely high losses even when victorious). So formula gets more complicated, harder to maintain and the general result is not necessarily better in an actual game: Fleet-based calculations are prone to use continuous math where the game works with discrete math, i.e. you need a high number of ships so that results become accurate and advantages really play out (unless differences are large enough of course).

[UncleFred]

I wonder about the impact of the combat system. It is not just if ships survive the turn to be repaired, but also (I think) when they are destroyed. Ships killed in the first or second round are no longer around to draw fire nor to do damage. If each of your ships fires once and each of my ships fires twice in a round I may have a better change to destroy your ships before the third round. Conversely, my advantage drops twice as much for each ship I lose compared to you. Now shields, reduce your ability to damage my ships, which makes my structure more "valuable" than yours but also more expensive. However if I want shields in late game I have to research the early shields anyway. Similarly if I'm developing the asteroid hull progression, what is the real cost of crystal armor?

In late game I tend to favor Quantum, Titan, and Solar hulls. Fractals can be configured for a lot of hitting power, but without shields have poor survivability. If I have a logistics hull along, anything that survives gets repaired immediately so survivability is key, which make fractals a poor trade off.

Perhaps it's not really damage x structure. Perhaps it's (total weapon count x expected per weapon damage) vs (opposing ship count x average ship structure). In other word how many of your ships can I expect to destroy in a round compared to how may ships should I expect to lose in a round. This is in part why shields have value in the late game. In effect my shields dramatically reduce your weapon count. This is also why combat bonuses/penalties can rather dramatically shift the equation.

There's a lot of value in your post, but I suspect that optimizing ship design in terms of winning combat is more complex than optimizing the metrics you've chosen.

[slv]

I wonder about the impact of the combat system. It is not just if ships survive the turn to be repaired, but also (I think) when they are destroyed. Ships killed in the first or second round are no longer around to draw fire nor to do damage. If each of your ships fires once and each of my ships fires twice in a round I may have a better change to destroy your ships before the third round. Conversely, my advantage drops twice as much for each ship I lose compared to you. Now shields, reduce your ability to damage my ships, which makes my structure more "valuable" than yours but also more expensive. However if I want shields in late game I have to research the early shields anyway. Similarly if I'm developing the asteroid hull progression, what is the real cost of crystal armor?

This is an effect which is very important for early game and for mixed fleets (fleets with different types of designs inside). In early game, it's not exactly a structure which matters, but rather a number of hits ship needs to take to die, 25 structure ship is the same as 30 damage ship, they die to an equal number of Laser IV (3 shots) and MD IV (5 shots) projectiles.

But outside of early game, most ships (or rather, most damage*structure/PPcost^2 "optimal" ships) have really high structure, which makes them dead only to ~8 shots of respective weapon. As soon as ships are that healthy an interesting effect appears, given that currently weapons choose a random target most of the fleet will be damaged in the same way. i.e. in 20 vs 20 ships battle where each ship carry two weapons most ships will be shot ~2 times in the first round, ~4 times in the second round. Of course there will be some flukes and few will be hit much more than the expected, but the overall distribution will be normal.

What this implies that in Fleet vs Fleet battles consisting of one design all ships will die in the same round, if you see a considerable amount of ships dying on first round that usually implies that your whole fleet died at the first round or have just a few ships with one-hit structure and they will die in the second round.

In late game I tend to favor Quantum, Titan, and Solar hulls. Fractals can be configured for a lot of hitting power, but without shields have poor survivability. If I have a logistics hull along, anything that survives gets repaired immediately so survivability is key, which make fractals a poor trade off.

I also tend to favour Quantums now (I was a Fractal guy before), but not for the reason you stated, I play with shieldless quantums right now.

So let's try to find out what happens if you see a fight between shielded and shieldless quantums. [Without any damage*structure things, just fleet vs fleet considerations]

Shielded quantum design constists of 3x Plasma Gun IV, 4Diamon Armour and a PlasmaShield
272 PP, 18*3=57 damage 122 structure, dies from 14 hits
Shieldless consists of 2x Plasma 5x Diamond
150 PP, 9*2=18 damage, 140 structure, dies from 8 hits

Approximately 5:9 ratio

What we will see in a real battle will be pretty much a tie. The Fleet of 5 shielded ships will be damaged by 80% on average (11 hits out of 14) and the fleet of unshielded ships will be damaged by 69% on average (5-6 hits out of 8 ).
And if we do include that damage is distributed non-randomly for small fleet sizes it will make it possible for the shielded ship to die, not large, but still greater than the respective chance of a nonshielded ones. The smaller amount of ships you have the greater variance in ship damage will be.

If instead both players have Good Pilots species, shielded ships will be destroyed, still not being able to destroy enough opposing ships. Great Pilots will make it so Shielded ships will be destroyed in Round 2 of combat (again, all at the same time).

Amusingly enough, if a person builds a
3x Plasma 4x Diamond ship (it's a bit non-optimal, 10% less optimal) with 182PP 9*3=27 damage and 122 structure it will be able to destroy opposing shielded fleet (instead of a tie).

And of course any Deflector Shield design will lose to shieldless without causing any casualties.

So I insist that even if you have shields researched against an equally teched enemy it is better to not put shields on your ships.

What happens in a real game that you sometimes have deflector/plasma researched when your opponents still have Laser II. In such cases it is indeed efficient to put them on your ship, though at this level of steamrolling you don't really care what is efficient and what is not.

Perhaps it's not really damage x structure. Perhaps it's (total weapon count x expected per weapon damage) vs (opposing ship count x average ship structure). In other word how many of your ships can I expect to destroy in a round compared to how may ships should I expect to lose in a round. This is in part why shields have value in the late game. In effect my shields dramatically reduce your weapon count. This is also why combat bonuses/penalties can rather dramatically shift the equation.

There's a lot of value in your post, but I suspect that optimizing ship design in terms of winning combat is more complex than optimizing the metrics you've chosen.

In early game it's rather damage*(number of hits needed to destroy a ship). In midgame there are two factors, one being a damage*structure, other being a damage/structure ratio.

If your ratio is too low (like "optimal" ships have) it makes possible to have "ties" when two opposing fleets are able only to deal 60% damage to each other. If you don't want this to happen you can swap a plate for a gun, winning a few otherwise tied battles with a cost of losing few otherwise tied battles. And "Ties" are not neccesarily a bad thing, you can retreat from an otherwise loosing battle or if you are winning first one the second one will be won in the first or a second round still not causing you to suffer casualties.

Also keep it in the mind that damage*structure "mnemonic" I was using isn't really applicable to mixed fleets. It can be applied only if all ships have the same (effective) structure, otherwise ships will not die at the same time (even approximately), and the later is one of the cornerstone assumptions of the model. I am not particularly sure, but it seems that having non-equal structure of your ships decreases your fleet strength, so mixing is usually bad. I usually try to have old designs work on a different theatre separately from new ones.

[UncleFred]

I'm curious why did you decide to move away from Fractals to Quantums?

[slv]

I'm curious why did you decide to move away from Fractals to Quantums?

I played with 6x Plasma 8x Diamond fractals before. (108/184 ship for Plasma IV humans).

At some point I realised that if I build two 3x Plasma 4x Diamond Quantums and group them together I will get a (108/244) fleet instead.

Differences:

1)One may think that sometimes fractal is better, i.e. when it barely survives the fight. Turns out that if enemy dealt ~180 damage (say 10 plasma hits) it is really unlikely that this 10 plasma hits will distribute as 7+3 to kill even one my ship. And if it takes 11 plasma hits the fractal will die while at least one of my ships will survive and in the most frequent 5+6 case both will survive.

And in case we are comparing 2 fractals vs 4 quantums the situation will be even worse, now there is a higher chance to lose fractal from damage discrepancy than to lose a quantum, and each loss is twice costly.

This is something most people don't realise, they remember cases when their titan hull survived on 4 hp but they forget times when they sent 2 titanic hulls to battle and one died while second had 50 hp. Discrepancy is actually decreased when you send more ships to battle, the chance to have one of your two 900-structure ship die to 30*50 damage is higher than the chance of any of four 450 ships to die from the same damage.

2)I can split my quantums, this one turns to be major since this ship is so big that they can operate in a cruiser-like fashion, you send just one "cruiser" against enemy fleet.

3)Quantums will win in a head-to-head fight.

4)Quantums are faster to build. This is actually important. I tend to order my researches in a way that I get Quantums,Plasma and Diamond in the same turn, so I have to wait till I get a technological fleet flying around with laser/zortrium ships, 7 turns sometimes feel like an eternity.

5)Quantums make a better troopships. Higher slot/PP ration, higher structure, smaller "overkill" while invading.

6)The only disadvantage is a fleet upkeep.

This is a crucial thing and it will overcome all other things eventually. Question is when does this "eventually" come. Honestly I skipped this step thinking "Well, unless I build 10 cruisers the effect is not noticeable and if I did then I won the game", but we can do it now.

Let's briefly check when does it make Fractals to be better in head-to-head fight. (Rough estimate giving the approximate magnitude of an effect)
I usually have a decent amount of non-combat ships flying around (and I am somewhat bad at scrapping old designs, usually send them to invade some backyard-AIs), say 20 of them.
So if I have 20 ships and decide to build x fractals I will pay 1.2 times the cost for the first one and (1.2+x/100) for the last one averaging at (1.2+x/200).

If I build 2x quantums instead I will average (1.2+x/100). For damage*structure/PP_cost -type estimate what we are asking is for what x does

(244/184)*(1.2+x/200)^2 / (1.2+x/100)^2

become less than 1.

And the answer is ~40. So until I start to compare 40 fractals vs 80 quantums the structure advantage of my quantums is enough to overcome fleet upkeep.

Right now after I did calculations in the initial post I am playing both with 2x Plasma 5x Diamond and 3x Plasma 4x Diamond. Second one has an advantage of not having "ties" and being better in planet invasion, first one is better combatwise.

On High planets I favour 3x Plasma design, you often have to invade 5-planet systems and it can be dangerous unless you have enough firepower to finish the shields quickly.
On low/medium planets 2xPlasma 5x Diamonds is good. I was terrified to think of "ties" and not being able to kill opposing fleet, but it turned out that

1)since AI quite often have mixed fleets when a "tie" happens you usually destroy weaker ones, making second battle easier. This makes you win battles you might otherwise lose

2)You can run away after the tie. Say if the battle was really close you can repair your ships.

3)You can follow the enemy further and within 2 battles and 6 rounds of combat it is not possible to have a tie. 120 speed makes it impossible to run away. Also sometimes AI doesn't retreat at all.

[AndrewW]

3)You can follow the enemy further and within 2 battles and 6 rounds of combat it is not possible to have a tie. 120 speed makes it impossible to run away. Also sometimes AI doesn't retreat at all.

Not impossible, one can use engine parts or other technology to speed up their ships.