Measuring efficiency of ship designs

For topics that do not fit in another sub-forum.

Moderators: Oberlus, Oberlus

Message
Author
defaultuser
Vacuum Dragon
Posts: 630
Joined: Wed Aug 26, 2015 6:15 pm

Re: New Fuel Settings?

#31 Post by defaultuser » Wed Sep 11, 2019 9:56 pm

Robotic hulls are best used with robotic species, so shields would vary by fleet size.

User avatar
JonCST
Pupating Mass
Posts: 91
Joined: Sat Dec 15, 2018 4:28 am

Re: New Fuel Settings?

#32 Post by JonCST » Wed Sep 11, 2019 10:05 pm

Hi all.
defaultuser wrote:
Wed Sep 11, 2019 9:56 pm
Robotic hulls are best used with robotic species, [...]
Robotic Hulls are best used with species who need a fast-to-build, sturdy, and capable warship, especially for breaking planetary shields.

IMO, of course.

Jon

PS: i almost never use robotic interface shields on them. It's PP cheaper to layer on DAP and invest in repair tech. Especially when facing mines and/or fighters.

PPS: at least, when playing against the AI. Haven't played enough PVP to make an intelligent comment on that, yet.

User avatar
alleryn
Dyson Forest
Posts: 206
Joined: Sun Nov 19, 2017 6:32 pm

Re: New Fuel Settings?

#33 Post by alleryn » Fri Sep 13, 2019 4:14 am

A few things (none of which really have to do with Fuel Settings. I suppose we really ought to shift this discussion to a different thread.):
  1. I think it's worth still considering the utility of Ophiuchus's max_hits_survival instead of min_hits_to_die to model structure in some scenarios. Specifically, especially early in the game, when you are taking out space monsters guarding systems you want, you may desire to bring enough ships not to just win the fight, but to win it with minimal losses (at least i think that's a good strategy).

    This is different from when you are battling another human player, where typically your forces are going to be pretty evenly matched (all things being equal), and your main concern is how to dedicate your PP to your fleet in order to obtain space supremacy.

    When fighting neutral monsters and wanting to preserve your fleet, i think this is where max_hits_survival make sense to use in calculations of the sort-of-minimum fleet you want to take to clear them out. Haven't done any real calculations here; just wanted to mention it.
  2. Regarding robo shields, i made an offhand comment about how them being cost effective in some situations.

    So i was looking over solar web during my last game, trying to remember how it worked (Briefly for anyone who doesn't recall: it flatly adds to laser damage on organic ships. The amount of the flat increase is determined by star color (1.5-4.5), and with the second tech, gets a mutiplier that increases linearly from 1x to 8/3x as you go from 6 to 16 ships, then with reducing gains like a sqrt function from 8/3x at 16 ships to 5x at 121 ships.)

    Robotic Interface Shields work somewhat similarly, with the effect increasing from 1 shield to 7 shields linearly with number of ships, then dropping off like a sqrt until maxing out at 20 shields at around 57 ships (19.97 at 57, 20 (max) at 58).

    Anyways, the robo shield is pretty pricey, but obviously if you build enough ships with them they will outclass a weapon tech of similar RP cost (i.e. lasers) and be cost effective (especially if they start reducing damage from 2 to 1 or 1 to 0).

    The solar web, since it's increasing instead of decreasing damage differential (so has decreasing marginal returns per point of damage||shield instead of increasing), doesn't really have the same allure, although it probably has a niche if you get outclassed by plasma shields or multispec from a ruin.

    Not sure whether this is related:
    JonCST wrote:
    Wed Sep 11, 2019 10:05 pm

    PS: i almost never use robotic interface shields on them. It's PP cheaper to layer on DAP and invest in repair tech.
    What is DAP?
  3. Built i think my first carrier ship ever in the game i'm playing now, so i'm having fun with fighters. I always imagined they were pretty much outclassed by beam ships, and i still mostly think so. I thought the general consensus was that fighter fighters were the strongest, but they feel like the weakest of the 3 types to me (mostly due to the slot cost associated with launch bays, etc), with really only interceptors in a mixed fleet looking cost effective, perhaps early on supported by a bomber ship as that will scale better into the laser era than mass drivers.

User avatar
JonCST
Pupating Mass
Posts: 91
Joined: Sat Dec 15, 2018 4:28 am

Re: New Fuel Settings?

#34 Post by JonCST » Fri Sep 13, 2019 5:35 am

alleryn wrote:
Fri Sep 13, 2019 4:14 am
JonCST wrote:
Wed Sep 11, 2019 10:05 pm

PS: i almost never use robotic interface shields on them. It's PP cheaper to layer on DAP and invest in repair tech.
What is DAP?
Diamond Armor Plating
alleryn wrote:
Fri Sep 13, 2019 4:14 am
[...]. I always imagined [fighters] were pretty much outclassed by beam ships, and i still mostly think so.
Some important points about fighters:

o They bypass shields.
o They act as "ablative shields", often taking hits instead of the carrier.
o They use the highest researched fighter weapon at no increase in PP cost.
o They don't incur "maintenance costs" for having more ships.
o They refill for no cost if the carrier is in supply.

Major fighter drawback: They do not participate in planetary attack or defense.

SGH ships loaded with fighters (not interceptors nor bombers) and one beam weapon is a potent recipe against an AI player.

Perhaps even stronger is a stealthed protoplasmic hull with 6 plasma cannon fighters and no beam weapons.

Future nerfing of stealth carrier or improvement of AI may change that, of course.

And, we should probably relocate this discussion, since it now has nothing to do with fuel.
:wink:

Jon

User avatar
Oberlus
Cosmic Dragon
Posts: 1490
Joined: Mon Apr 10, 2017 4:25 pm

Re: New Fuel Settings?

#35 Post by Oberlus » Fri Sep 13, 2019 10:37 am

alleryn wrote:
Fri Sep 13, 2019 4:14 am
we really ought to shift this discussion to a different thread.
JonCST wrote:
Fri Sep 13, 2019 5:35 am
we should probably
I can do that. I'm a bit busy and groggy, what would be the best post to split this thread?
Pleasepleaseplease, join fifth multiplayer slow game!

User avatar
alleryn
Dyson Forest
Posts: 206
Joined: Sun Nov 19, 2017 6:32 pm

Re: New Fuel Settings?

#36 Post by alleryn » Fri Sep 13, 2019 11:29 am

Oberlus wrote:
Fri Sep 13, 2019 10:37 am
what would be the best post to split this thread?
Let's move #22 viewtopic.php?p=97091&sid=a71f8b2c358b5 ... ddb#p97091 and after, i think.


The last post that was on topic was #20: viewtopic.php?p=97087&sid=a71f8b2c358b5 ... ddb#p97087 (but just parts of it).

Starting with #16: viewtopic.php?p=97064&sid=a71f8b2c358b5 ... ddb#p97064), the conversation began to diverge from "Can the robo hull make do with 1.5 fuel or does it need 2?" to "Could the robo hull still serve a combat niche with 3 external slots instead of 4?". (As there was a possibility of trading an external slot for additional fuel).

But i think between #21 and 22 is the cleanest break. After this the conversation is clearly devoted to what the measure of cost efficiency should be, with no further reference to fuel specifically that i can see.

User avatar
Oberlus
Cosmic Dragon
Posts: 1490
Joined: Mon Apr 10, 2017 4:25 pm

Re: New Fuel Settings?

#37 Post by Oberlus » Fri Sep 13, 2019 1:54 pm

alleryn wrote:
Fri Sep 13, 2019 11:29 am
Let's move #22 viewtopic.php?p=97091&sid=a71f8b2c358b5 ... ddb#p97091 and after, i think.
I went a bit farther and splitted at #18. Let the party continue.
Pleasepleaseplease, join fifth multiplayer slow game!

User avatar
JonCST
Pupating Mass
Posts: 91
Joined: Sat Dec 15, 2018 4:28 am

Re: New Fuel Settings?

#38 Post by JonCST » Fri Sep 13, 2019 1:57 pm

alleryn wrote:
Fri Sep 13, 2019 11:29 am
[...] After this the conversation is clearly devoted to what the measure of cost efficiency should be, with no further reference to fuel specifically that i can see.
Maybe two threads, since i don't think the fighter stuff fits the cost efficiency discussion. Or, maybe it identifies a factor which needs to be included. Dunno.

Jon

User avatar
Oberlus
Cosmic Dragon
Posts: 1490
Joined: Mon Apr 10, 2017 4:25 pm

Re: Measuring efficiency of ship designs

#39 Post by Oberlus » Fri Sep 13, 2019 2:15 pm

If what you are looking for is a single index (number) to measure the overall efficiency of each design against all the rest, you won't find it.
We can blame the intrinsic rock-paper-scissors nature of the combat system due to the non-linear correspondence between damage per shot of a weapon and structure loss at the target (a 30 damage shot to a ship with 10 remaining structure points does 10 damage, etc.). So we get ship designs for which A>B>C>A, and there is no single number to express that kind of multidimensional ordering.

However, we can get efficiency indices for specific pairs of ship designs (and even pairs of fleet compositions). In other words, you can have a measure of the efficiency of [robo hull + 2 laser + 2 plates + shield] vs [multicell + whatever], and a different one vs [robo hull + 1 laser + 2 plates + 1 launch bay + bombers], etc. but we cannot get an infallible measure of the abstract efficiency of the ship design vs everything at once. No matter what equation you use, you'll find scenarios where the efficiency index will fail to predict the result of a battle (I mean, will give a wrong estimation of what is better than what).

In order to do so, we can develop complex formulas that take into account the number of ships and their structure, number of weapons and their damage per shot and fire rate, the randomness of the targetting, etc. Or use combat simulators, which is my kind of ideal. Get 100 battle results and calculate the average (or median, or mode) number of PPs of the surviving ships of each faction compared to the initial ones to compose the efficiency index.
Pleasepleaseplease, join fifth multiplayer slow game!

User avatar
Oberlus
Cosmic Dragon
Posts: 1490
Joined: Mon Apr 10, 2017 4:25 pm

Re: New Fuel Settings?

#40 Post by Oberlus » Fri Sep 13, 2019 2:17 pm

JonCST wrote:
Fri Sep 13, 2019 1:57 pm
Maybe two threads, since i don't think the fighter stuff fits the cost efficiency discussion. Or, maybe it identifies a factor which needs to be included.
Your last words. I think figthers (and any other stuff that affects defense/offense of ship designs) must be considered when estimating the efficiency.
Pleasepleaseplease, join fifth multiplayer slow game!

Ophiuchus
Programmer
Posts: 966
Joined: Tue Sep 30, 2014 10:01 am
Location: Wall IV

Re: New Fuel Settings?

#41 Post by Ophiuchus » Fri Sep 13, 2019 4:38 pm

Oberlus wrote:
Fri Sep 13, 2019 2:17 pm
JonCST wrote:
Fri Sep 13, 2019 1:57 pm
Maybe two threads, since i don't think the fighter stuff fits the cost efficiency discussion. Or, maybe it identifies a factor which needs to be included.
Your last words. I think figthers (and any other stuff that affects defense/offense of ship designs) must be considered when estimating the efficiency.
The structure based cost efficiency gives an estimate for the damage-taking part against fighters. Also if the ship has no shields.

The min_hits_to_die gives a good estimate if you have shields against ship weapons.

The max_hits_to_survive could be useful for finding a good fleet against fleet estimate.
E.g. you take the sum of min_hits_to_die values of the oppenent fleet; then you compare the sum of the max_hits_to_survive values of your fleet. If your max_hits_to_survive value equals the min_hits_to_die of the opponent I guess you will win without many losses. This is my theory at least :wink:

In my sheet I have the three metrics for ships and structure based for fleets next to each other. This gives quite an impression of the value/strengths/weaknesses of the various ship designs.
Any code or patches in anything posted here is released under the CC and GPL licences in use for the FO project.

Furthermore, I propse... we should default to four combat rounds instead of three ...for the good of playerkind.

User avatar
alleryn
Dyson Forest
Posts: 206
Joined: Sun Nov 19, 2017 6:32 pm

Re: Measuring efficiency of ship designs

#42 Post by alleryn » Sat Sep 14, 2019 6:12 pm

Oberlus wrote:
Fri Sep 13, 2019 2:15 pm
However, we can get efficiency indices for specific pairs of ship designs (and even pairs of fleet compositions) [...]

In order to do so, we can develop complex formulas that take into account the number of ships and their structure, number of weapons and their damage per shot and fire rate, the randomness of the targetting, etc. Or use combat simulators, which is my kind of ideal. Get 100 battle results and calculate the average (or median, or mode) number of PPs of the surviving ships of each faction compared to the initial ones to compose the efficiency index.
Yes. I think it's useful both to generate the formulae (in order to hopefully gain some insight into the "why" as well as learning the "which" (fleet is better)), and to have the combat simulator (to check our formulae and because once we have it, of course, it's a lot less work).

There is morlic's old combat simulator: https://github.com/freeorion/freeorion/pull/1175 as well as his/her work on one in an alternate branch https://github.com/freeorion/freeorion/pull/1072. I'm not sure what state they are in, but i think you mentioned you are thinking of developing a simulator (and i've been thinking about this too). If so, those resources may be useful.

Ophiuchus
Programmer
Posts: 966
Joined: Tue Sep 30, 2014 10:01 am
Location: Wall IV

Re: Measuring efficiency of ship designs

#43 Post by Ophiuchus » Sat Sep 14, 2019 9:24 pm

alleryn wrote:
Sat Sep 14, 2019 6:12 pm
There is morlic's old combat simulator: https://github.com/freeorion/freeorion/pull/1175 as well as his/her work on one in an alternate branch https://github.com/freeorion/freeorion/pull/1072. I'm not sure what state they are in, but i think you mentioned you are thinking of developing a simulator (and i've been thinking about this too). If so, those resources may be useful.
Morlics simulator is reimplementing the combat system in python. While I like python I would go another way. I want to use the real combat engine as a simulator in order to have always the "correct" results even if the combat system changes (and not having to do each combat system change twice, doubling mismatch/bugs etc).

The fastest way to do this is to change the c++ backend to provide somewhere an interface where you put in the combat universe and let it process the combat and gives back the combat result (surviving ships and values) and a combat report.

Another way to do this is to reimplement the combat system in another language and change the c++ code to use this "combat system module".
Any code or patches in anything posted here is released under the CC and GPL licences in use for the FO project.

Furthermore, I propse... we should default to four combat rounds instead of three ...for the good of playerkind.

User avatar
alleryn
Dyson Forest
Posts: 206
Joined: Sun Nov 19, 2017 6:32 pm

Re: Measuring efficiency of ship designs

#44 Post by alleryn » Sun Sep 15, 2019 11:06 pm

Ophiuchus wrote:
Sat Sep 14, 2019 9:24 pm
The fastest way to do this is to change the c++ backend to provide somewhere an interface where you put in the combat universe and let it process the combat and gives back the combat result (surviving ships and values) and a combat report.
That doesn't sound too difficult. I suppose the toughest part would be figuring out how to input the fleets to test.
____________________________________________________________________________________________________________________________________________________________________________________________

Plans for this thread going forward:
I want to suggest a general framework about how we approach the question of efficiency of ship designs.

First of all, i suggest we limit ourselves (in this thread) to the question of combat efficiency. Considerations of fuel we already have another thread for. And speed, while certainly an important consideration, is going to be really difficult to evaluate in relative importance in any quantifiable way, i think.

For the most part i think we should also assume that we have the prerequisite buildings finished for whichever hulls we are evaluating. Trying to factor in cost of building a shipyard might be possible, but when we start looking at turns to build each shipyard, that's getting pretty tricky. Maybe we could look at this but it should be pretty low priority or maybe in a separate thread entirely if we want to start factoring that stuff in. Research costs would, i think, fall under the same/a similar umbrella.

Fleet upkeep i think is too essential a consideration to ignore, so i think it should be considered here, but it is fairly nebulous (insofar as it involves "higher order" thinking, the strategic needs of how big a military does one need vs the tactical needs of what fleet is better right here right now), so i'd suggest making that a somewhat low priority.

I don't think we need separate threads for fighters, shields, etc. There's always a tradeoff between having too much disparate info cluttering one thread and too many threads cluttering one subforum. So we can try to find a balance.

I'd also suggest that we lmit this thread to analysis of the current state of affairs. "What if" scenarios are fine as long as we don't get so far into them that we start actually doing calculations that may get confused with calculations about the combat system as it stands.

So far, for my part, i've pretty much just been tossing whatever comes to my mind at the wall and hoping it starts to form some kind of pattern. And i think that technique has been moderately successful. I feel like we've come to a rough consensus that for beam ship vs beam ship battles, (dmg*min_hits_to_die/cost^2) is a reasonable metric for performance efficiency, and i understand a little more about (again for the limited beams vs beams case) how effective a greater number of small ships are vs "an equivalent" smaller number of large ships (though there may be more worth learning here).

I'm most interested in the early game (after all, a thread about efficiency is intrinsically a thread about balance, and if the early game isn't balanced, the late game will probably never happen), and right now i'm mostly interested in studying small fleet battles of varying compositions. Things like 1 beam ship vs 1 carrier ship, stuff like that, and then slightly more complex scenarios with mixed fleets. So that's probably what i'll be focusing on in the near future, moving forward, in this thread.

If things go well i think i should be ready to post in the near future at least some rudimentary stuff about beams vs fighters (really need better terminology to avoid the overlap between fighter as interceptor/fighter/bomber vs fighter-fighters) and some rudimentary stuff about non-shielded beam ships vs shielded beam ships.

Ophiuchus
Programmer
Posts: 966
Joined: Tue Sep 30, 2014 10:01 am
Location: Wall IV

Re: Measuring efficiency of ship designs

#45 Post by Ophiuchus » Mon Sep 16, 2019 8:21 am

alleryn wrote:
Sun Sep 15, 2019 11:06 pm
Ophiuchus wrote:
Sat Sep 14, 2019 9:24 pm
The fastest way to do this is to change the c++ backend to provide somewhere an interface where you put in the combat universe and let it process the combat and gives back the combat result (surviving ships and values) and a combat report.
That doesn't sound too difficult. I suppose the toughest part would be figuring out how to input the fleets to test.
The structure is not really supporting this refactoring so i would say medium difficulty for changing the backend.
And then a lot of easy work in order to provide API for the input part(s).
Then some work to make AI use the simulator.
Then a lot of UI work if we make the simulator available to the player inside freeorion (On fleet "Add/Remove fleet to/from battle simulation"; on design window "Add ships of current design to battle simulation", On map/system "Start battle simulation here". Also either "Remove all ships from battle simulation" or "Edit battle simulation" maybe with save/load simulation).
I would love some text based network API (http/JSON) where you drop in a list of fleets and planets, maybe ship designs (else ships should contain their contents), maybe a location/system description and get back the battle report and survivors output (probably this should be async hmm).
Maybe in the end API to put in designs, locations, and fleets with design and location references, and then one API in order to start the battle which returns the battle result if ready, else a 503 or 302 with Retry-After header.
Ok. Way too much stuff. I should stop in this thread.
alleryn wrote:
Sun Sep 15, 2019 11:06 pm
I want to suggest a general framework about how we approach the question of efficiency of ship designs.
..
If things go well i think i should be ready to post in the near future at least some rudimentary stuff about beams vs fighters (really need better terminology to avoid the overlap between fighter as interceptor/fighter/bomber vs fighter-fighters) and some rudimentary stuff about non-shielded beam ships vs shielded beam ships.
Agree 100% with your framework.

You can use the "space boat" term I use or the "drone" term which Oberlus uses if you like.
I dont think those will become part of the in-game terminology but for the discussion they are quite fine.
Any code or patches in anything posted here is released under the CC and GPL licences in use for the FO project.

Furthermore, I propse... we should default to four combat rounds instead of three ...for the good of playerkind.

Post Reply