These turn times are killer

[unmerged(400937)]

Actually abstraction means a close approximation to a full simulation. So inaccuracies are not just side effects, but the whole point. Stormers are a problem but the distance thing can be easily fixed as well as damage over time stuff...

Personally i'd rather take inaccuracies than a 30m turn - because i am not gonna wait for that turn to compute, i am going to play something else, which is by far the best abstraction ;P

i didn't say an abstraction was impossible, but SOTS' variety means that it will always be susceptible to mistakes, sometimes quite massive ones, in odd combinations of circumstance.

let me tell you about a fight i had in SOTS 1. I had six fission engined Liir UV and plasma cannon cruisers hanging around at a world, a relic of a defense fleet thrown together to defend against a silicoid queen. they'd been there for a good fifty turns when the Puppetmaster showed up.

Okay, I thought, there's no point in fighting this, but this is gonna be exciting, never met the Puppetmaster before, I'll start putting together a fleet group to defeat him!

so I click autoresolve, to let them die so I can get on with it.

one of them does indeed die. SO DOES THE PUPPETMASTER.

there were no stormers, no PD issues, no range issues, no maneuverability issues, none of the normal things that people say autoresolve doesn't take into account; just a dead Puppetmaster and some proud UV laser gunners who had no explanation.

I have never since seen the Puppetmaster.


That's an abstraction for you; no matter how close you get it to the real simulation (and getting it close to the real simulation requires getting close to the processing power and time needed by the real simulation) there will always be odd errors.

I'd rather wait for a reasonable simulation; I like it when the AI isn't custom-built to account for the weirdness of autoresolve (as in SOTS 1, where it's forced to alter its weapon choice if it's mostly fighting autoresolve battles), when the AI can fight combats amongst itself with a reasonable chance of realistic results, when I can use autoresolve for any combat I could fight manually and expect the same results.

in short, I prefer delaying my turns than, well, SOTS 1 autoresolve.

 

[eRe4s3r]

So we agree to disagree then, for me SOTS2 is unplayable if these turn times remain (and grow) like this ;( Even if combat results are fancy accurate, in reality 99% of those battles have no effect on me whatsoever. A good abstracted simulation has to be measured over say, 100 combats - if on average the results are close to player battle then we have a winner. Its clear that is not the case in SOTS1 however... so i don't want that abstraction back, i want a better one. But i do want one.

 

[Myrrdin]

The big thing is though, that the SotS AI constantly redesigned their ships to counter whatever it was up against by analyzing its combat results, and I'm sure the same will be true for the finished SotS 2 AI. Thus, an AI that mainly had fought a player in actual combats would often make a frightfully competent showing when it came to ship design. Conversely, an AI mostly having fought other AIs using autoresolve, often put up laughable resistance at first and seldom managed to change their ship designs before getting totally overwhelmed. Something that almost negates a really neat feature like this, sticks out like a sore thumb. I am sure you could make a quick simulation a little more accurate, but inconsistencies would add upp as far as the AI design goes. Given that getting new designs and fleets full with them up in SotS 2 is a decidedly slower process, this would essentially neuter this aspect of the AI.

 

[unmerged(400937)]

So we agree to disagree then, for me SOTS2 is unplayable if these turn times remain (and grow) like this ;( Even if combat results are fancy accurate, in reality 99% of those battles have no effect on me whatsoever. A good abstracted simulation has to be measured over say, 100 combats - if on average the results are close to player battle then we have a winner. Its clear that is not the case in SOTS1 however... so i don't want that abstraction back, i want a better one. But i do want one.

they DO have an effect on you, since if they have an effect on ANYONE, they will eventually have an effect on you. you may not notice it, but they still do, such as the one Myrddin just noted; that you can murder an AI empire in SOTS 1 before they change their weaponry.

the problem with averages is that they don't make any sense here; combat results are more than just "who won", even before you start to consider that you could have one of those hundred combats involve one battlerider destroying fourteen full Leviathan led fleets and it wouldn't skew the average more than a little bit.

so you'd have to average each weapon damage, each ship type, each model and possible design, each fancy section, each tactic, which'd be a massive amount of work; and getting it close to accurate, getting it to the extent where the simulation is an effective duplicate of the manual combat results REQUIRES THE SAME PROCESSING POWER.

if there was a more efficient way to produce an accurate simulation of manual combat than manual combat, we'd use that for manual combat. you have a substantial gap between manual combat and your simulation if you want to have a simulation that requires any appreciable amount less processing power.

 

[eRe4s3r]

Well its the job of the coder to make sure that the average doesn't involve SUCH drastic outliers

Basically you are saying: "It takes effort to make a good abstraction" and I agree. But reality is that unless Kerberus puts this effort into an (potentially optional) feature then SOTS2 will become unplayable the longer you play. And with unplayable I literally mean "time where you don't play"

At some point, SOTS2 takes longer to calculate a turn than you needed to do everything IN that turn. And thats the point where I hit the delete button and start a new game (or drop the game)

 

[unmerged(400937)]

Well its the job of the coder to make sure that the average doesn't involve SUCH drastic outliers

Basically you are saying: "It takes effort to make a good abstraction" and I agree.

no, what I'm saying is "there is no good abstraction". there will always be outliers because SOTS is complicated and abstractions are flawed. the need to do it with DRASTICALLY less computing power if you want it done quicker and without interfering with you means it will HAVE to be critically flawed.

But reality is that unless Kerberus puts this effort into an (potentially optional) feature then SOTS2 will become unplayable the longer you play. And with unplayable I literally mean "time where you don't play"

At some point, SOTS2 takes longer to calculate a turn than you needed to do everything IN that turn. And thats the point where I hit the delete button and start a new game (or drop the game)

I play multiplayer. I have to wait half an hour sometimes for a turn to be finished by other players, so it's not a massive dealbreaker for me. and that's just in games with my roommate - he's a Hiver player who insists on checking every single thing, every single turn. i've got used to keeping a book or a forum or a comic or just alt-tabbing to the internet.

that said, I will miss what I used to do in SOTS 1, which was play Neverwinter Nights 2 while I waited for him to click end turn.


EDIT: oh, probably worth mentioning; turn time delays are definitely gonna go down, there's currently an artificial delay inserted to prevent a bug, and speed of the game in general is likely to increase over subsequent patches. can't promise it'll get good enough for you, but I can say you should wait to see before making an assessment of the turn delays.

 

[Myrrdin]

I would be surprised if the current turn times won't get some of the time shaved off, just don't have any idea of how much. It can be noted that even today, as an example, Civ 4 turn times late game on a huge map and with a full complement of AIs is pretty bloody painful, no matter what kind of computer you throw at it. The solution, if you like the game and can't stand the turn times: Play smaller maps with less opponents.

 

[Ruanek]

If it's a large problem then Kerberos could make a battle calculator. Yeah, it wouldn't be perfect, but it's another alternative at least.

 

[eRe4s3r]

Well i am waiting for the game to even include all the features (like I have a choice ,p) so we'll see what way turn times go. But at least Kerberus knows that some people care...

 

[Tairneanach]

Well, lookee there:

- Some AI processing speed optimization.

From today's patch notes. I wonder how much of an impact this is going to make.

 

[namuseta]

That's besides the point, nothing really actually USES more than 4 at this point. It isn't needed, and we don't have anything outside of intense graphic programs or video design programs that require it.

It can be 16 recommended but that doesn't mean it can use it.

It's always recommended to have more than you need but the infrastructure of our programs does not use more than 4 for much.

i am using about 3 gigs of ram normally not even playing anything. when playing something it can be 5-7 gigs in use so having 8 gig nowdays is really usufull... i some times have 2 games running at the same time

going from 4 to 8 is a good idea what gives you alot of benifits.

 

[unmerged(402700)]

And that's an EXCEPTION which isn't normal for the architecture.

You don't often have 3dMAX and MAYA running at the same time and if you did then yes you need 16 gigs but if you're playing ONE GAME despite its intensity, it won't actually take more than 4 gigs except in rare circumstances where 6 is the cut off point.

If your computer is using 3 gigs by doing nothing then you have problems with your computer.

 

[ArmchairMagpie]

It seems the AI is using the same UI functionality albeit hidden as the player does. UI generally becomes more and more slower and less responsive as time goes on. Whatever is causing the latter has a cumulative effect on the former.