View unanswered posts | View active topics It is currently Mon Dec 09, 2019 12:00 am



Reply to topic  [ 44 posts ]  Go to page Previous  1, 2, 3, 4, 5  Next
Tournament desings up to 4000P 
Author Message
User avatar

Joined: Sat Jun 06, 2015 3:04 am
Posts: 457
Reply with quote
Quote:
though it sometimes loses...


There seems to be more variance with fleet battles. Which is why testing has been such a b****.
I'm pretty sure if I ran that tournament again, the top ships would switch places.

Also, fleet size seems to make a difference. When I did the 'one fleet to rule them all' article, I was thinking one good design when multiplied would still be good. However, scale seems to change the dynamics. I've had different designs win-out against other fleets, even though they lost to those same ship designs when the fleet size was smaller.

_________________
https://discord.gg/BM4eeqS
http://reassemblyerect.site


Wed Sep 23, 2015 3:13 pm
Profile
User avatar

Joined: Mon Aug 17, 2015 4:44 pm
Posts: 669
Reply with quote
I've had suspicions that there is an apex point at which adding more ships simply doesn't improve effectiveness. Due to the demands of geometry, space, and the reality of taking incoming fire I think there's a point at which more small ships just simply no longer adds appreciable value. For instance, in my mind, 60,000 P worth of Blizzballs would, I suspect, have difficulty against 60,000 P worth of ships with Obliterators, as the Blizzballs would end up getting cut in half pretty rapidly overall, and if the Obliterator ships had some decent armor they could easily shrug off incidental damage. I could be wrong, of course, but it's a thought.

Also, I've been running around in my head about the idea of building purposely slow ships, but tournament results have suggested that this would be a bad idea. Purposely slow ships would be forced to stick together, though, and would make it easier to concentrate fire, distribute incoming damage from swarms, and overall seems like a pretty neat idea. At the same time, being able to avoid a Tinkrell fire hose is worth every point you've invested in thrusters. Just something I've been thinking about.

_________________
NMSS Reassembly Tournament Archive

We have a Discord, too


Wed Sep 23, 2015 7:47 pm
Profile
User avatar

Joined: Sat Jul 12, 2014 3:36 pm
Posts: 326
Reply with quote
MonsPubis wrote:
I've had suspicions that there is an apex point at which adding more ships simply doesn't improve effectiveness.

Personally I think it's more down to the ratio between fleet sizes, the more ships you have the higher chances of the enemy getting distracted and not shooting at the ship(s) it should be.
Some ships though do become more/less effective when you change the number that are deployed by merit of the mechanics of the weapons they use though.

I think that one of the key advantages of smaller ships over larger ones is that they only have to move say 3/4 block lengths to completely avoid an incoming shot, while a larger ship has to move a much greater distance by merit of its increased size, meaning that in order to dodge the same weapon a larger ship needs to have better acceleration than a smaller one would.

MonsPubis wrote:
Also, I've been running around in my head about the idea of building purposely slow ships,

This would only work if they are armed with a longer range weapon than their enemy, as otherwise the enemy could just stay out of range and pelt them with impunity, otherwise it could be interesting to consider.


Wed Sep 23, 2015 10:23 pm
Profile
User avatar

Joined: Fri Jan 09, 2015 11:34 am
Posts: 19
Reply with quote
Thanks guys for a new ideas :-) I hade for you one oliberator fleet. It can lose against f4 snipers, or nuke ones, but is extremely effective against smaller targets.


Attachments:
LK_OliCruiser.lua.gz [2.12 KiB]
Downloaded 150 times

_________________
Lord Kikin
Thu Sep 24, 2015 11:37 am
Profile
User avatar

Joined: Sat Jun 06, 2015 3:04 am
Posts: 457
Reply with quote
Quote:
Camo5: I'd like to note that the fleet I entered has beaten every fleet I posted 3 times in a row, though it sometimes loses...


For the record, I'm having a terrible time beating one of your test designs.
My last test run it went 22-0. ... so yeah, none of my designs can get a single win.

Razz'in frazz'in! I say, razz'in frazz'in!

:)

_________________
https://discord.gg/BM4eeqS
http://reassemblyerect.site


Thu Sep 24, 2015 5:38 pm
Profile
User avatar

Joined: Fri Jan 24, 2014 10:11 pm
Posts: 1295
Location: Ames, IA
Reply with quote
Posting my 35k spinner fleet here, as well as the crystalline fleet I fought and somehow lost to on deluk's game...


Attachments:

_________________
The game where I originated as a member of this community. . VISUAL BASED MOD KIT!!
Shellcore command EP2
Thu Sep 24, 2015 7:16 pm
Profile
User avatar

Joined: Sun Sep 06, 2015 2:31 pm
Posts: 34
Location: UK
Reply with quote
Camo5 wrote:
Posting my 35k spinner fleet here, as well as the crystalline fleet I fought and somehow lost to on deluk's game...
Not sure why you tucked your post into this thread, but here's copy of my reply posted on the Deluks YouTube Video comments:
Quote:
So, downloaded and done some experimenting. For me Camo's Ultimate fleet is the nominal winner, with about 50 ships to spare, but... I get a lot of variation in who win under various conditions: For a start, turning the physics right down (to minimum of 15/s) massively favours Lucky's missile swarm (winning with almost 200 ships left). Higher physics (e.g. max of 120) doesn't seem to benefit Camo's spinners any more. I'd conclude that the cross-sectional (collision) co-efficient between the slightly larger spinner ships and (on average) slightly larger (and slower moving) crystalline missiles is less sensitive to bigger time steps than the smaller, slower moving Lucky ships and smaller, faster moving Terran missiles. (Nuclear physics is all about these kind of variables.)

However, I'm pretty sure +Deluks Gaming didn't have his setting low, because that also shows up when you try to pan the cam, making it choppy and slower (also, he has great moral fortitude and never makes mistakes. ;o)

Smaller screen resolution (windowed) also screw things around, with a marginal Swarm win over spinners, for example. View zoom might even have an effect, although I'm not prepared to say I've found anything definitive here. Not sure speeding/slowing time has any particular effect. Lower res certainly reduces the stated amount of video memory used (in settings screen). And zooming out to the whole arena halved my frame rate down to ~30 (compared to a more limited, zoomed in field of view, at the full 60fps), as shown by fraps.

What setting do you each have frame rate set to: "60FPS cap, No limiter, Vsync, Adaptive Vsync"? Adaptive default? I think the only solution is for all future tournaments to be run in 'headless' mode...... ;o) Or maybe someone with an inside line should fire off a question to Arthur, see if he can shine a light?

Side note: is it my imagination, or does the first letter of the fleet name determine left/right start position: earlier in alphabet on the right? There might be subtle symmetry breaking in the simulation to account for too...?

_________________
Image
Youtube Channel (Terraria Wiring mostly) | T-MEC - My Terraria Sub-forum | Blog (sci/tech/culture).


Fri Sep 25, 2015 2:06 am
Profile
User avatar

Joined: Mon Aug 17, 2015 4:44 pm
Posts: 669
Reply with quote
ZeroGravitas wrote:
So, downloaded and done some experimenting. For me Camo's Ultimate fleet is the nominal winner, with about 50 ships to spare, but... I get a lot of variation in who win under various conditions: For a start, turning the physics right down (to minimum of 15/s) massively favours Lucky's missile swarm (winning with almost 200 ships left). Higher physics (e.g. max of 120) doesn't seem to benefit Camo's spinners any more. I'd conclude that the cross-sectional (collision) co-efficient between the slightly larger spinner ships and (on average) slightly larger (and slower moving) crystalline missiles is less sensitive to bigger time steps than the smaller, slower moving Lucky ships and smaller, faster moving Terran missiles. (Nuclear physics is all about these kind of variables.)

However, I'm pretty sure +Deluks Gaming didn't have his setting low, because that also shows up when you try to pan the cam, making it choppy and slower (also, he has great moral fortitude and never makes mistakes. ;o)


It's due to the way spinners function. If there are 15 frames per second of simulation and the object is moving at 3 rotations per second, it's traveling 72 degrees per frame of simulation. If it only has a 12 degree window of opportunity to fire and hit the target, it likely will not fire at all. Ever. At the same time, if the simulation was running at 180 frames per second it would travel 6 degrees per frame. It will fire every time the weapon is able, more or less. This is why spinners see such a sharp drop in performance as physics simulation fidelity is lowered (and why Camo saw his fleet so much more effective when he tested it than when I did in my tournament -- he used 120, I use 60). For reference, I also intend to use 60 FPS as my standard frame rate in the future as well.

ZeroGravitas wrote:
Smaller screen resolution (windowed) also screw things around, with a marginal Swarm win over spinners, for example. View zoom might even have an effect, although I'm not prepared to say I've found anything definitive here. Not sure speeding/slowing time has any particular effect. Lower res certainly reduces the stated amount of video memory used (in settings screen). And zooming out to the whole arena halved my frame rate down to ~30 (compared to a more limited, zoomed in field of view, at the full 60fps), as shown by fraps.

What setting do you each have frame rate set to: "60FPS cap, No limiter, Vsync, Adaptive Vsync"? Adaptive default? I think the only solution is for all future tournaments to be run in 'headless' mode...... ;o) Or maybe someone with an inside line should fire off a question to Arthur, see if he can shine a light?


Engine/Simulation/Physics FPS and graphics FPS are not linked. They can be running at completely separate rates. If there are issues related to whether or not something happens 'on screen', as would be suggested by the resolution producing different results, it would lead me to believe the game approximates details in the background, but there's not much reason to do that and I'm inclined to believe it just fully simulates everything, since it's all one sector of game space anyway. The game also tends to pseudo randomly arrange ships in the starting area for each fleet, which can cause pretty dramatic variation sometimes as well.

ZeroGravitas wrote:
Side note: is it my imagination, or does the first letter of the fleet name determine left/right start position: earlier in alphabet on the right? There might be subtle symmetry breaking in the simulation to account for too...?


Fleet starting position is chosen by position in the pool or bracket. The 'active' fleet in a pool will always be on the left. I'm not certain about bracket, but I'd assume something similar.

_________________
NMSS Reassembly Tournament Archive

We have a Discord, too


Fri Sep 25, 2015 4:03 am
Profile
User avatar

Joined: Sun Sep 06, 2015 2:31 pm
Posts: 34
Location: UK
Reply with quote
MonsPubis wrote:
If there are 15 frames per second of simulation and the object is moving at 3 rotations per second, it's traveling 72 degrees per frame of simulation. If it only has a 12 degree window of opportunity to fire and hit the target, it likely will not fire at all. Ever.
Except that this fleet of Camo's primarily uses mines and homing missiles (+ some non-homing torpedos? I'm least familiar with Terrans), which will get shot regardless of orientation, right? More to the point, it shouldn't matter how far the aim rotates per game tick, presuming it targets on a tick-by-tick basis (either fire or don't fire). Provided it doesn't get phase locked to an exact multiple such that it only ever points in exactly 2 or 3 discrete directions, it will still be pointing in every possible direction, but just in a staggered, non-sequential order. There will have to be a fair sized margin of error programed in anyway, since the slightest of movement would prevent perfect shot alignment on a somewhat distant command module, anyway...?
MonsPubis wrote:
he used 120, I use 60

I'm going to assume that Camo's a pretty smart chap and will have matched the stated fidelity for Deluks (although I know nothing).
MonsPubis wrote:
Physics FPS and graphics FPS are not linked. They can be running at completely separate rates.
Sure. No direct link. I was thinking more of indirect via CPU load. But also, just as the camera pan movement is jerky in line with the simulation steps (not display FPS, for me anyway), there may be some other idiosyncrasies in there: something weird like some length scale fidelity being set by zoom level (well, I don't know what, but something). I'm definitely not saying it 'approximates' off-screen details. Also, a beauty of simulating swarm fleets is that variations in starting positions should average out more to have near negligible effects.

Edit: actually, thinking about timing fidelity - "physics" has exact settings, that may well be rigidly adhered to, but what about AI? I've noticed big groups of laser weapons (e.g. Obliterators) often fire in sync when it's vanishingly unlikely they should have done so. Does this mean the shooting decision making happens at a much slower time scale? Does it even get a little 'stuck' some times? Is this a factor that gets squeezed by a lack of CPU cycles? (Also, how much does this conversation need a seperate/different thread?!)

_________________
Image
Youtube Channel (Terraria Wiring mostly) | T-MEC - My Terraria Sub-forum | Blog (sci/tech/culture).


Fri Sep 25, 2015 5:07 am
Profile
User avatar

Joined: Mon Aug 17, 2015 4:44 pm
Posts: 669
Reply with quote
ZeroGravitas wrote:
Except that this fleet of Camo's primarily uses mines and homing missiles (+ some non-homing torpedos? I'm least familiar with Terrans), which will get shot regardless of orientation, right? More to the point, it shouldn't matter how far the aim rotates per game tick, presuming it targets on a tick-by-tick basis (either fire or don't fire). Provided it doesn't get phase locked to an exact multiple such that it only ever points in exactly 2 or 3 discrete directions, it will still be pointing in every possible direction, but just in a staggered, non-sequential order. There will have to be a fair sized margin of error programed in anyway, since the slightest of movement would prevent perfect shot alignment on a somewhat distant command module, anyway...?


A 12 degree margin is pretty large, I think, but I'm not sure what the actual value is. I expect Arthur tested it pretty extensively. I was just explaining that for anyone who may not have understood how that works. Getting frame locked is certainly possible, and I suspect you can see it sometimes when the ship appears stationary and all its thruster puffs travel in one direction.

I'm not sure how Mine and homing missile AI work. I think homing missiles and mines will launch anyway, but I haven't really tested. Thinking about it more, I do think you're onto something with collision detection too. I suspect if a projectile is moving fast enough to travel completely through a ship in a single frame, it may pass harmlessly through it. This could also lead to some other shenanigans, I bet, like ships getting stuck outside the arena (something else I've witnessed, but only in extreme circumstances).

ZeroGravitas wrote:
I'm going to assume that Camo's a pretty smart chap and will have matched the stated fidelity for Deluks (although I know nothing).


Maybe. I didn't even think about mentioning physics FPS until after the tournament had concluded it and it was pointed out that some simulations were producing results very, very different from mine with a high degree of consistency. I don't know if Deluks specifies his.

ZeroGravitas wrote:
Sure. No direct link. I was thinking more of indirect via CPU load. But also, just as the camera pan movement is jerky in line with the simulation steps (not display FPS, for me anyway), there may be some other idiosyncrasies in there: something weird like some length scale fidelity being set by zoom level (well, I don't know what, but something). I'm definitely not saying it 'approximates' off-screen details. Also, a beauty of simulating swarm fleets is that variations in starting positions should average out more to have near negligible effects.


The simulation should simply slow down if the CPU load is too great. You see it all the time in the titan tournament Deluks ran. It may step down to 10 engine FPS, but still calculate as though it were at 60, which would lead to 1:6 time dilation. If that makes any sense.

If we are seeing a lot of variation in results using swarms of ships, I suspect there is something amiss but what it is is hard to identify, exactly. I haven't run a lot of tests myself, but in my own tests results are generally pretty consistent (Fleet A beats Fleet B every match though sometimes individual fights Fleet B will win by a hair or by points, for example).

The fleets you were testing with were Camo's Ultimate 35K and the torpedotest that were posted in this thread a little bit ago?

EDIT: Just tested Torpedospamtest and Ultimate 35K in four 1v1 tournaments against each other. Each of them came away with two wins in 2-0 matches. Every time the fleet on the left was the victorious one. When Ultimate 35K started on the left, it always engulfed Torpedospamtest and disintegrated it from multiple sides. When Ultimate 35K started on the right, it mostly smashed into Torpedospamtest and got slaughtered wholesale for doing so. It's a small sample pool, but I am seeing something of a trend here. Tested at 60 Physics FPS and had no performance issues to speak of.

_________________
NMSS Reassembly Tournament Archive

We have a Discord, too


Fri Sep 25, 2015 5:21 am
Profile
Display posts from previous:  Sort by  
Reply to topic   [ 44 posts ]  Go to page Previous  1, 2, 3, 4, 5  Next

Who is online

Users browsing this forum: No registered users and 1 guest


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
cron
Powered by phpBB® Forum Software © phpBB Group
Designed by ST Software