Post Reply 
 
Thread Rating:
  • 0 Votes - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Possible bug: strange asymmetry in battle
03-24-2013, 12:55 AM
Post: #1
Possible bug: strange asymmetry in battle
Hi rocco et al,

I found a strange asymmetry in the Pass-and-play Test Game scenario of version 1.2.0.b1.

Create fleets like this for both Player1 and Player2:

6x Mammoth 12Ion3 3Armor2 2Bulk2 Shields4 3ECM Engines

Send both fleets into battle, with orders "Charge Close, Target Weakest".

The result? Player 2 wins with 5 ships surviving, 71% health! How is that lopsided outcome possible, given the same fleets with the same orders?!?

Am I losing my mind? I ran it 4 times, got the same result each time. Can one of the other Beta Testers try to reproduce this battle?

Bug report submitted.
Find all posts by this user
Quote this message in a reply
03-24-2013, 04:12 AM
Post: #2
RE: Possible bug: strange asymmetry in battle
(03-24-2013 12:55 AM)MarcusVictor Wrote:  Hi rocco et al,

I found a strange asymmetry in the Pass-and-play Test Game scenario of version 1.2.0.b1.

Create fleets like this for both Player1 and Player2:

6x Mammoth 12Ion3 3Armor2 2Bulk2 Shields4 3ECM Engines

Send both fleets into battle, with orders "Charge Close, Target Weakest".

The result? Player 2 wins with 5 ships surviving, 71% health! How is that lopsided outcome possible, given the same fleets with the same orders?!?

Am I losing my mind? I ran it 4 times, got the same result each time. Can one of the other Beta Testers try to reproduce this battle?

Bug report submitted.

I got the same, but it's not from a bug, just from the ships following different paths resulting in uneven fights. In mine, the blue ships all stick together while the red end up spread out more after the first pass, leaving it weighted in blue's favor. I tried with primary targets set on both side and it's a far closer match. It's gotten down to 2 blue ships with 14% total fleet health remaining vs 1 red ship with 12% total fleet health remaining. Trying to get it to finish now, they got to the point that they just kept circling each other and couldn't get into range cause their turning radius is too large.

GC ID - VanderLegion, GMT-9. Sandbox GC ID (Beta) - VanderLegion
Find all posts by this user
Quote this message in a reply
03-24-2013, 06:01 AM (This post was last modified: 03-24-2013 06:02 AM by rocco.)
Post: #3
RE: Possible bug: strange asymmetry in battle
@VL is likely correct; Ion and Gauss weapons are less forgiving than others when it comes to positioning and distance. Running a similar test with a longer range weapon like lasers or plasmas would likely produce closer results.
Find all posts by this user
Quote this message in a reply
03-24-2013, 01:27 PM
Post: #4
RE: Possible bug: strange asymmetry in battle
Thanks for both your responses, and relieved to see it's reproducible.

[rant]
It may not be a BUG per se, but isn't it a HUGE PROBLEM? I suspect many players, like I, spend a fair amount of time thinking about the best fleet design and combat orders, discussing these strategies on the forums, and play testing them against each other. I'd like to think that the time so "invested" translates into some kind of edge in actual games. To find such an lopsided outcome in matched fleets is possible because of blind luck, is demoralizing.

Actually, it's not even blind luck. Imagine an alternative combat resolution mechanism, where we just put 6 red balls and 6 blue balls in an urn, and drew one at a time to find out the next ship to die. The probability that you'd have 6 red vs 1 blue is ... 1/132, less than 1%. It's too improbable to be blink luck.

Ship vs Ship combat is the very core of Starbase - it seems like there is still some work that could be done here. This should be highest priority!
[/rant]

I do see that the blue and red ships follow different paths - one zigs where the other zags. Are these decisions determined by calls to a random number generator? If so, why not have each fleet have it's own instance of a random number generator, initialized with the same seed? Then the asymmetry would be gone.

On rocco's suggestion I ran the same test with Gauss/Laser Mammoths:

6x Mammoth 10Gauss3 2Laser3 3Armor2 2Bulk2 Shields4 3ECM Engines, orders="Closest, weakest"

Result is Blue wins, 3 ships left, 47% health. So still asymmetrical, but not quite as severe. Blue wins again - I'm definitely choosing Blue for my empire color next game. Wink

I also confirm that if both sides chose primary targets, the battle outcome is closer - but unlike Vander I had Red win with 2 ships, 20% health left. It seems to depend in some unpredictable way on which ship you chose to primary target. Remember, all the ships are identical - should primary target really matter so much?

I'd like to hear other emperor's opinions...
Find all posts by this user
Quote this message in a reply
03-24-2013, 03:27 PM
Post: #5
RE: Possible bug: strange asymmetry in battle
(03-24-2013 01:27 PM)MarcusVictor Wrote:  I also confirm that if both sides chose primary targets, the battle outcome is closer - but unlike Vander I had Red win with 2 ships, 20% health left. It seems to depend in some unpredictable way on which ship you chose to primary target. Remember, all the ships are identical - should primary target really matter so much?

I'd like to hear other emperor's opinions...

It's not ship differences that make a difference with PT here. It's how your ships path based on what PT you picked. If you pick one on the edge, your ships all aim for that one, then when they're turning back around they do so based on where that ship is. If you pick one in the middle, same thing. For the test I did with PT, I used ships in the middle on both sides as the initial PT.

GC ID - VanderLegion, GMT-9. Sandbox GC ID (Beta) - VanderLegion
Find all posts by this user
Quote this message in a reply
03-24-2013, 04:13 PM
Post: #6
RE: Possible bug: strange asymmetry in battle
Quote: Am I losing my mind? I ran it 4 times, got the same result each time. Can one of the other Beta Testers try to reproduce this battle?

Keep in mind, if you "run it 4 times" in the battle sim you should make sure you do it on different turns, and not just replay the same turn. All turns now use the same PRNG seed, so only a different turn would give a different result.



With that in mind, I tried to reproduce it here. I did two battles (ran them on separate turns), with the equipment mentioned (I assumed 3ECM engines was actually 3 engines). In the non nebula system my results were 16% and 27% still alive (favor of blue). In the nebula system is was all red dead, and 30% blue.

I also ran the test I actually suggested, which was to use a long range weapon (in my case I did LS3), as the long range weapons are less susceptible to positioning that the short range weapons. The results were 26% / 30% (blue's favor).



I do admit that it is quite odd that blue continuously seems to have the upper hand. I will investigate as to why that may be.
Find all posts by this user
Quote this message in a reply
03-24-2013, 06:03 PM (This post was last modified: 03-24-2013 06:21 PM by MarcusVictor.)
Post: #7
RE: Possible bug: strange asymmetry in battle
(03-24-2013 03:27 PM)VanderLegion Wrote:  
(03-24-2013 01:27 PM)MarcusVictor Wrote:  I also confirm that if both sides chose primary targets, the battle outcome is closer - but unlike Vander I had Red win with 2 ships, 20% health left. It seems to depend in some unpredictable way on which ship you chose to primary target. Remember, all the ships are identical - should primary target really matter so much?

I'd like to hear other emperor's opinions...

It's not ship differences that make a difference with PT here. It's how your ships path based on what PT you picked. If you pick one on the edge, your ships all aim for that one, then when they're turning back around they do so based on where that ship is. If you pick one in the middle, same thing. For the test I did with PT, I used ships in the middle on both sides as the initial PT.

I DO understand it is simply a function of the course the ships take that determined the outcome. When I showed the original battle to my Starbase-playing 9 year old, he predicted a tie (logically enough). His analysis after seeing blue's massacre of red's fleet, was that red's ships were "being stupid".

In the test with Primary Target I mentioned above, I had chosen the first ship (top ship) as the target for both. I reran, choosing the third ship from the top as Primary Target for both, and got blue winning, with 2 ships, 20% left.
(03-24-2013 04:13 PM)rocco Wrote:  
Quote: Am I losing my mind? I ran it 4 times, got the same result each time. Can one of the other Beta Testers try to reproduce this battle?

Keep in mind, if you "run it 4 times" in the battle sim you should make sure you do it on different turns, and not just replay the same turn. All turns now use the same PRNG seed, so only a different turn would give a different result.

I did run it on two different turns (on 400.4, and on 432.6), figuring that the random number generator would be in a different state, and hence would alter the outcome, but it didn't. There seems to be something deeper going on.

Quote:

With that in mind, I tried to reproduce it here. I did two battles (ran them on separate turns), with the equipment mentioned (I assumed 3ECM engines was actually 3 engines). In the non nebula system my results were 16% and 27% still alive (favor of blue). In the nebula system is was all red dead, and 30% blue.

No, they were 3 Electronic Counter Measures, 1 Combat Engines. I was trying to be "realistic", in the sense that the Ion fleet player needed to consider the possibility that he'd be facing a Proton fleet.

Quote:I also ran the test I actually suggested, which was to use a long range weapon (in my case I did LS3), as the long range weapons are less susceptible to positioning that the short range weapons. The results were 26% / 30% (blue's favor).



I do admit that it is quite odd that blue continuously seems to have the upper hand. I will investigate as to why that may be.

It seems like a hard problem... but I think it's worth solving. Best of luck, let me (us) know if you want us to Beta test.

Can I make a suggestion that might help reduce the randomness factor with Ion specifically? Get rid of the double damage/half rate of fire feature of Ion. I think it increases the variation in the random outcome of Ion vs Ion battles - because, by pseudo-random luck, your ships might end up very close to opponent's ships when it's your turn to fire, and it's not opponent's turn. The feature was put in a long time ago (1.1.4??). I think the reason was that it was thought that it would encourage larger ships. I don't think that logic works. It messes up the Weapons Damage chart - very misleading for newbies(not documented in the Ship Designer description), adding complexity for no good reason. Begone with it.
Find all posts by this user
Quote this message in a reply
03-25-2013, 03:12 AM
Post: #8
RE: Possible bug: strange asymmetry in battle
I did some test runs too.

I used 3 combat engines, 2 ECM, 2 HA2, 2 BH2 for IPC. Charge close, weakest. Blue won handily. Tried again, using charge close, weakest for Blue, charge close, closest for Red. Blue won, but Red was a lot closer.

Then tried IPC mammoth vs Plasma mammoth. Both identical systems (same as above). Plasma was keep long, closest, IPC was charge close, weakest. Plasma Blue won with five survivors. IPC Blue won also (!?). That didn't make any sense.

At this point, I am thinking maybe blue has combat engines or something. I try the old standard: two scouts. Both die, and no evidence of combat engines.

So I tried 6 LS3 mammoth. Identical. Both hold together, weakest. Red won (!) with one ship left, only 4 hitpoints remaining. Remarkable, really.

BTW: I think "hold together" would be better if it was "stay in line" regardless of whether there is a starbase or not.

The only factor that I think might explain why red was losing (and it didn't matter if they went first, or what turn it was on) was that I assume that all of the crew wore REDSHIRTS (!). Everyone knows that means instant death. Rocco, take the code out that kills redshirts. I know, I know... classic and untouchable truth of science fiction lore, but really now.
Find all posts by this user
Quote this message in a reply
03-25-2013, 06:18 AM
Post: #9
RE: Possible bug: strange asymmetry in battle
Ok, so this intrigued me enough that I spent most of the day looking at it. As I expected it did boil down to better positioning, but it was due to a logic flaw in the combat processing code that resulted in one side consistently being able to get the better positioning.

Here is how combat is currently implemented:

Firstly, know that under the hood there is a full grid-based, turn-based combat system in place. Each ship is assigned a number of movement points each combat turn (based on ship size and other factors). Each ship is also assigned an initiative value (based on their movements points). Each combat turn has several phases, the first phase is where all ships get to move (ships don't fire until all movement is done). The ships are sorted by their initiative, then each ship (in order) gets to expend all of their movement points to move themselves wherever they need to.

In this case of these "mirror matches", all ships have the same initiative so the ships array contained all ships of one empire, followed by all ships of the second empire. Which meant that all ships of the second empire moved after all ships of the first empire, causing them to be more reactive and position themselves better.

Note that this issue is most noticeable in these test matches where all ships have the same initiative.


Here is how I have changed it to fix the issue:

Firstly, I changed it such that ships do not get to use all of their movement points all at once. Instead, each ship in combat gets to use ONE movement point at a time. This allows ship movement to be more simultaneous, and allows a little better positioning across the board.

The second change is that each time through the loop for all ships to use one movement point, the array is randomized. So even at the one move at a time level it is mixed up such that all of one side does not get to always move all their ships before the other.

The second change means that the concept of "initiative" is out the window. This should be fine as now the higher MP ships will be able to spend the rest of their MP after the slower ships run out; allowing them to finer position themselves in relation to the slower ships.

My empirical tests with this specific scenario went very well; all of the battles were much closer (to within 1-3 ships of each other), and there was a good mixture of red wins along with the blue wins.



My gut tells me that this is better logic overall, but the change is about as low level as they come so there may be unexpected consequences. I will rely on the wonderful beta group to help test it more out!
Find all posts by this user
Quote this message in a reply
03-25-2013, 07:38 AM (This post was last modified: 03-25-2013 07:40 AM by VanderLegion.)
Post: #10
RE: Possible bug: strange asymmetry in battle
(03-24-2013 06:03 PM)MarcusVictor Wrote:  No, they were 3 Electronic Counter Measures, 1 Combat Engines. I was trying to be "realistic", in the sense that the Ion fleet player needed to consider the possibility that he'd be facing a Proton fleet.

Actually, if you want an at all "realistic" gauss fleet, you need way more than 1 engine so you can actually get in range to do damage. You drop some armor/bullkheads for the ECM (and no shields)

As for the other stuff, it all make ssense now rocco. Change sounds good Smile

GC ID - VanderLegion, GMT-9. Sandbox GC ID (Beta) - VanderLegion
Find all posts by this user
Quote this message in a reply
Post Reply 


Forum Jump:


User(s) browsing this thread: 1 Guest(s)

Contact Us | Starbase Orion | Return to Top | Return to Content | Lite (Archive) Mode | RSS Syndication